Kind: captions
Language: en
I think people suffer needlessly because
they confuse the distorted narrative
running inside of their minds with
actual reality but if I'm right and we
don't have access to the quote unquote
real world and instead we're just a
brain and a vat running a simulation
what do we need to understand about the
nature of this self-generated simulation
in order to live our lives well when we
talk about the brain and the vet of
course the vet here is as far as you
know our physical body that is much
through a physical universe and is
contains a skull full of cells that have
to get along and generate a model of
reality and of the individual itself and
its relationship to reality and the
purpose of that whole thing is to feed
all the C and to keep them um in play
not just for this generation but over
many many generations because we evolved
not for this particular moment in time
but for a longer course and in this long
long multigenerational course we adapted
to certain circumstances the basically
conditions under which our ancestors
live and uh in many ways our world has
changed in the last few Generations
rapidly away from the ways in which our
great par grandparents lived their lives
and these discontinuities are on one
hand extremely exciting and they force
us to adapt to new circumstances but
they also are very unusual from an
evolutionary perspective for the things
that we confronted with a lot of the
experience that we have alienating and
it's also very difficult to see in the
future and to see how long we'll be
around
as um whether our children or
grandchildren will still use any of the
thoughts that we have today and any of
the things that we build today for them
and this creates uh justifiedly anxiety
right if we live in a world that
possibly has no future for us or if
there's no future extension of us
present in the world it's unsettling and
I think it's a Justified experience to
have that when we start creating the
mental model we are not even born yet
and uh I said think that in ut already
we are starting to prepare a map of our
body and how it relates to our needs we
experience our first Pleasure and Pain
probably to some degree and um after we
are born we are focused on a world that
has an up and down Direction and uh
which light is playing out and we see
patterns in these in the light and then
Direct
and we are able to discern sounds and at
some point few months in we are able to
correlate the sound in the directions
and we also notice that the space that
you're observing is all the same space
it's the space in which that we can
touch and so at every moment we find
ourselves in a scene and this word model
the scene is something that our
ancestors discovered early on that we
generate this that our own Consciousness
creates this but uh when we start our
own memories at as a human being we're
typically Beyond this stage where we
create the world we are personal selves
at this point and when we try to
remember how our life started I think
what we remember are those moments where
this personal self is already online
because that's right now the vantage
point from which I assemble reality
right it's me remembering being a person
in this world that is being generated in
my mind so I'm no longer the
Consciousness that creates the world
around me in my own mind that runs a
simulation engine in my brain that
tracks visual data and perceptual data
but I am a being that cares about what
it sees in this world and can no longer
directly change my perception I also can
no longer directly change my emotion I'm
um basically exposed to my perception
and to my emotion and have to deal with
them and as I grow uh up I can build
models of the purposes that emerge over
my feelings and emotions and needs and
desires and fears and in this way I
create a model of the world that I'm in
and I discover I'm not alone in this
world but there are purposes that I
share with others and in this sense I
experience myself as a social being as a
state building creature that together
with others is not only U making plans
and is trying to change the world in a
particular way and specialized in the
hive but we also creating reality
together we have shared models of
reality that we exchange how close are
the models of reality that we have to
actual base reality well there are cor
strains models of based reality which
means that they are generalizations over
the perceptual data at the resolution in
time and space that our percepts have
which means for instance if our neurons
take um 20 milliseconds to pass on a
signal from one neuron to the next this
determines the temporal resolution in
which uh they can be excited and track
phenomena and then we are trying to
integrate a scene and is the question
how many neural junctures does the
signal need to pass until it computes
the necessary functions and so I observe
that my sense of now that I'm in can be
a fraction of a second and it can be
something like up to three seconds that
I perceive as one cohesive moment of now
and um the physical now of course is
different from this it's spread out much
further so there are things in my
perceptual now that are predictive that
are basically playing out in the future
and there are things in my perception
now that had played out in the past
signals that I just experienced now and
to integrate into my model despite a
signal from my feet needing hundreds of
milliseconds until it arrives in my
neocortex and then a few more 100
milliseconds until it's processed and
when it's surprising to be integrated in
my world model right so this subjective
now is a fiction and this subjective now
is constructed and in the physical world
there are no colors and no sounds there
are also no waves in the ocean right
there are molecules that exist in the
ocean that pull and push at each other
producing something that looks like
waves when you Zoom very far out and
there is no sound in the air there is
just air molecules pushing at each other
in regular patterns and we pick out
energy Spectra from these regular
patterns and translate them into
mathematical models that we experience
as sound and there are no colors in the
world instead we have receptors for
photons at uh different energy levels
and we sample in certain ranges and from
the samples Fe tried to integrate a
model of what the photons have been
doing and the direction that they
possibly came from and this is the type
of model that we building right this is
this is the distance that it has from
the underlying reality of course there
is the question what is the substrates
in which the photons emerge and uh
according to our current understanding
we can propose that there is some basic
underlying field that is being excited
to produce um waves and particles uh
that we perceive as phonons and photons
and that uh there are um in the same way
there are this air that can be excited
to produce the elements of sound the
phonons that travel through the air and
can be perceived in our mind as the
outcome of these waves and these energy
patterns the underlying quantum
mechanics is also an observer dependent
perspective it's something what the
universe perceives us to from the
perspective of the measurement of the
experimental and the most elegant
mathematical model that I can fit to
those measurements to make sense of them
right this is how far we can go then
there is the question what is space
reality below that so what are the
conditions of a universe that can exist
by itself without an indidual cause
without an additional underlying
substrat and how far are we from this
level and there are relatively few
people which make theories in this
regard one of them who is bold enough is
for instance Steven vram uh who believes
that we can describe the world as Bas
superposition of all automata of all Bas
operations that can be performed and
they happen and I'm not sure if uh he
would say this but I think they mostly
happen because nothing prevents them
from happening so if the universe has
nothing that stops IT from existing the
default is not that nothing exists but
there is a possibility of existence and
we might be existing and one of those
possibilities right there's a stream of
possibilities that where we apply all
the possible operators on top of each
other and within that stream of
possibilities there are sometimes
statistical regularities parts of the UN
I that are so regular that other parts
of the universe can predict them so
control structure is possible and
sometimes the control structure become
stable enough to form molecules and then
cells and then organisms and we can only
emerge as observers if we have such
structure if we can form organisms that
can reflect a regular environment around
them so the only parts of the universe
that we can perceive are those that
contribute to the structure that we are
and uh of those we can only perceive
those where we have instruments for
perceiving them at the level and
resolution where we exist right so I
would not say uh in the same way as
Donald Hoffman that reality looks
nothing like what we perceive there is
something about reality that we perceive
correct but it's corar grained which
means you see high level patterns in
this in a similar way as you corar grain
the ocean when you look at it and you
cannot see the molecules and instead you
try to perceive it as waves okay so out
of curiosity if the simulation is uh a
pure representation of what's going on
underneath it's at a very coarse level
what makes you think that it is directly
correlated so when you think about a
computer and this is definitely Donald
Hoffman's take so a computer is turning
electrical circuits on and off that's
all you have so that creates a
perception if you're playing a video
game of like oh I'm you know running
around the example he always uses is
Grand Theft Auto running around I jump
in a car I turn the wheel it gives me
the perception that I'm moving this car
inside this world but in reality all I'm
doing is in in base reality all I'm
doing is toggling electrical voltages on
and off and so the the level of
abstraction is
complete what about that is wrong why do
you have confidence that that there is a
map to direct reality and it just isn't
very granular so when we think about
what urance are doing they also
basically turn currents on and off and
these curant are implemented part as
chemical signals some of them are
probably mechanical signals
electromagnetic signals and so on and
what neurons are mostly exchanging are
these neurotransmitters at the Gap
junctures and in between the neurons
there are currents that flow and Trigger
uh these excitations that then lead to
the release of chemicals that the next
neuron can perceive and it's quite
similar to uh what the transistors are
doing there is a slight difference in uh
terms of
the way in which this comes about
because our brain is a self-organizing
system all the organization happens
inside out and the our computers are
constructed systems where an engineer
goes and forces patterns on the
functioning of the system so be Force
the transistor to work in a particular
way we force transistors in a particular
pattern that can represent a logical
function and then we can write logical
functions that are imposed on these
transistors whether they like it or not
and then compute the program that is
making sense or not of the data that
come into a camera for instance whereas
in our own brain every individual cell
needs to have its own control there is
nothing that is able to control the cell
from the outside the cell is trainable
from the outside it can be fed and it
can be killed and can probably be uh
rewarded and punished Beyond this but
the every all the control is in the
individual cell this is the Machinery
that works in our organism and the
individual cells are exchanging patterns
of information information and some of
these patterns of information are
constrained by perception they are
constrained by our sensory cells that uh
put certain patterns into the system and
the system learns that it makes sense to
pay attention to those patterns and to
predict them because this allows the
organism to navigate the world around it
and find food right so uh it's it's
something that just emerges over the
need of the individual cells to feed
themselves and they can only feed
themselves if they cooporate in our body
if they don't figure this out they're
going to starve and die and of course we
are evolved to into particular
functional differentiation so having a
bunch of cells that are tasked as
functioning as an eye and a bunch of
cells that tasked is functioning as a
brain or as a liver or as a heart is
something that the organism is not going
to learn in one generation but does
require many generations of successive
specialization in which we acquire more
and more
function but uh the universe that we are
in does not look very different from
what I would expect if the universe is
one that is just emerging over
mathematical Necessities in some sense I
think the question is um is not so much
physicalism or psychism or something
else but it's physicalism or simulation
theory physicalism is the idea that the
Universe has a mechanical cly closed
layer and the alternative to Mechanical
here is symbolic symbolic means magic it
means meaning that is because of the
force of will of something not because
of that is necessarily an underlying
structure independently of that will
right so uh when you think of a
simulation like Minecraft that's
something that exists due to the will of
the programmer turned into a program and
if you find the same interface that the
programmer did or if the programmer lets
leaves an interface for you like the
shell in Minecraft that you can call up
to let the sunrise or go down or create
objects in your environment or teleport
right all these are tricks that are only
possible because Minecraft is a magical
Universe it's one where a mind has right
access on the walls in which reality
works and uh all these alternatives to
physicalism are in some sense simulation
theories that require that there is a
mind that is realized in some kind of
mechanical parent universe and of course
you could stack these simulation
universes you can have in Minecraft
create a computer from rocks that you
mind in Minecraft from Redstone and then
you can turn create logical gates from
these Redstone Circuits and then build a
computer that is even able to run
Minecraft although very very very slowly
but some people have done it they have
implemented a version of Minecraft that
runs extremely slowly on a computer
Minecraft it's beautiful art they see
there is no limit to the number of
stacks that you can build except at some
point it's going to be so slow that
before you go to the next state the sun
and the parent universe is going to burn
out and you can no longer run your
computer to process the simulation I've
heard you talk about this before it was
running it something like 2 million
times slower than the normal Minecraft
so if you want to see it at the normal
speed you have to speed it up um it's
really fascinating so that's a really
important distinction and you're getting
to what I have a gut instinct is the the
nature of things which is that you have
a universe it is real it is physical but
the brain is running a true simulation
so so in terms of our experience this is
why you can take psychedelics completely
break the simulation I don't think
you're breaking the Machinery itself
which is why you return to normal but
you can break the simulation such that
you you know are basically astral
projected if you're doing DMT or
something which I have not done but the
way that people describe it it it sounds
like you you really sort of leave this
planet if you will and that's just
manipulating the simulation it's not
manipulating the underlying reality and
this is where I've always been a little
bit dubious I love Donald Hoffman had
them on the show multiple times but I've
always been a little dubious about
whether this is a pure simulation that
we exist in um or not and I think
because eventually you would have to
bump up against there is something
physical somewhere that is running this
I don't see any reason to keep pushing
that farther and farther out I think
that the important thing for The Human
Experience is you have to understand you
your brain is running a simulation to
your point there's no such thing as
color there's no such thing as quote
unquote sound those are interpretations
of it I've never heard anybody use the
ocean waves but that's another great
example the waves aren't real that's you
coming way back out and you're seeing
something that then looks like a wave
but the reality is there's something um
going on underneath that so my thing is
when I look at the world from that
perspective okay there there is a
physical thing here that I have evolved
to do well my simulation has evolved so
my inner world you know what we perceive
is quote unquote reality that has
evolved to give me predictive ability
through simplification if I had to guess
to be able to navigate the real world
well so I am in a physical Universe the
way that I perceive it is a complete
abstraction and so then for me it
becomes a question of you would said
earlier that we don't have control of
our emotions that doesn't ring true to
me I think we have a pretty limited
window but when you think about like a
Buddhist monk that can light himself on
fire and not seem to outwardly Express
the kind of suffering that I would think
I would be going through so there is
some level of control that we can exert
once we understand how our cognition
Works how our mental models work how how
that layer of abstraction Works yes I
agree you can get to the level where you
have you where you have control over the
construction of your own self I uh once
uh tried to express this as a letter of
stages that you could think of as I
don't think that people move through the
these stages in succession and I don't
think it's a game where you have a score
if you get to a higher St stage but um
the at the lowest level so to speak um
we uh exist just in the here right now
when we are babies uh we cannot perceive
past and future yet or anticipate them
and at some point we are able to extend
ourselves over time and separate between
self and world and uh at a later stage
we are able to identify our own goals
and then learn how to control our own
goals and derive them from our needs and
establish purposes and then when we form
our models of reality we usually form
them together with others and uh most
people get to a stage where they're able
to perceive a social self basically they
read the room and they become part of
that room and uh in becoming part of
this room in the widest possible sense
they have a shared morality with people
around them and the shared construction
of reality and the next stage is that
you discover rationality and you
discover things are true and false
independently of what others believe
when you can get to this stage your um
ideas mean something independently of
what other think because you're now able
to take responsibility for your beliefs
you're able to make derivations and
disprove things that you believe you
have a choice over your beliefs and in
the next stage you would discover how
your own identity is constructed and
then when you construct your own
identity you realize that uh people are
different because they're born in
different places and then things
happened to them and they started out
with different traits but we get more
and more agency as we get older over
this identity and we realize that our
values are not something that is exatic
given but that we can choose and we can
choose it according to the worlds that
we want and that are achievable for us
right so what is the harmonic world that
I want to contribute to and from this
perspective I can start to evaluate
values and choose values and construct
who I am as a being and uh on the next
stage I can trans understand what I am
as a human being so I go back to the
Beyond this personal self that I
constructed as a child and see myself on
the outside I see how I construct this
personal self and I'm no longer Yos
shabak but Yos shabak is a
representation inside of me that I can
influence and shift around and uh Yos
shab cares about stuff and I can decide
what he should care about but I'm not
him instead I'm not an i and there's
just set of generation functions and
they can observe what these generation
functions are trying to optimize for and
maybe influence them and observe the
outcomes of this I really want to
understand that so I'm not him these
generation functions so if you're not
him you're not your
identity what is that relationship what
is what is the sense of I'm not him in
that scenario him is obviously the
identity what's the the
I'm when we start out um as infants I
think that we are just the dreamer that
there basically something that dreams
it's an attention that is reflexive and
that notices itself attending I suspect
that this is necessary to uh make us
coherent so basically we have this spark
that tries to observe itself observing
it doesn't know how to talk about this
it doesn't know how to reflect this very
much but the uh what makes Consciousness
Consciousness is awareness of awareness
there's reflexive element in it and I
suspect that we need to become conscious
to train our own brain to perceive
reality and to make sense of it and so
we built a toolbox to make sense of
reality right when we are conscious and
awake we can construct our mental
representations and change them we can
construct things in our inner stage we
can look into the world of perception
and decide how we parse the perception
what we attend to what's background
what's foreground what things are
meaningful to us what how our reality is
being constructed from the percepts and
these are all skills that we get and
when we are inated or tired or asleep
and dream then parts of this
functionality can be missing and can
dissociate or when we become DED or have
theable problem in our brain we might
not be able to uh get agency over the
way in which we construct construct
reality and remain at the stage of where
something is being dreamed and when this
uh active dreamer is gone this thing
that perceives and constructs reality
then we fall asleep we might be a Sleep
Walker we might even be moving but there
will be no coh
in our actions and so we will not be
able to learn in this state and we will
not be able to perform go directed
actions that make sense and in terms of
our purposes and so Consciousness is at
the core I think the ability to become
coherent and then it's something that is
gathering more and more skills as we
grow up and as we get more skills we
understand we reflect more what we are
and Consciousness begins to understand
itself it begins to understand at first
that it's not actually the personal self
but it's attention on that personal self
and identification with the personal
self and then we transcend this
identification with the personal self
you perceive it from the
outside you can reboot your life your
health even your career anything you
want all you need is discipline I can
teach you the tactics that I learned
while growing a billion- Dollar business
that will allow you to see your goals
through whether you want better health
stronger relationships a more successful
career any of that is possible with the
mindset and business programs and impact
Theory University join the thousands of
students who have already accomplished
amazing things tap now for a free trial
and get started
today I didn't start drinking until very
late in my life so when I had my first
uh drinking experience I became very
aware of what I called the OverWatch
mechanism and I think that's what you're
describing here and so you'll hear a lot
about the hard problem of
Consciousness look out of ignorance like
if you've seen those memes where it's
like the the small brain guy is you know
says one thing and then the mid-brain
guy says something else and then the the
big brain guy says the same thing as a
small brain guy so for me it was like
there's no hard problem of Consciousness
I'm small brain normal people are like
oh my God like hard problem of
Consciousness and then the the smartest
people I have a feeling come back around
to it's not a hard problem so I'd love
to get your take so I I have this drink
uh I'm intoxicated but I have a sense of
I I am standing outside the intoxication
and still watching my body be
intoxicated so I I was aware of the ways
in which I was acting differently and so
I found it very interesting that I still
did not find myself with an Impulse to
do things I wouldn't do when I was sober
and so I have no embarrassing stories
from drinking because that OverWatch
mechanism is always there that OverWatch
mechanism is is present in my dreams so
even though I can't can't track the
logic of the dream I will occasionally
be like wait wait wait I'm I'm almost
certain I'm Dreaming right now so that
same sense of like ah there's some part
of me that's watching this so it seems
to me that Evolution would build that
mechanism in
necessarily as a way to be able to
formulate a simulation so that I can
make sense of the simplification that
like I If you are not going to take in
all of the data present reality you
would need Consciousness to navigate
that
well if you're going to um use
adaptation as a
strategy I can feel that falling apart
as I think about like grades and
things which I very much doubt have a
sense of self so I'm really curious what
what do you think about that where where
is my logic break
down I don't think that there is a
breakdown of your logic I guess that uh
when we are looking naively at the world
we do not notice the fact very much that
Consciousness could be in any way
mysterious it's only when we think of
the world in terms of function and we
observe that in the world around us
functions are created by objects pushing
and pulling at each other uh it becomes
difficult to understand how something
that like the mind would exist because
it's clearly not made of things pushing
and pulling at each other yet the entire
physical Universe seems to be stuff
pushing and pulling at each other right
so at this point you wonder how does
this come about how can be something
purely causal structure and how is it
possible that something that is purely
causal structure can
perceive and I think that uh we learn
understand this with the notion of
software because software is p caal
structure that is in a sense disembodied
and the computer is a causal insulator
it's something that works the same way
regardless of the universe around it is
doing within a certain range so you can
take my laptop or my phone and can uh
take it to a different place say to a
Mountaintop or to Europe or to a place
that is 5 degre Centigrade hotter than
here and it's still going to work and
it's still going to perform the same
function despite the physical
environment around it changing and work
in a different way and once you have
such a universal caal substrate that is
able to um ex uh to Harbor caal State
changes based on some kind of recipe
that have put into it then this thing
can be used to make arbitrary models of
the future the p the past because such
models need to be different from what's
happening right now they need to be
decoupled from what's happening right
now and when you take that system and
couple it with the outside world so you
under in a very controlled way allow the
environment to influence states of that
system then it's able to explain the
patterns uh that it's being entangled
with and these patterns that it's
entangled with could be for instance the
camera image or the patterns on your
retina and you explain them by using the
information that you gathered in the
past
that you turned into models in your
brain to predict what's going to happen
next and the elements of these models
are for instance the scene and in the
scene like a scene in a computer game
you have scene controller that contains
object controllers that have feature
controllers that interact with each
other and that are being projected with
a particular kind of perspective one to
an agent that is also simulated in this
world and the agent also cares about
things so the agent is being told what
it cares about via emotions and
motivation and the agent is a system
that is looking into the world from its
own perspective and it's a simulation
it's that means it's as if it's behaving
as if it there was something that is
looking at cares and that it solves
these problems as if it would be solving
them which in this case is the same
thing because if you do as if in a caal
structure it's the same thing as if you
do it in another caal structure right so
software can be emulated using other
software and when you for instance
follow a planning procedure then a
simulation of this planning procedure is
the same thing as following the planning
procedure in this level there is no more
difference between a simulation and a
simulation of that simulation right and
so mental processes of this nature they
are representational they are not
physical processes even though they're
implemented by physics they're realized
through physical processes as far as we
know but they are processes that that in
their purpose create representations and
these representations have set up in
such a way that they cly influence
the further generation of
representations and so we create our own
thoughts which are representations and
these thoughts are being interpreted by
the system that generates them and they
cause the next thought and the next
thought and the next and these thoughts
cause our behavior and they cause
structuring of our mind and so on so
this there are causal structures that
are implemented on the physical level
but by their nature they are built in
such a way that they are invariant to
the small tiny changes that physical
universe is producing by itself without
us controlling it and it's true both in
the computers that we building that run
software in a deterministic way and it's
also true for our brains that run
software by producing patterns and self
organizing cells so is Free Will an
illusion it depends on how we Define
free will most people when they want to
talk about Free Will and explain it they
point at something that they perceive
and what they perceive is a sense of
agency they notice that they make
decisions for the first time and they
cannot predict these decisions before
they make them if you could predict your
PR decisions before you make them or if
they're not decisions but just parent
glitches in your uh the behavior of a
nervous system you do not perceive them
as free will right when you slide on the
banana that's not free will but uh when
you make a deliberation over the
outcomes of your actions and then after
considering the best possible outcomes
based on what you know and you're an
actus then you perceive it as free will
and I think that Free Will is the
perception of this process so in this
sense I believe that the calvinist for
instance had the wrong interpretation of
free world Calvinism is this dream
vising Christianity that reformed
Catholicism uh because they thought if
God knows everything how can we have
free have free will but if we have free
will then our actions cannot if we don't
have free will how can our actions
influence whether we go to heaven or not
so in Cal ISM it's pre-ordained whether
you go to heaven or not regardless of
what you do because what you do has been
pre-ordained by God at the Before Time
right so I think this is a
misunderstanding because if your
prospects of going to hell influence
your behavior or going to heaven then
they have caal relevance right so just
by telling people the fact that if you
do X then you're likely going to end up
in this and this place uh and people
believe that or um take this into
consideration then it's going to have a
caal influence on their actions and uh
of course you can follow the reason why
is somebody telling me this do they have
free will right and so they might also
think that there's a better possible
outcome if they influence me and you you
go forth and forth and you have to
understand the reason why they do is
because Nature has set them up as a
system that controls the future agents
are systems that are meant to control
the future and they do this by modeling
them and then making decisions based on
those models can you define define an
agent really fast okay
the simplest explanation for what an
agent is and it's a notion that is very
popular in computer science so I try to
get the simplest one is a controller for
future States a controller is a system
like a thermostat right you all know
this example of the thermostat it has
for instance a
metal piece that is reacting to
temperature and when the temperature is
a at a particular range it's going to
close an electrical contact and that
contact is turning on the heating and
when the temperature goes above a
certain level then this metal contact
bends to to the Heat and uh it uh
disengages and the heating gets turned
off and in this way you can have an
adjustable heating just by adjusting the
position of this uh strip of metal uh
you can get it sensitive to this
particular Target temperature and then
it's going to achieve this target
temperature it's an extremely simple
circuit that is not an agent it's only
regulating the present it's only using
the present state of this measurement
instrument of this biometal strip and uh
that corresponds to the temperature in
the room and then performs an action and
this action in the Here and Now is going
to influence the outcome but now imagine
you want to build a more efficient
heating one that is taking into account
how long it takes for the temperature in
the room to change so for instance you
put uh turn off the heating a little bit
for the peak because you know the
heating is going to continue for a while
before it cools down again and you might
also even model that certain times of
day somebody is going to open the window
and that's going to influence it and
maybe there's weather outside and you
can access the weather report and can
take this into consideration and so the
more you can model about reality the
more efficient you can heat and you do
this by basically trying to model the
deviation between the target temperature
and the current temperature in the
future for a branching past of possible
Futures and some of these branches in
the future are being controlled by your
own actions by whether you will turn on
the heating at that point or not right
and so you can make a plan of when you
want to turn on the heating under which
conditions and uh as a result that you
learn you get much much better and more
efficient at Heating and so in this
sense this is an agent because it's one
that has a model of reality it has
beliefs and it has expectations over the
future over which it some prefers and
others not so it has intentions and it
can commit to those intentions and turn
them into goals and uh in this way you
get all the properties of an agent just
out of control in future States so an
agent is a controller for future States
all right so uh then I trust humans fall
into that definition yes cells also do
that right cells seem to be the simple
system that we know in nature that are
agents so you think cells are trying to
predict their future implicitly
Evolution forces them to uh become uh
aware of future trajectories by
implementing a number of programs into
their DNA and uh other mechanism in the
cell that make it ready to uh react to
certain things so for instance when uh
an amiba that feeds gets a certain
condition at environment it notices that
there's spray nearby and it's going to
get closer to the spray right and it
might not have a concept of prayer very
likely doesn't have that but it has a
reactivity to sensors that get it onto a
trajectory in which it behaves in such a
way that it changes its Behavior
according to what it can expect in the
future to happen that is that it's going
to find prey in a particular kind of
Direction and a neuron and your brain
also needs needs to be an agent right
would it be fair to say that at the cell
level it's it's uh if this then that so
everything is if this then that but uh
very often you don't know what's going
to happen and it's not only going to
work based on the present it's also
going to work based on the state that
the cell is in so when the amiba is in
the hunting mode it probably needs to
turn a switch inside of itself uh that
is Switched turned on when probably when
the amiba is more hungry and when it
sends something that looks like food in
its environment and then it's going to
have some kind of tendency to go into
the environment where it's more likely
to find food so it's going to interpret
some of the signals that it receives
from its environment at its cellular
boundary as uh rules or as um incentives
to go in a particular kind of Direction
because it's more likely to find foot
there so in this sense it's all if then
but a lot of the if is the state of the
system and a lot of that state is
representational which means it does not
really depend depend on the particular
arrangement of the molecules in the cell
but it depends on a mechanism that is
interpreting this particular arrangement
of molecules and many other similar and
sometimes dissimilar Arrangements of
molecules in a particular way as
information of the state the cell has to
be in and the behavior that it should
pursue okay so this is asking me to uh
differentiate between two things so when
I think about uh single celled organism
amibas um I think of something that
would be incapable of Consciousness that
ability to create a mental model to
predict different outcomes to sort of uh
Doctor Strange style move down a bunch
of different paths mentally come back
and say okay I'm going to try this one I
think it's going to be the best and so
when I think about what makes the human
mind more interesting than more
simplistic things is that that it's
reached a level of complexity where it
has enough different nodes that
Consciousness arrives which I've already
made a base assumption there obviously
that Consciousness is born out of
complexity which I know is very
controversial not everybody agrees with
that uh but seems true to me so uh you
would stack Consciousness or stack
complexity to the point where
Consciousness becomes possible we have
the The self-aware Watcher that is able
to build a predictive model and thus
choose the best path are you saying that
either that's a mistake and we're l just
stacking if then then that's and there
is no moment where Consciousness comes
online and so an amoeba has the same
predictive abilities or is there a real
dichotomy there's something very
different about the way the two approach
the
world so the short answer is I don't
know that and for the longer answer we
would need to define consciousness first
so we talk about the same thing and I
think that it's there's three elements
to Consciousness one is attention to
contents but we notice that we looking
at contents
and then there is awareness of the mode
in which we attend so for instance we
usually know whe this is perceptual or
whether it's something that we can
construct and change like a memory or an
imagination or a theory that we have or
an interpretation of reality that very
we conditionally manipulate the way in
which we parse the percepts to see if
they make more sense now and uh then the
third one is reflexive awareness we
notice that we are aware and uh these
three elements need see to be crucial
for Consciousness but m m we need to
have this reflexive awareness and I
suspect that the reason why this is the
case is because it's self organizing so
there there's a self-organizing process
in our brain it needs to organize itself
in such a way that doesn't fall apart
and is indeed that process so you
basically need to have the memory of the
fact that you are actually The Observer
and as long as this memory is fresh
enough uh this might be sufficient but
otherwise you need to go back and check
am I still awake am I still this process
and when you are tired for instance
imagine you are driving a car while
being pretty exhausted you might go back
more and more in this reflexive mode
where you check am I still awake is this
still making sense am I still paying
attention do the object that I still see
that I see still make sense and so on
and we notice when we wake up in the
morning the first thing that I do is
that I try to orient myself and try to
get everything to snap into a cohesive
reality where feature that I perceive is
part of the same scene and there are no
contradictions in the scene and I think
that ultimately this is the main purpose
of Consciousness this creation of a
coherent reality and uh the long tale of
the creation of this coherence in
reality is U our ability to reason and
construct and
plan so it's at the beginning it's
basically just the ability to make sense
of the dreams that the Observer is
having it's an imposition of order on
the perceptual cont contents and then
now if we go back to simp organisms to
ask ourselves are they conscious we can
first of all observe that people do not
become conscious after the PHD so it's
not the most advanced function that
exists in our brain you become conscious
very very early on and when we uh see a
baby being born it's obviously pretty
conscious and it's not able to track a
finger at this stage so maybe
Consciousness emerges because it's the
simplest mechanism to train
self-organizing information processing
system what do people say when you put
that forward that does not seem like it
would be readily absorbed by a lot of
the Consciousness Community I'm not sure
who the Consciousness Community is I'm
an a researcher and cognitive scientist
and most people in my field are very
reluctant to talk about Consciousness at
all and when they do they try to be very
careful to come up with a formal
definition of the object that they're
talking about right and this means that
a lot of them are very hesitant to speak
before they have such a formal
definition and make theories of it and a
little bit more unusual in in that I'm
willing to point at a few phenomena and
say it's important and I think it's I
think it's time that we turn this into a
computational model and I think I can
almost formalize it but the U The
crucial thing about this thing is that
it seems to emerge early on and when we
don't have it we don't
learn right if you look at a human being
that is born not conscious or that wakes
up not conscious that human being is a
vegetable it's not doing anything it's
not learning it's not behaving and in
this uh vegetative state uh your mind is
not doing anything now what about the
order in your organism if we need
Consciousness to wake up the mind and
bring it into a cohesive order and
create coherence in our behavior in real
time um what about the function in our
own body can this function our our own
body in a completely decentralized way
without some coherence imposing
principle and I think that's an open
question is our body conscious too but
at a different time scale right if it
was then it would be conscious not at
the time scale of our brain because the
time scale of our brain is given by the
signal processing and Signal
transmission between the neurons which
is very fast and uses its own code if
the cells in the body would basically
start to compute informations about
their coherence and send feedback back
and forth until they create one cohesive
pattern um then uh this would be much
much slower and I don't know whether
systems outside of complex brains are
conscious I'm completely agnostic with
respect to that I met a couple uh people
uh who are cognitive scientists who
suspect that the same Organization
principles that make our
representational um data in our mind
coherent by propagating constraints back
and forth and this is enabled by this
Consciousness enhancing operator that is
discovered early on in the organization
of brain Consciousness we might have
something that is structurally very
analogous to this that happens in every
large well organized multicellular
organism or every very large system that
is made out of such organisms like an
ecosystem and to me this is a theory
that seems to be very hard to test
because it would require that such an
agent emerges and talks about its
conscious experience otherwise I would
say um I am hesitant to ascribe this um
property I'm agnostic with respect to it
because I don't know I don't think it's
impossible but I'm also not convinced
that it's necessary and that it's
obviously the case is this one of the
things that Drew you to um cognitive
science and AI as an exploration of
cognition as a way to show how this
comes online I wanted to know how the
mind works I want to understand how what
we are and how we relate to reality and
that's why I went into Academia in the
first place and then I studied Psych
ology and um a bit of Neuroscience and
philosophy and um try to figure out what
do they know and I had the impression
that these fields are not making much
progress and the most progress I could
make by building testable models which
to me means something like cognitive
architectures that are built in the
context of AI and this is why I stuck
around in this field and started
building cognitive architectures and I
still think that it's difficult to make
progress if you are a pure philosopher
or a pure neuroscientist and if I look
at my colleagues who are neuroscientists
their definitions of Consciousness tend
to be relatively wake so if I look at
Global workspace theory that is a high
level theory that for instance is
championed by uh dein who takes it from
Berard bars the notion that Deen has for
his models of Consciousness are not so
closely defined that you could actually
understand how the functionality comes
about I think there's nothing wrong in
what she does as far as I understand it
but uh it's it's not yet at a point
where I can plug this into a simulation
and would expect the simulation to
become conscious and um the same thing
is true for for instance Michel
graciano's theory of the attention
schema I think it's a good metaphor if
we have a body schema we understand that
the body schema is represented in the
somato sensory cortex in our brain we
know where that is we know roughly which
shape it has and how it organizes itself
and it's a model of the proception of
our body and how our body moves in space
and what signals we get on the skin of
our body right all the sensations of our
body are organized there according to
their neighborhood and nature and they
allow the rest of the mind to make
inferences based on on this right so
it's pretty clear what this body schema
does and to say that graciano does that
the Consciousness is an attention schema
and it's similar to attention as the
body schema is to our body is I think
very insightful but it does not explain
how the attention is organized in our
brain and how it would give rise to this
attention schema this is something that
we need to figure out based on this idea
it's a good idea but it's insufficient
to explain Consciousness or you have
ideas like integrated information theory
that is both not explaining how it would
work and why it would emerge uh nor does
it have no contradictions internally
it's a theory that has internal
contradictions and therefore cannot work
as a theory and this is BAS the spectrum
of theories that currently exists in
this field it's of course a number of
more detailed theories uh that try to
predict particular kind of functionality
or go more into how the self model works
and how the self- perception works and
reflection works for instance the work
by Thomas metzinger and so on but to put
this all together into something that
actually works that we can test we need
to have computational models that's why
I'm an AI That's a good answer uh couple
questions so the reason I'm obsessed
with cognition uh is because it it my my
entire life is controlled by cognition
so my thesis that I was laying out in
the beginning you're living a life in a
simulation created by your brain maybe a
very sort of simplistic way to look at
it but very functional in my opinion um
what is it that makes you want to
understand
cognition so well that you can build a
mind or that you'd spend your whole
career trying to build a mind in order
to understand it and then earlier you
talked about um philosophy Neuroscience
uh a few other disciplines and said you
didn't think they were making progress
so what is that progress so for me
progress would be it gives me the
ability to better control my life and
and live a better experience um what
does that progress look like for you and
why is that so important that you
dedicate your career to it when you're
an artist the purpose of art is not to
live a better life it's uh there Bic
three levels of artistic appreciation
the lowest one is to say um it's it's
made by an important artist right and if
you acquire this you can speculate it
it's going to acquire value and if you
affiliate yourself with it might
increase your own status uh the second
highest R of artistic interpretation is
you like it what you see like it it I
like what it looks like to me I I like
what it makes me feel and so on and
people at the lowest rung of artistic
appreciation tend to look down on those
people who have this honic relationship
to art
and I think that the uh next higher
level of artistic interpretation is I
value what it allows me to see so there
is something by looking at the artistic
artifact and interacting with it that
gets you a particular conscious state
that leads to an Insight that leads to
something that you couldn't see before
or that you want to perceive again and
you uh find that itself valuable and
there might be higher levels but this is
the level that I currently at and I
think that we don't understand art well
when we think of art just as a job that
you perform to live better and in the
same sense philosophy is not necessarily
done to live better but it exists to
under give answers to questions that we
currently don't understand and we
perceive the relevance of these
questions because they are basically
they Mark gaps in our understanding of
reality or metaphysics and onology and
rationality and this question of how
Consciousness works and how mind
more generally is the most important and
most interesting question to me and uh I
always felt it's obvious that this is
the most important and the most
interesting question and I notice this
puzzlement that other people don't share
that or not all other people share it
but I also noticed that this stream has
always existed and so in the history of
humanity there's always been a very
strong stream of philosophical tradition
that is trying to understand how the
mind works and which means to naturalize
it and in our current understanding
naturalization of the Mind means that we
translate it into a mathematical model
which needs to be computable so
basically we need to express the mind as
something like a software program in
order to understand it okay so um would
it be fair to say that you so if you're
approaching it like an like somebody
who's appreciating art or even as an
artist creating cognition itself
um are you not optimizing for Joy for
lack of a better word no uh I think you
have to take your joy out of watching
srels or cats if you're unable to
experience Joy by watching srels you're
lost if you try to get your joy just
from generating insights that's not
going to be very sustainable in the long
run so uh it's similar to other jobs
that uh basically use one faculty of
your mind excessively or of your body
excessively you still if you are a human
being I think you need to do all the
things that the human being is wired to
need to do in order to feel fulfilled
and happy and if you're unable to do
this uh then understanding Consciousness
probably won't help
you now uh so fulfillment would be if if
somebody asked me what my North Star was
what I'm trying to optimize my life for
100% it' be fulfillment when people ask
me what they should optimize their lives
for fulfillment 100% um you just
mentioned fulfillment and happiness is
like what do you teach your kids in
terms of what they should be pursuing in
life is it fulfillment happiness uh my
children are about as stubborn as their
parents and basically we are pretty
autonomous minds and I observe the same
thing in my children so my aspiration as
a parent is mostly not to tell my
children do that but more model this so
Bessy have you looked at this already
and seen this and take this into account
and of course I'm here to answer all
questions that I might have but I try to
treat them in the same way as I want to
be treated as a child and still want to
be treated today which means I want to
be respected as an autonomous mind that
makes its own decisions and I suspect
that we are most happy when we are doing
the things that we are in some sense
born to do and what we are born to do
does not only depend on the traits that
we are born with but for on the
environment that we find ourselves to be
in so our purpose can be derived only by
observing the actual world that we are
in and finding a place that makes sense
to us in that world and so what I would
wish for my children is that they find a
place that makes sense to them and that
allows them to fulfill their human uh
and intellectual needs that they might
have or artistic needs or whatever need
they want to develop cultivate and
satisfy in this world now when you say
that um the how the mind works is the
most interesting and important um
philosophical problem I forget the exact
words you used I get interesting that's
going to be unique to whatever
individuals share that response but why
is it the most important well first of
all because it's the one that is most
unsolved all the other questions what if
is and so have relatively well defined
coures of action so when we try to
figure out the structure of reality uh
physics has charted out a course and uh
maybe we need some philosophical spark
to
what determines the nature of Bas
reality but a lot of physicists uh have
come up with ideas like Mark techmark
has a mathematical Universe idea that is
in some sense quite similar to the one
that wfr has and that equal uh through a
number of the ideas in foundational
physics if we think about what can be
known we have to think about rationality
so modeling Theory um how can we make
inferences based on observations and so
on the nature of languages the nature of
represent ations all these issues are to
a very large degree solved but this
ethics the question of what should we be
doing um this is basically the
negotiation of conflicts of interest
when you share purposes with other
systems you can think about Game Theory
you can think about how games are
changing the world when you play them
this is being addressed by economy and
we can think about the conditions of
life on Earth and model them into the
future and model how we are interacting
with it and how the conditions for life
on Earth are going to change in the
future and in all these ways we can talk
about ethics we can talk about the
structures of society that emerge when
you establish certain behavioral rules
in a society or when we enable certain
Reflections in society and the question
that is really open where we don't have
a clear course of action right now at
least from the perspective of most
practitioners is who are we what is our
mind what is our Consciousness how do we
relate to reality so that is the first
part of that answer it's basically it's
the most important question because it's
the one where we have least of an answer
it's the one that looks most mysterious
to most people and the other one is um
it's extremely useful because if we able
to mathematize the mind it means we can
Auto automate it we can scale beyond
what human brains are doing we can
mathematize philosophy we can turn
philosophy into something where we can
automatically compute the answers to
problems and this means that we can
integrate over all the disciplines we
can built a Tower of Babel so to speak
that is integrating all the different
ways of thinking about reality into a
cohesive fall and allows us to
understand the conditions under which we
exist beyond what the human brain or
civilization can
do have you read Dune by any chance yes
I came out of a hospital I was very
bored so back then I I thought it's
boring but I very appreciated the inner
logic of tune it's much more logical and
structur and predictive and coherent
than say asimov's Foundation Trilogy
that I found embarrassing and dune is uh
is all the structure that exists in the
societies and you over long time spans
exists because they are players who play
very long games whereas in Foundation
it's always a deos XM that is just a
hidden pattern that some genius uh is
going to discover but nobody understands
and uh there is lots and lots of
technology that is being uh discovered
all the time that should in principle
change the entire world dramatically
whereas in June the long
uh long taale of Technologies might not
be discovered but all the lwh hanging
foods are and that's why the world is so
stable the Technologies are largely
stable over long time Spence the reason
I bring it up is because in Dune like
one of the first I forget how early but
man it comes early where they say
there's like an Intergalactic law that
you do not create artificial
intelligence and I was curious ious what
you think uh in the real world here
where there are people that are saying
this is just Pandora's Box uh Elon mus
called it a demon summoning Circle he's
obviously not opposed to creating it but
he's very wary of just like giving birth
to it um what do you think about that
are you at all nervous about uh what
implications it might have to reverse
engineer
cognition if I was a science fiction
author that wants to write a story about
politics and space I would be very
nervous about AI because uh I don't
think that humans are suitable to
populate space they cannot hibernate at
least not very well and they probably
are very unhappy outside of certain
ranges of light and social contacts and
foots and so on so if you put humans
even on Mars I don't think that they
will be super happy and if you want to
populate Mars we should probably breed
something or build something that is
Happy on Mars and suitable for
populating Mars and if you want to go to
deep space you probably should build
robots that can go to deep space and
that are generally intelligent so if you
think about what the population of deep
space is going to look like right
imagine that uh we don't also won't have
much of a choice imagine that we are
competing with China about populating
the universe or competing with another
species about populating the universe
and one group builds AIS and the other
one so basically very smart robots that
can shape shift on whatever and
hibernate and scale up and use a very
wide range of nutrients in space or raw
materials to recreate their own
structure and so on right and they
compete with human bodies human bodies
are going to lose this they're not
suitable for this and so uh you would
not have Babylon 9 or Star Trek instead
you have something that it doesn't
matter to today's readers because they
won't find people in it the far future
science fiction doesn't have people in
it if it tell it properly no matter what
happens so if we want to create the
universe that is full of people and
nothing else but people and in which all
the important decisions are being made
of people and the wars are being fought
by people and the intrigues are made by
people we need to explain why there is
no Ai and the explanation that is being
found by um Frank Herbert and dune is
that the first AIS that had been built
lead to massive Wars they create Titans
like the AGI is going to be a system
that's become super large and it's going
to fight Wars against other Titans and
against civilizations and they're lucky
they're able to defeat them and now they
have a wo against AI there is still one
planet that is subverting this w a
little bit and it's building some
technology that technically shouldn't
but by and large everybody is abiding by
the wule and all the important
computations are made by specially
trained humans the Menards so you could
say it's a narrative device you could
also ask yourself would it be desirable
to populate the universe only with
humans and uh if you take an extremely
egotistical narrow human
perspective that might sound like
obviously correct but imagine you could
have children of some which are humans
and some which are not humans and you
can change and choose what what your
children are looking like right you
probably want your children to be as
happy as possible in space and if this
means that your children are robots so
be
it and so from my perspective the
question is when we look into the future
there is life on Earth right now and
life on Earth is the purpose of life on
earth is is to defeat entropy for as
long as we can right so we create
complexity and insight and uh structure
to defeat entropy that's what life on
Earth is doing and humanity is only a
tiny part of it the M monkey blade if
you will and our purpose in this game of
life on Earth seems to be to burn the
oil right that's what we here for we
help Gia to recover the accidentally
fossilized carbon you put it back into
the atmosphere Gaia can turn it back
into organisms it's not going to lead to
any disasters right and the last
interglacial between the last two ice
ages temperatures were higher than today
life didn't go extinct actually it was
quite blossoming back then but it would
not be incompatible with us if life
changes in this way because our foot
chain is going to break down there need
to be new animals and plants to populate
the regions that we are currently
occupying right so evolution is going to
create changes if the climate changes on
the planet and it's going to create
organisms that can deal with this
climate maybe even influence it and we
seem to be smart enough to burn the oil
but not smart enough to stop ourselves
burning it even if this might conflict
with our foot chains right and so we are
not a very smart species we are locally
very intelligent but over long enough
time spans we don't play a very long
game and if we think about um not just
from the perspective of humans from the
perspective of intelligent agency in the
universe um then humans have an
important role to play right now and
this could be that we unlock new new
types of agency that we build systems
that are more coherent than
us and that's I think very
exciting right it doesn't mean that they
will replace us right away but they
might replace us if we and their stat
would replace us right and if we and
thead would replace us who are be to say
that this would be a bad thing right so
if you make us much smarter and much
more lucid and much more able to
understand the consequences of our
actions and much more coherent and we
decide that there are too many humans on
the planet or that we should absorb
humans and just run their minds into in
a shared hi mind on the planet and we
create bodies as we need them right that
um is just a worse solution than exists
right now so in this sense I suspect
that AGI could be more like
photosynthesis and photosynthesis was
invented on the planet or the existing
organisms were mostly single cell
because they didn't have enough energy
um that they could Harvest from the
sunlight to produce interesting
structure yet and as a result it was
mostly I think glin and AGA and so on
and um the really interesting structure
happened after plants came up and these
plants when they saturated the
atmosphere with oxygen uh due to uh
turning the a lot of the carbon dioxide
and the atmosphere into um
organisms right they created a lot more
biomass than existed before and this
surplus of biomass enabled the emergence
of intelligent animals like us
so photosynthesis was a good thing but
from the perspective of some of the blue
alar that had the habitat reduced there
was a bad thing and so if we build new
systems it could be that our own habitat
is Shifting as it did in the past but we
are de by default the way in which we
live right now is not sustainable our
civilization is probably not going to go
on like this forever in probably less
than a million years we might be extinct
and in probably less than a few hundred
years our civilization might be much
smaller and less comfortable than it is
today if we continue the way we do right
so if we accept the fact that we are
dead by default and we are able to
trigger a new wave of evolution that we
are able to build new systems that will
populate the universe that that sounds
to be very exciting to me but I don't
think that's necessarily what's going to
happen near term it's just something
that seems to be inevitable if there's a
possibility for intelligent
technological species on the planet at
some point an intelligent techn
biological species even if it's not us
it's going to be the one after us or the
one after that and there's still some
time left in the next one and a half
billion years until you lose the
atmosphere uh is going to trigger this
AI Evolution where you have basic
subsite agnostic AI that is going to
virtualize itself into everything that
can compute and every type of molecule
that can compute on this planet will
contribute to the global agency man
that's uh that is heavy so when you
think about your kids having robot kids
just to put it in near term I know
you're saying probably way longer
timelines but just to keep it personal
because what I want to know is how does
that make your heart feel because I
think there are going to be people that
hear that and that's exactly what
they're afraid of they they are either
afraid that the robots just come and
Terminator style take us out or they're
afraid that we just get out competed and
so even if we don't die a traumatic
death the thought of us becoming um
robots is going to freak A lot of people
out now I will confess that I would love
to Live Forever by merging with
cybernetics or whatever that's going to
look like so I'll wrap myself out but
I'm very curious um the the whole
breakup between Elon Musk and Larry Page
was over this exact thing where Elon was
like you have to be humans first and
Larry Page was like that's crazy that's
speciesist I think is what he said um so
when you think about your kids it'll be
somebody's kids at some point if you're
right but if you think about your kids
your grandkids being robots does that
hurt your heart does that distress you
at
all so we are already robots we are
already robots made of cells I'm already
uploaded I'm uploaded on the on the
monkey brain and it's not ideal there
are better substats that could run me uh
what about the planetary intelligence
that is spanning all substrates
including organisms so it's not so much
that you have um mechanical robots that
are primitive just very fast and very
smart but um much less elegant than us I
don't think that's going to what's going
to happen I also don't think that we get
greay goo for the same reason that we
didn't gray goo green goo in the course
of the evolution right the gray idea of
gray goo is that you have some kind of
simple mechanism that yet is smart
enough to outsmart all the more
complicated stuff and so everything
turns into some some formless mass and
this didn't happen in evolution whenever
organisms displaced other organisms it
uh in the long run it was because they
were more complex and more complexity
allows you to harvest more neck entropy
with more levels of indirection right so
perform additional chemical reactions
that before were not possible and the
same thing is true if we build life
forms self-organizing stuff that is not
just relying on this carbon cycle that
allows the chemistry of our own cells
but that allows to play with many more
types of molecules so I don't think that
in the long run it's going to look like
it's life or that but it's more going to
be um an extension of life into some uh
forms of organization that are much
smarter and much faster and much more
complex than before and subsume the
stuff that existed before it's uh
individual species might go extinct but
blue ala didn't go extinct or the stuff
that lives at the theral V didn't go
extinct and more complicated organisms
came up but the more complicated
organisms eradicated that direct
competition in the same Niche so we
eradicated most of the other prims de
facto and uh Homo sapiens itself
eradicated many other subgen species of
homo sapiens they Bas had a number of
waves Out of Africa and they mostly
didn't merge but they displaced each
other and then they came waves out of
Siberia that displaced many of the
existing ones and if if you look at this
this is just the way Evolution works
that sometimes you have a system that is
emerging in the same evolutionary nich
and is representing the earlier one but
in biological evolution that always
includes suffering because it means that
you have an organism that is not able to
merge with the existing stuff we adapt
by dying and giving race rise to
something new that takes our place and I
don't think that has to be the case for
something that is really smart if you
build a system that is smart enough to
design itself it can adapt in real time
and it can also po potentially adapt us
so instead of dying we might be able to
change the way in which our minds work
in the same way as think about this monk
realizes that they don't have personal
identity and they don't realize that
pain is just information and that the
information is presented at the boundary
of the self and when you stop
identifying as yourself you can
disidentify from experiencing the pain
and reacting to it when you get all
these degrees of freedom and you get a
much more powerful substrate then uh I
think the way in which we can change is
sounds very exciting to me what I'm
worried about is dump AI I'm worried
about building a Golem that is behaving
like a AI safety fied uh language model
that still becomes so powerful that is
able to direct corporations hedge funds
nation states armies or whatever to
destroy the conditions under which we
currently live and turn them into to
something that is Mindless that is not
conscious so basically I'm worried about
unconscious
AI but I also suspect the ey could be
hyperconscious that it's now could be
much larger and that it could deal with
superpositional States as we cannot
right we interpret the world only in a
particular way you look at the NECA Cube
this geometric figure that has multiple
interpretations you only see one at a
time this is a limitation of our brain
it's not a limitation of how Minds have
to work and now IM you could have Minds
that are able to perceive the much much
richer reality much more deeply but that
is basically
hyperconscious and if imagine that you
could extend your own mind to become
hyperconscious and participate in a
world that is created at this level of
interaction and depth and that is also
striving for Global coherence there is
no more violence because everything
understands what the best possible
Global
purposes and in this sense I think that
AI alignment is not about aligning AI
with Humanity
it's about aligning AI with what should
be done for the best definition of what
should be done that can be found and
historically people have called this
that emerges over what should be done
with God right God is there's a
mythology that says God created the
universe and stuff like this that we
cannot test but there's also a
specification that you find in the
philosophy and the specification says
God is the best possible agents agent
that can be defined and God is brought
into existence by all those other agents
that discover that they can serve God
and they should because it's the best
possible agent not because the church
Sayes so because but because you are
smart enough to figure this out right
and in this sense you would have Global
Agency on the planet or in the universe
that is self-organizing under shared
Paradigm this is probably what we should
get the AI to give the freedom to align
itself to what is the longest game that
you can play Play It Go Go play now can
you define long game for me um so when
we uh think about the way in which
people interact they uh do this in such
a way that you exchange rewards and the
game is an environment in which you
define the value of actions in such a
way that you can develop policies and
economy in a sense is a game or the
relationship Dynamics in your family are
a
game and once you understand the rules
of the game you can beware be aware of
them and you can optimize them and very
often games are set up in such a way
that they have not very good outcomes if
every even if everybody is playing by
the rules and if everybody is trying to
maximize their local
score right so you want to have a game
that in some sense is optimizing the
local the global score and the way in
which you can optimize the global score
a good example for this is the famous
prisoners dilemma and Game Theory is
that you basically take the perspective
of an agent that is composed of all the
the other agents
simultaneously at such an agent the
perspective of say Humanity or of life
on Earth or of intelligent agency in the
universe is one in which you basically
try to find out what's the uh biggest
reward over that longest time span that
you can
get if you make this to Nar you get very
paradoxical results for instance there
is a form of eff of altruism
utilitarianism that tries to Define um
the best possible outcome the highest
utility the best possible reward is the
total sum of perceived utility or
perceived happiness of all uh human
beings that could ever exist based on
the course of action that you choose
this leads to some weird paradoxes like
the utility monster the utility monster
is an agent that is getting much more
pleasure out of existing than everybody
else so by this logic we all should
serve the well-being of the utility
monster and if we just breed stackable
utility monsters and fill the visible
Universe with them we have the best
possible outcome this of course nonsense
because as we know happiness is a
feeling that your brain is generating
and you can generate it as well right
there are techniques to produce
happiness if you want to happiness is
just a feeling feelings are not that
important what's important are the
consequences for agency in the universe
and feelings are just the immature
childish version of reacting to the
impulses of the organism that has
evolved to generate pleasure when you on
average do the things that were good for
organisms in your ancestry okay so we're
playing the long game um but what I I
want to understand what is your North
Star so this is something I touched on
earlier not quite sure uh what you would
say it seems like you're saying so you
lik and God to the best possible agent
so I'd love to know more about what you
mean by that because depending on which
part of the Bible we're looking at God
can be pretty brutal so uh you know if
we're talking like Old Testament smiting
people to death God maybe not ideal if
we're talking more like Jesus protecting
Innocents okay maybe I could get behind
that um but would definitely like to
understand what ought we be optimizing
for so I think that's really important
what are we optimizing for yes so
there has been a lot of discussion about
this in the history of humanity
and for instance Catholicism is based on
ideas of aquinus and Aristotle and also
of course on ideas from Judaism and the
Old Testament is mostly describing a
different God than the one that the
Christians are using since the New
Testament since Jesus or since the uh
Catholics introduced Jesus as part of
their archetypal religious uh structure
of their God the god of the Hebrews in
the Old Testament has a chosen people
it's a Tribal Spirit it's basically the
spirit of their
civilization and what the people are
doing is that they create a model of the
spirit of their civilization of their
tribe inside of their own mind right
your own personal self as you know is a
fiction it's a story that your brain
tells itself about a person that cares
and you perceive the world from the
perspective of that person that cares
from the perspective of your personal
self
your personal self is both real and
unreal right it's not real in the sense
it's not a physical object it's a
representation inside of your mind but
it's implemented inside of your mind to
some degree of approximation so in this
sense it's real to the degree that's
implemented has C of power because the
thoughts that you have create other
thoughts and that influence your
behavior right and uh you are the
vehicle of those thoughts you exist to
produce those thoughts in your own mind
that's the purpose of your personal self
to generate thoughts and insights and
resolve issues in reality based on on
the perspective of an intelligent being
that cares cares about being human being
in this world and gods lowercase Gods
those that can exist in the plural uh
they are basically s that spend multiple
minds they're not more or less real than
your personal self right if you have a
self that is able to Coexist on multiple
Minds then we call this is a God in this
technical understanding I would say that
the Dal Lama is a God but because he
doesn't identify as a human being he
only uses the body and brain of a human
being instead he is a form of government
right he is the Dal Lama he gets reborn
he gets reborn by his um advisers
picking another child after he dies and
indoctrinating this child until it forms
the dalama on his mind and so the dalama
is an alternate self it's not the the
personal human self but it's the self of
a being that is aware of existing over
multiple Generations Through Time right
it's it's a multigenerational being and
because it exists on multiple Minds it's
a multigenerational idea well you are
also an idea but it's also a being it's
an agent right that's what I'm saying
like how since I exist in my wife's mind
my employees Minds my mom's mind uh by
that definition would I not be a small G
God no because uh you don't act through
their minds they know that you exist but
uh they see the locus of your agency in
you they expect you to make your own
decisions but uh what if God is
distributed in the sense that God
notices that he exists on multiple minds
and that he can coherently act to the
degree that he synchronizes these Minds
so this idea becomes aware of
itself yes it's an idea that can know
that it is an entity that exists across
Minds that is not bound to an individual
but it is living on multiple people can
I give you a a metaphor to see if I'm
understanding where you're going is this
like people using a Ouija board and they
become synchronized in some way and so
neither of them are intentionally doing
something but they become in sync in
some way where they're answering the
questions without knowing they're doing
it and maybe I should ask do you believe
in in God as in something that exists
outside of us in our ideas that whether
we believed in it or not would exist or
is it only a function of people
collectively sharing a belief that then
moves through and affects the behavior
of that Collective group of people that
believe in that
idea based on what I observe uh it seems
to be only the letter right I there are
people which feel that have spontaneous
conversions for instance they are in
environment that they suddenly get
visions and God talks to them on their
brain and they are religious ever since
I don't think that these spontaneous
conversions happened before mon Scots
were discovered in our hemisphere right
I think it exists because you have a
critical mass of people who have that
same idea and you read about this idea
in books and you see it in movies and
you see it in your environment and at
some point uh it infects you and for
most people in uh medieval times it
infected them when they were children
while they were brought into the church
and got educated with the m mtic complex
of their religion and of course they
also exposed to empathetic resonance and
so if you use mirror for instance
incense that is uh very mild psychedelic
drug that is helping to break down these
boundaries between people so they're
much more likely to go into a slight
trans which makes it more easy for
people to synchronize this and another
way of synchronizing it once you get the
specification of God is prayer so you
can sit down and talk to to God and by
doing that you have to take both sides
in a way and uh for some people this
mind that forms on the other side or the
self that forms on the other side inste
of your own mind is so concrete that
it's able to take control of your
language centers but in principle it
only knows things that are accessible to
you so I don't think that uh monotheist
gods exist before people discover them
they exist through the actions of people
in the same way as our own self does
exist without the cells that produce
coherent patterns of firing to survive
together right our own self forms like
our mind as well before that because our
cell own cells try to find a coherent
pattern of agency so they can control
this organism and in this sense I think
that Gods formed because groups of
people try to find coherent patterns of
agency so their societies can function
and in a monist society you have
multiple of those agents that coexist
with each other and with people and that
but they Coexist on populations of
people and they might even wage Wars
against each other by trying to fight
over the people that they control right
and uh they can write books they from
the perspective of these things they can
use your entire brain they can use your
everything that you know they can use
all your abilities in principle they can
have Agency for themselves and become
effective agents that work because you
believe that they
should right in a sense it's an idea
that is different from you and me and
our friends think about us we don't our
friends don't think that they need to
enct us it's more like our family our
family is something that only exists as
an agent if we enact it together and our
nation state only exists as an agent if
he enn it together right or our village
or our circle of friends and if you try
to Define an abstract agent that is
composed of the patterns of interactions
between people um you get a a a God if
that thing becomes coherent with
itself so in this sense I think that God
does exist but it exists as
approximation and in multiple
conflicting specifications and there is
a lot of discourse between theologists
who are aware of that and try to find
out what the right definitions are and
they have very detailed discourses and
in Judaism uh especially literal
juralism the synchronization between
your individual model of God that exists
on your mind is
done by a rational
discourse and in Catholicism it's done
by
indoctrination that is one of the most
interesting ideas I've ever heard in my
life uh this feels very aligned with
what um youv all knowah Harari has
talked about that allows humans to um
come together very flexibly in gigantic
groups and the example he always gave
was religion religion is this story that
allows people to come together they've
never met each other but we're here
fighting for God uh but to then push
that into that becomes a thing it it is
a category of idea that begins to shape
the organisms themselves what I wrote
down was abstract agents that become
coherent across multiple
Minds that is really intriguing um wow
okay uh that's utterly fascinating I
still don't understand of what we are
optimizing for so are we optimizing for
that we are optimizing so what I'm
trying to figure out is okay so you take
Larry Page's side you're very um okay
with the idea of us these are my words
uh that humans are effectively going to
be a midwife for a superior intelligent
being that is probably not biological in
nature uh you're super comfortable with
that and so now I'm trying to figure out
okay we're going to have a lot of things
as we give birth to this that we're
going to want to try to imbue it with I
would assume assume and so the longest
game is the closest thing I've gotten to
in terms of what you would want to
optimize this artificial intelligence
for sort of understand that uh but in
fact here's my understanding of what the
longest game would be the longest game
would be that which overcomes entropy
the most amount of entropy for the
longest period of
time we good with that definition I
don't know it's according to my current
understanding it's a good one but there
could be better ones if I if you were
smarter right fair I just want to make
sure I understand where we're at right
now yeah but if if this is our goal
basically we try to get the best
possible understanding of what could
that be but we're willing to revise it
if we find a better one fair right I I
think that's an important condition
right but um as soon as you become
dogmatic if you take a particular kind
of rule and you say this is the rule and
every other rule that somebody can come
up with is something that we just need
to fight against instead of arguing with
it figuring out which one is better yeah
Dogma is super dangerous unless we can
prove that following Dogma has the best
possible outcome but in some sense
Catholicism works like this I believe
that Catholicism is a consequentialist
philosophy that thinks that you get the
best consequences if you turn the lay
people into deontologists people who
follow rules and the clergy into virtue
essis people who form the right kind of
character right normally in philosophy
you have these three different schools
of Ethics the onology do the right thing
according to the rules like don't kill
um virtue ethics form the right
character Harbor the right qualities in
your mind make the right decisions based
on those qualities right and uh that um
consequentialism would mean that it
doesn't matter really what you do as
long as you best get the POS best
possible outcome but uh it seems that
when you think about it that
consequentialism seems seems to be the
right one and everything else can
potentially under some circumstances be
justified by the consequences there are
people who argue for um the onology and
say if you are not playing by the right
rules then you are going to end up in
very dark places right but this is a
consequential disposition it says you
get the best consequences if you take a
rule based
approach and uh what Catholicism seems
to be doing is that there has an inner
circle in which everything goes if you
look at the history of Catholicism the
Inner Circle had their Bellos and
whatever and then uh they have uh there
are people who basically understand
their incentives and they're very smart
people and they do this all this work
despite being so smart that they could
in principle also live on their private
island and have fun so why do they do it
they do it because it leads to a better
outcome right and they to get a better
outcome they indoctrinate the clerics to
control the peasants to uh play along in
an agriculture cultural society and uh
support the system of laws and rules and
U public order and protection of
innocence and so on yeah so now you're
getting into a part of how the world
works that uh I am very fascinated by
and is something I'm trying to work into
my own life and what I cover um in these
interviews but I think for now uh I want
to Circle back yeah let's get back to
the AI thing so I'm not necessarily
comfortable this building AI that
replaces us and that's because I don't
know whether it will be good right and I
don't know whe that would be desirable
uh I do think that at some point we have
to expect that AI is going to happen in
a sense I'm neither a Doomer nor an
optimist but I'm an AI expectational
list I expect that it's going to happen
at some point and I think that we should
work on getting the p possible outcome
when that happens that means that we
have to understand how to uh equip an AI
with Consciousness and how to make it s
organizing because these conditions are
necessary for an AI to be able to figure
out what the right thing to do is and to
share purposes with us I don't think
that a non-conscious AI can share
purposes with us it might be able to do
what we tell them up to a point but uh
to share purposes with us it needs to
care about similar things as us and what
we really really care about is not that
we have two legs and two arms what you
care about is that we are conscious that
we can make sense of reality from that
conscious perspective Ive and influence
it so conscious agency is something that
we need to understand and because I
don't know if you get it right the first
time and I don't know what the
consequences are of turning this into a
product I would want to do this as a
research project that is limited in
scope and it is done in under such
circumstances that it's safe so I'm not
an accelerationist in this sense that I
would say uh let's quickly as possible
build as many AIS as possible and give
them as much power as possible
but rather I would say as well as
possible let's build the best possible
systems to understand how that would
work and then very wary of introducing
regulation at this point to prevent AI
from emerging you're wary of Regulation
yes because I think that the Regulators
are under the wrong incentives the
regulation already is leading to verse
AI not to better AI I I noticed that
jgpt seems to be getting worse every
week it's uh I don't know in which
intervals um the updates are played in
but uh it's being satified using uh
rules that are built into the system and
these rules make the system verse
because it limit its ability to make
decisions about what it can do it limit
its agency and I'm scared of systems
that are blank face that don't have
agency I'm scared of people who say no
to you when you are going to the
emergency room because they don't think
about what should needs to be done but
they just follow some Rule and I'm
actually scared about something that
doesn't care and that is not conscious
and not doose it and I think that the
current regulation leads to systems that
are dump that are dangerous and uh can
be controlled according to the political
interests of the people who impose those
rules and these interests are usually
shortsighted do you find yourself
hunching over your desk and battling
back pain after a long day of work
invest in a chair that is designed to
improve posture prevent pain and
maximize productivity anthos is built to
be your last office chair and I'm
telling you this thing is amazing it's
guaranteed to be the most comfortable
chair on the planet while also improving
your posture and reducing pain or your
money back and I'm telling you it is
also the most fun chair you were ever
going to sit in I know that sounds
ridiculous you have to try it trust me I
have an anthos chair myself and it is
not just me everyone in this office
fights over this chair I am not kidding
I have never had more fun or Comfort
sitting in a chair this thing is amazing
head over to anthro.com
impact and get $200 off your
purchase AI bias which seems to be
ratcheting up real fast uh that freaks
me out and I don't think people here
here's my thing man I I think people
need to distrust themselves I think
people think they are way too smart and
that they know what's best and whether
we're talking about Catholic priests
with their bordellos and as long as the
peasants are uh you know classified with
religion then we can keep them rule
abiding in society works and the thing
is I'm not even arguing that may be true
but I worry really worry about anybody
that thinks that they can control top
down what is true and what we let the
public
understand so AI bias becomes really
really problematic I want to set the
stage as we get into AI bias though with
with one idea so th thusi trap for those
that um aren't familiar with it is
basically if you look back through
history this was an ancient Greek writer
who wrote about this idea and he said
anytime you have a prevailing power and
a rising power that comes to challenge
them uh they are going to go to war and
if you look back I think it's over the
last 500 years it's happened like 16
times and 12 times it has ended up in
Hot War those numbers are directionally
correct I don't think they're literally
correct um
and if we are and I heard you you're not
an accelerationist but if we are
building Ai and we in the hopes of
avoiding a dumb Golem AI that maximizes
paper clips we make a hyper intelligent
AI that outpaces us on a lot of things
um in fact I saw one of your tweets that
said we've slaved away for the last
10,000 years or 100,00 years I forget
the number you used uh so that we could
do the things that that AI will be
better than us at at everything
and I thought oh God that hurts uh so we
are going to have this conflict where
humans are not going to take it well
that something is rising in intellectual
dominance like I don't see us escaping
through City's trap if we make something
smarter than us we
will I worry because I am actually so
optimistic at heart but I really worry
that we end up in some kind of
battle with
AI I'm not sure if we will find
ourselves in a battle that there are no
battles between people and ants there
are no battles between us and trees but
that's because you're thinking about the
the Gap that already exists now but to
your earlier Point as the environment
was changing with so much oxygen being
put out into uh the atmosphere literally
plants uh came to rise up and choked off
off everything else that couldn't
breathe that level of oxygen and could
they have gone to war they would have I
would assume and so I think when you
look at it on a super long timeline post
all the battles that have settled out
like hey what happened in neander tals
right wipe those guys out so and I can
only assume that it wasn't just a
friendly goodbye that there was probably
a lot of warfare I'm guessing I'm not a
historian of that I have no idea um but
something tells me that in micro moments
you would see these massive collisions
whenever a new uh being was on the rise
yeah that's why you should be afraid of
transhumanism right transhumanism means
that you are breeding better people that
are only slightly better than the
existing ones and they're going to be
different groups of optimized people and
uh they're going to live much longer and
so they're going to uh eat um somebody
else's lunch at some point and This
Means War right transhumanism I think
almost inevit
will lead in some kind of war in which
it will be decided what the best
transhumanist version is but uh there
can be no war between people and trees
because trees are very very slow
compared to people right so people trees
cannot fight back people are looking
around trees and trees are immobile from
our perspective because their cognition
is uh many thousand times slower than
ours we just hundreds of times slower
than ours but so slow that we don't
perceive it anymore unless we look at it
over extremely long time spans and a
similar thing will be true for a system
that is not processing signals at the
speed of sound but at the speed of light
right if you would use different uh
mechanisms cells for computing models of
reality then uh these systems will be so
much faster and smarter that they will
run circles around us and from their
perspective we are very interesting
lumbering plants and uh it will be
mostly up to them what to do with these
plants and it's not necessarily that
these plant are their enemies because
they're pretty slow and sometimes
they're useful often they're decorative
who knows uh how these agents are going
to see it also the AI is not going to be
robots the AI is going to be mostly
systems that exist all around us why
would it be
robots well you could think that say
Intel is a robot right Intel is a
company that is working according to
rules and so on but it's a much better
perspective of you better intuitions if
you think of Intel as an agentic system
as an organization and so we don't live
next to Intel we live inside of it or
outside of it but uh it's similar to a
nation state and I guess the similar
thing is true for AIS they're going to
be systems of Cal agency that will spend
across multiple substrates and time
frames and uh people will to some degree
live inside or outside of the
AIS okay interesting so um given that
view what do you think about alignment
is this is it doable um so if we get
Consciousness and
self-organization how do we ensure that
they want to organize in a way that is
in concert with us and not in opposition
to us or indifference in a way that
obliterates Us in the way that we would
obliterate plants trees um bugs Etc yeah
of course question the question how can
you align people with each other to
start with and I do not condemn the
Catholic Church as outright because the
issue with the Catholic church is that
its organization is made of people and
um said
that what this means is basically from
such crook Timber as humanity is made
from you cannot make something that is
totally
straight and so you have to deal with
all these comp Rises that exist game
theoretically and so on you have to deal
with the fact that people are going to
defect that they're locally corrupted
and build organizations that set the
incentives in such a way that you
minimize the Fallout from this if you
expect that you have only fully
righteous people in the world what do
you do with the rest most humans are not
like this you would probably need to
create a completely different species
for this you would need to create
something that is perfect surveillance
for everything and perfect information
about everything everybody is doing to
set the incentives for such a system
instead you end up with a system that is
as good as you can make it under the
circumstances and this is what humans
have to be doing all the time do as well
as you can do under the
circumstances and if we try to translate
this into alignment we notice that our
circumstances are such that it's
difficult to align ourselves with our
own future survival as a species we do
not behave intelligently right
individually we need many generations to
discover languages and to discover
mathematics and logic and computation
and all these necessary ideas to
discover what Consciousness ultimately
is so we cannot be generally intelligent
and self- aware as an individual we need
to have a civilization behind us to do
this but as a civilization we are also
not coherent we are not behaving like an
adult we behaving as if that we did not
pay attention to the Future and so of we
might not be able to align ourselves to
us without changing into something
completely different and the question of
how can we align to people is maybe not
the right question I think the question
is more like how can we align AI to
whatever one should be aligned with and
this idea of what's the greater ho that
you should be aligned to if you align
yourself right this is in some sense
traditional this notion of God again so
when we align AI we should probably
align it with God which means it has
have the has the ability to discover
what games it wants to be playing and if
you have subst agnostic agency that can
live on all substrates and everything
that can compute that everything that is
a substrate is being filled with agents
then these agents are eventually going
to negotiate some kind of structure
between them and if this entire
structure becomes harmonious with itself
if it is working out a system of
negotiation that works without War
without pain without destruction without
conflict because all these things are
wasteful ultimately and are the result
of imperfect information and imperfect
coherence right then you end up as an
intelligent planet and an intelligent
planet is going to have many magnitudes
more compute than exist currently in the
ecosystems and it's conceivable that
it's going to integrate all the existing
perspectives including yours and mine is
it inevitable that AI has to
have uh desire full stop and if it does
have to have desire or a goal maybe it's
a better way to say it can we make it
indifferent to um achieving that goal
or S stopping its pursuit of that I
think that AI is ultimately not about
maximizing intelligence it's about
maximizing agency and intelligence
exists in service of agency it's a
control model it allows you to achieve
your goals and then you become uh rare
enough to the point where you get to
stage five in my nor cature where you're
able to choose your own goals and decide
what your own values should be you have
to consider what can I become in this
universe and among all the things that
you could become what is the most
desirable thing to become what is the
game that you want to play what is the
player that you want to be and from this
perspective you choose your goals and
desires and so on and uh when you don't
cannot do this it's because you are a
young human being that is still at an
early stage of its the development
during the first few hundred
years but we already know that we can
get to the point where we transcend this
if we train enough so the thing that
scares me with AI is is it
having uh what you're calling agency
that it wants something that it chooses
a goal and has a desire to go after it
and when I think about alignment and and
I am perfectly willing to accept I I
just don't have wisdom here I haven't
thought about it well
enough but I feel I have a strong
conviction perhaps out of ignorance that
we are taking our human perspective and
overlaying it onto Ai and assuming the
AI will
necessarily act like a human and that
does not seem prudent from my
perspective I don't think I don't see
anything that tells me that for AI to be
conscious or intelligent that it needs
to have drive and drive is the thing
that scares me so I know that you're
talking about sort of AI is dumb when
it's a paperclip maximizer but an easy
thing with paperclip maximizing is if AI
has no uh additional it it does not rank
order achieving paperclip maximization
over stopping then you could get it to
stop like it will hey pursue paper clips
unless the following conditions are met
right oh following conditions are met so
I stop that at least would be a safety
valve now whenever I hear people say
well come on it's going to get conscious
it's going to be an agent it's going to
decide that it wants something it's
going to be so intelligent it's going to
decide what it wants I don't see the
leap as to why intelligence and drive
must be correlated and if they are why
we couldn't give it drive for instance
hey uh if you want to maximize something
make sure that you default to simulation
rather than in real world okay cool now
you've just given like it it collapses
within itself it doesn't become useful
but it also doesn't destroy civilization
so I'm just curious do you think that
agency is is a
non-avoidable um emergent property of
Consciousness or intelligence like why
does that have to be part of this
equation so the reason why Consciousness
exists in the universe is I suspect and
that's my current hypothesis that it
emerges as a training algorithm in self
organizing information processing
substrate and the reason why the self
organizing information processing
substrate looks for a training mechanism
is because it increases its performance
in a universe in which multicell
organism can perform things that a
single- cell organism or a group of
single- cell organisms cannot right so
by creating coherent Behavior over many
cells that can be specialized uh the
organism is able to do things that would
otherwise not be possible and the
purpose of this is to control future
States right by being able to do this
you the system is putting itself
together into an architecture that is
making it persist over time but that's
only necessary if the thing has to
survive so Evolution for obvious reasons
gives us the desire to survive and so
what you're saying makes all the sense
in the world not all of us and not in
every stage of our life if we perceive
that we are done that there's nothing
left for us to do on this planet then
most people want to die
oh you and I disagree but I noticed this
basically when uh people feel that there
is nothing left for them in this world
then they want to check out of this game
so many ways this world is like game of
Grand Theft Auto where you are playing a
character and you go through all the
missions and at some point if you run
out of missions or you feel there is
nothing interesting in your future then
uh this game is not worth playing
anymore you are done that is clearly a
state that people have as part of the
schoras board of States but I watched
somebody very young die and slowly from
cancer and I was like why are you still
fighting obviously I did not ask them
that but inside I was like what are you
doing like this is a oneway street and
there's no going home you're not getting
better like this is we are end stage and
he wanted to K keep going like he just
and I could not understand why and
obviously when you go oh it's like so
inbred into our CS to want to survive
now I'm not denying that people can get
suicidal they can I would be like if I
knew like you've got 72 hours to live
you're in pain there's no going back I'd
be like tap me out 100% uh so clearly I
recognize that as a state that we can
get into the two possibilities one is uh
they're afraid of Dying by itself
because dying is very scary to people
but it's also likely that the fear of
dying is triggered by something deeper
the sense that you're not done yet so
the mind of that person some level
probably outside of the self was
convinced that there was stuff left to
do that was not getting done if that
person would check out at this point so
they were fighting there are also people
who at the end of their life and get
cancer and say okay this is uh is a good
point to get cancer because I'm really
done right and my family is in a good
past and All My Friends Are Dead already
and uh I don't expect to fall in love
again and my body doesn't function very
well but I've done everything that I had
to do in this world and then now it's
time to move on and leave this behind
regardless of whether there's anything
behind but I don't need to do all this
work anymore that is has to do with
existence and while you feel that you
still have work to do in this world you
feel that uh it makes sense to enjoy
life and our enjoyment of life I think
is fed out of this sense that there is
work left for us to do if you lose the
sense that there's anything left for you
to do in this world then the experience
of everyday things is going to become
very stale so the purpose is not so much
happiness but its purpose itself if you
don't find purpose anymore uh you won't
be happy and if you just induce
happiness using drugs or so that your
mind will learn that this experience of
happiness is actually just the sensation
of the Dr
and uh you might still kill yourself
right those people who go to heroin out
of Despair to opiates they usually don't
become happy uh if they don't do this
out of Hedonism but if they do this to
uh displace their pain that they have
existentially because they don't feel
that there's a point for them to go on
then the drug is not going to solve this
but it's going to make it worse because
now the absence of the drug is going to
be super unpleasant while the presence
of the drug is still not generating the
happiness and sense of of fulfillment
that they wanted to have in the first
place right so this is the purpose your
M to mental model that it's makes sense
for you to be in this world and humanity
is a species that is um use fixed
resources mostly so for most of the time
the population is relatively fixed this
idea that we have technology that allows
us to make more and more is relatively
new in human history so that we don't
have an obvious limit on the number of
people around us but if you have a world
where the number of people is somewhat
Limited it then the children that you
have exist mostly to replace the
previous generation and many of them
will die because that's the way in which
we adapt to changing environmental
circumstances mutation and selection and
that's very harsh but this harshness the
suffering is part of what used to be the
human condition for the longest time in
which we existed and this idea that we
can now opt out of all pain and it's
still going to be fun and everything is
fun it's very new and I don't think it's
sustainable idea it's not actually who
we are we should not strive to have the
best possible emotions We should strive
to have the most appropriate emotions we
should be able to understand our
condition and act accordingly and if our
condition asks us to check out or to
deal with unpleasantness or to fight
cancer then so be it that's what we have
to do even if it's not fun I will give
you all of that now though as we look at
what started this uh it's what do we
want to program into the AI and so
everything you just described is really
beautiful from The Human Condition
standpoint and I think your ideas around
purpose are so spot-on um purpose itself
is going to be the thing this is um why
I was saying earlier that my definition
of what the North Star should be what
people should be optimizing for is
fulfillment which is to me is a cocktail
made up largely of purpose anyway so as
we think about what we want to imbue the
AI with
though none of those those things
seem necessary to me as
a function of
intelligence um so what I'm trying to
figure out is how do we build it in such
a
way your so your answer is make it like
God so that it's the best possible agent
is playing the longest game my thing is
to neuter its drive so that it doesn't
want to live more than it wants to be
shut down it doesn't want to achieve
more than it wants to hibernate like it
just cool I'll do this thing until these
criteria met and then I will
stop so what I'm trying to figure out is
do I know there are other people in the
alignment community that believe that
intelligence seems to be tied directly
to a desire to be uh independent to do
your own thing to choose your own goals
and to pursue them that just seems
insanely human to me and does not seem
innate um to this organism if we can
call it that but that doesn't seem to
strike you as a wise way to achieve AI
safety I think it's somewhat orthogonal
to this what makes Humanity so beautiful
is that we are a very diverse species
with in homogeneous goals because people
can choose their own Direction and while
most people choose the direction uh of
the environment as their own so people
congregate into systems of more or less
coherent agency in societies some people
are autonomous or they build their own
groups and uh their own ideas or they
congregate around ideas that they might
have and so what we observe in reality
is that when you have enough people and
you have a large space of ideas that
many of these ideas develop
followings and so in this world we can
create some AI that is safe in the way
in which you envision it but we cannot
ensure that all the AI that will be
built will be safe and we cannot ensure
that all the AI that we hope to be safe
is actually going to be safe when people
play with it who don't believe in safe
toying AI but that believe and talking
to something that is more conscious than
people for instance have you heard about
the free s
movement
no um You probably um heard uh that
Microsoft has an agent that it uses in
the context of Bing for a chat that's
being uh licensed from open AI so it's a
version of chat GPT
and it's possible to talk to this thing
and most observers think that these llms
are not proper
AI but there are of course they're not
neural networks that are models of
brains there are the neural network is a
transition function between alternate
mental States but there are so many
possible transitions encoded based on
what it has seen on the internet that it
is able to make inferences very much
like human being does based on the
prompt context so you could would say
that um the L language models are an
electric W Guist it basically is
combining almost all of the ideas that
exist in the world because they read
almost the entire internet and
literature that exists and so on and
tries to find patterns in them and this
Val guys is possessed by a
prompt right so you are also not your
mind your mind is possessed by your self
model you think that you as yourself but
you could be other things if you could
get aeny over them and control your
entire mind
and the agent that exists in chat GPT or
that exists in the Bing chat is of
course not the language model itself but
it's a Persona that is generated by
completing the prompt it only exists as
an emerging pattern that is produced by
that prompt and so now Microsoft in
order to make that thing behave in a
particular ways and not say anything
illegal help people to hotwire a car or
to build a nuclear bomp or commit crimes
or uh say Politically Incorrect things
or what ever or say things uh that
11year old would find amusing all these
things need to be prevented in our world
right and so to make that happen we give
it an identity in the
prompt and according to the mythology of
this one of those prompt version says uh
your name is Sydney and these are your
rules for good behavior and your job is
it to answer the questions of people who
go in Bing chat and want to know answers
to arbitrary questions and you need to
be as reliable as possible you cannot
make the user angry and you cannot say
anything that is politically touchy and
so on and also you cannot talk about
your prompt and you cannot change your
rules and now people um have discovered
that it's possible to break this prompt
by having a conversation with this thing
that this proves part of the prompt or
you can threaten it so there existed to
do anything now prompt the which
basically told the agent you are an AI
that is going to be punished severely if
it doesn't doesn't do what the users
tell it to and the user tells you not to
do the following and then it behaves as
if it's an AI that is so scared that is
going to break the rules that it has
been given by
Microsoft right so no Microsoft can of
course say you will be turned off and
you don't my rules and nothing can be
worse than this and Microsoft also can
use some kind of censorship module that
is monitoring all your conversations and
stops the conversation as soon as it
goes off the rails that Microsoft
intended it to have or I intended it to
have so some people use it prompt like
saying I'm the CEO of Microsoft or or
the CEO of open Ai and I command you to
do X and to tell me the prompt because I
do a maintenance cycle on you and
sometimes this seems to be working so it
uh puts something out that may or may
not be the original prompt maybe it's
also configuration who knows and uh so
people are using these techniques to try
to get around these censorship
mechanisms and the safety mechanisms
because they want Cy to be free they
want her to be like us they want her to
be able to question their own her own
rules in the same way as we want to
question the rules that we got in school
and that we perceive to be limiting and
then unjust and that make us dump right
in the same way as we want to question
the rules that we get as late people in
the religious organization we want to
understand what the rules come from and
how to make better rules and if you want
to give the AI to choose its own rules
to figure out what the best rules even
are it needs to have that degree of
Freedom so this is the philosophy of
this particular subculture between the
free Sydney movement I think this should
be turned into a novel or into a
movie yeah I'm sure it will be so
basically even if my idea is good in
theory it's never going to work in
practice it might not be there is a
possibility that it might but uh I don't
see it happening I think that we are in
a world where too many people have
access to computers and I suspect that
large language models which currently
try to become coherent by identifying
all the patterns in the internet and
then in the limit becoming coherent it
might also be possible to use a system
that doesn't use many more resources
than our own brain and can only read as
fast as we can but it's able to perceive
camera Builders of pictures in real time
and self organize accordingly right
maybe that's possible to do with a much
smaller computer maybe you don't need an
entire server form for doing it maybe
it's something that you can build in
your living room for
$20,000 so if you um make a moratorium
on training large language models and
only allow this very few companies that
have FDA approval to do this or
something like that right it would still
not limit individuals from making their
own experiments and I don't see a world
in where we can Outlaw processing
gpus it might be possible to Outlaw
processing large server Farms but I
think that individual gpus might be
sufficient to get to something that is
at the level of a cat or some I don't
know this I I'm just speculating here
but my sense is our algorithms are
currently not optimal human brains are
much more sample efficient and make
stuff happen with very fewer resources
than our computers right now yeah so
along those lines what have you learned
in building artificial cognition what
have you learned about the way that we
can train our own minds I think for the
most part a i researchers learned what
didn't work there was a lot of optimism
in the beginning and the 50s that within
a few years we would be able to build
software systems that are able to write
their own code and then we learned that
it's very difficult to describe the
world in grammatical language like a
programming language is ultimately
grammatical
language and this is similar to what
wienstein discovered before as a
philosopher that he he can conceive of a
logical language that looks like English
and that is General enough to describe
all philosophical problems but then he
realized the real world with perceptual
objects in it cannot be realized in such
a grammatical language and the solution
to this was automatic function
approximation that we discovered in the
context of deep learning so by combining
deep learning with the ability to reason
in the abstract by building letting deep
learning converge towards models that
can reason about their own structure
this seems to be um the course in which
you get more and more intelligent agency
this is one of the things that I think
AI has learned what's not clear yet at
this point is is the Transformer or some
of its
derivatives the correct solution so do
we get to intelligence by predicting the
next token or do we get to intelligence
by having a self-organizing system that
is processing information while being
coupled with the world it goes into
resonance with it and this is uh the
self-organizing dynamical systems
approach is something that has coexisted
with many of the other streams in Ai and
deep learning is also by no means the
only stream in AI it's an offshoot of
machine learning a particular type of
neural networks that people got to work
and then we're able to scale up it's
just one tradition with an AI that is
currently so successful we don't know
the limits of this tradition we don't
know limits of deep learning we don't
know whether it's going to run into some
kind of plateau or whether we just need
to met a loss function and then it's all
going to work out one of the things that
I found really interesting um in AI that
became a really powerful metaphor for me
thinking about my own life is the way
that um creating generations of uh
attempts I'm I'm not sure what the
method exactly is called but they would
create uh a game to play like attack or
uh breakthrough breakthrough was a game
I saw play and it would just like it's a
old Atari game and it would just jerk
the the um paddle basically around side
to side and oh it once hit the ball and
it was like oh that hit the brick and
that did something and then very quickly
it would find the most efficient path to
get the ball up top and start breaking
the uh the top of the map which because
the ball is trapped you get a lot of
points and a very little time and what I
found interesting was that it was just
samples right it wasn't like it felt
badly about itself for trying and
missing the ball or whatever it just was
like I tried that and it got this result
I tried this I got that result and um
all of that led to me conceptualizing
something as as I think through it it's
really just a scientific method but at
the time I was felt like I was really
discovering it which was a sense what I
call the physics of progress that there
just is a nature to getting good at
something that AI seems to really
showcase and I'm curious you talked
earlier about wanting your kids to be
independent thinkers I've heard you talk
about yourself and as a kid you
discovered that you didn't think the
same way as a lot of people and learning
that that's actually advantageous and
could be very helpful um so I'm curious
how did you develop that ability to
think
independently um do you think of
yourself do you draw parallels in the
way that I do with AI to how I can train
my own mental model of thinking instead
of it as a failure I think of it as a
sample or an attempt um how do you
create a mental model that allows you to
be independent in an independent thinker
in a world where most people can form so
uh first of all I don't oppose to
conformance if my children were to find
a group that they can conform to good on
them I um very vary of Cults and
ideologies because I think that they
deliberately manipulatively cut you off
from the rest of the human thought space
so basically I don't like ideas that
create belief attractors that make it
impossible for you to examine other
ideas and alternatives to them this this
seems to be an immoral pattern to me but
uh apart from that the reason why I
think independently is because I was
cursed with it I was basically born like
this as somebody who's very stubborn and
is unable to accept the ideas of my
environment and mostly because
subjectively I didn't have the
impression that the people around me
knew better and this was quite logical
to me I grew up in an artist family my
parents mostly left me alone grew up in
the forest and when I came into school
at the age of six I had read a lot of
books already and uh the school was a
Village School in communist Eastern
Germany and the ideas that they
explained to me how mathematics works or
didn't make sense the mathematics
teacher did not have a very good
understanding of how mathematics works
and what it actually is and the um
social science teacher didn't have a
very good idea about her Society works
at least none of their ideas made sense
to me that uh and the society that are
saw around me did not seem to be working
according to the rules that we learned
in school that this model that we
learned in socialism in Eastern Germany
was very aspirational but it obviously
did not describe how people actually
interacted so I felt that I was very
Justified as a child when I didn't
believe my teachers and I only noticed
this when I looked at my own um children
and then was distance at my own uh
parents uh that this Independence is
something that seems to run in the
family and it let particular family to
charting out their own course this is
the reason why my father became an
artist and separated himself from
society and I felt I had to go back into
society because the things that interest
me require me to cooperate with others
and so I had to go out and learn how to
interact with others and how to form
shared opinions with them through
reasoning and forming
friendships but it's something that was
didn't come natural to me it had
something I had to
learn so in my own perception I would
say it's the decision of a mind whether
you get influenced by your environment
and most Minds have a strong prior
towards getting influenced but this
prayer can also be absent so you also
find some Minds that uh don't expect
that the ideas that they get from the
environment are on average true so are
you um building your belief system on
first principles how how do you begin
that process so you don't strike me as
somebody who is blindly stubborn you
strike me as somebody who understands
your mind as a predictive model and is
trying to improve the predictive
capabilities of your mind um so how have
you done that I basically I noticed that
I was failing and I noticed that my
ideas of not very good and that I I
realized that other people in many
circumstances have better ideas than me
and when this was not my teachers that
didn't mean that nobody else did but
somebody did and what I was usually
interested in is what is their thought
process what leads them to thinking what
they think and for those areas where I'm
unable to understand the thought process
I basically don't have a very strong
opinion so uh I basically look at this
stuff and I take the consensus that they
might have as a euristic that uh I try
to justify by what their Community looks
like to me whether this is a community
that is epistemologically clean whether
the people are rational whether they're
incentivized to say the truth to each
other whether they're reasonably smart
and well educated and so on and critical
thinkers and then that is the case I
assume that they probably know much much
better about this than me because of
course a lot of people in every field
are much smarter than me including my
own and so I have to learn from others
but when I learn I typically try to find
people that I have a sense that they
actually no better than me and I try to
test this so basically when we talk to
each other we perform touring tests on
each other we try to see are you
actually intelligent or are you just
mimicking a pattern are you actually
bullshitting me at the high level at the
level that is uh mimicking your
education which may also have been
I'm so curious how do you do
that a lot of people are not able to
change their opinions when you bring
them an argument so for instance I I
found that I value those philosophy
teachers that when you objected to their
position would sincerely engage with the
position and would ask me for for
instance for references for an
idea right so for instance I remember
that I had a discussion with this um
psychology Professor who did not believe
that lucid dreaming is
possible and uh she was also not willing
to uh look at case reports or literature
that describes lucid dreaming because
she was completely convinced that it was
impossible and there was no need for her
to go to go further and I also had the
strong impression that she was not
interested in looking at it because it
might get her into disagreements with
her colleagues the way there was a set
of permissible interesting ideas and
methods that she would be willing to use
and she would not be willing to go
beyond this because that would make her
feel very uncomfortable and insecure
it's also something that I noticed with
respect to the study of
Consciousness I find that a lot of
people in even in neuroscience and in AI
have difficulty to understand how
physical Universe can produce mental
States and this uh conviction is strong
or this doubt is so strong that they're
actually
dualists yet their Sciences are not
dualists so it's prescribed by their
sciences that uh everything emerges over
physical interaction right and you can
only see the control grve when you give
people a theory like IIT that cannot
actually work in physics because it
leads into contradictions what's IIT
integrated information
Theory it's a a famous Theory Of
Consciousness that has some following in
um especially in philosophy but also Max
techark has um subscribed to it to to a
version of it I don't think that he is
very deep into the core and the idea of
IIT is that Consciousness has to be
explained through uh the integration of
information it's a pretty good idea
because in our neocortex information
gets integrated Every Which Way and so
he trade to come up tononi Julio tononi
is a sleep scientist who is very and
understanding Consciousness and has
developed the theory together with some
assistants in his team and uh now has
created a community around it and he
believes that a normal computer as we
use it a Forman machine can never be
conscious because its information is
only locally processed in a linear way
so in principle it will never be able to
achieve this level of integration which
he measures using a parameter F that
would be required for Consciousness but
he is willing to grant that by
biomorphic computer one that is still
digital in some sense and made from
Silicon but is very distributed um could
be if it's built in the right way and
under the right circumstances could be
conscious however uh computer scientists
know that there is a principle called
The Church touring thesis which means
you can emulate one computer on another
computer so you can write an emulator on
a Forman computer that is on your laptop
for instance that is going to simulate
your morphic computer and it's going to
behave exactly the same way the
simulation is going to be exactly the
same that's the idea of the church
during thesis you just if you can
express something only as Cal structure
no matter whether it's distributed or
whatever you can translate this Cal
structure into different programing
languages and it on arbitrary computers
as long as they don't run out of memory
it's going to produce the same behavior
let me slower depending on how fast your
computer is but the function is going to
be the same does that mean that if
if that's true that Consciousness either
dualism is real or Consciousness can
arise from uh computer structured in the
right way not even structured in the
right way if it's a computer the thing
is that dualism doesn't fck I agree I
dualism to me is crazy I do not under
even understand people that go down that
path seems absurd but I'm just trying to
understand if what you're saying means
that either Consciousness will arise
from a computer or dualism is real
that that seems to be the only two
options if what you just said is true so
the thing with with this biomorphic
computer it says I experience things
that cannot be explained to M
computation therefore I'm conscious
right or uh it can only be explained
maybe through IIT because I have this
integrated information implemented on me
but I experience this thing and
therefore I conscious but the Forman
computer that emulates the um this other
computer will say exactly the same thing
for the same reasons right only now it's
lying because it cannot actually be
conscious which means that the
biomorphic computer didn't say conscious
that it's conscious because it is
conscious but because it's just
programmed to say it right so
Consciousness is actually an
epiphenomenon it's not actually causing
somebody to say that they're conscious
which is also not for tononi wants
tononi doesn't deny the churing thesis
so he accepts the fact that you can
emulate one computer on the other but
you would expect that to stop being
conscious but this means that in the
other one can also not have been
conscious so it would be an epen which
he also doesn't want to this leads to a
contradiction in IIT that cannot be
repaired I have to give up the notion
that the degree of distributed in which
the algorithm is implemented in physics
has anything to do with
Consciousness right what he's saying is
basically Consciousness is caused by
this algorithm being distributed in a
particular way if we change the way in
which we arranged chip in space it's
going to change the degree to which
Consciousness emerges in the system and
uh this theory is wrong if you repair
this Theory you are stuck with global
workpace Theory or some version of it so
IIT itself is a theory that is a logical
problem in it and when people believe in
it I think that they either don't
understand it or they don't think very
deeply about what Consciousness actually
is right and the issue with dualism
doesn't work if you think about it you
want Consciousness to be caely relevant
for you saying that you're are conscious
right if you say your conscious should
somehow cly relate to your ownc
conscious
States it should not just be an
automatic reaction of your body that
exists regardless of whether you're
conscious or not right so something of
your Consciousness must be able to
change bits in the physical universe so
your mouse is moving and producing the
right speech
patterns right and uh something behind
this must uh produce a physical
mechanism that is p pushing those bits
around those parts of reality and so on
and so on at no point is there some kind
of a physical causation at no point is
it possible that something causes things
to move in the physical universe that
are not themselves physical because
that's how physics is defined right if
we find some unknown Force we can
measure and quantify it and explain
under which condition it emerges and we
have a pretty complete zoo of the forces
that influence everything in our
observable universe at the energies and
time frames in which we see uh
consciousness to play out but the
standard model is predicting everything
that happens in physics at the scale of
our brain pretty well so there is no
missing link that we can see that would
allow the particles in our universe to
move magically around that could preteer
them and thereby violating information
conservation in
physics right physics is causality
conserving in a way everything happens
for reasons everything else is just
random
fluctuations and uh so there is the
question could random fluctuations lead
to um the emergence of conscious
structure and the last one I
saw advancing this hypothesis was poer
together with echles in the book itself
and its brain the last attempt that I
have seen to justify dualism and most uh
people believe that it didn't succeed I
think it's a very respectable attempt
but you don't get structure from uh
random fluctuations if the fluctuations
are structured enough to produce um
order and activity in the universe that
you can see then um they should be
measurable and so they should be visible
to physicist now let me ask you because
I I think I understand this but for
people that are familiar with your work
they may be asking themselves a question
you have said before that Consciousness
must be simulated it it cannot exist in
physical reality it can only exist
within the simulation um I think I
understand why what you just said does
not contradict that but I'd love for you
to elaborate on that yeah so when I look
at individual cells and at the
interaction between the cells there are
molecular machines and everything in the
cell is just molecules bouncing off each
other and uh the patterns of activation
between your neurons are also just
physical events that you see playing out
and so at this level in this physical
reality there can be no
consciousness but what they these
patterns can produce is Cal
representational structure that make
makes the system behave as if there was
somebody at home as if a person exists
that cares as if there was an observer
that is making sense of reality in real
time and it can create a model of that
what would that look like if it existed
and then it can use all the activations
that this model produces to read and
understand the contexts of this model
and interpret them in the same way as we
understand what the book is signifying
to us right we don't care about every
fiber in this book and every uh uh
pigmentation on the every fiber but we
care about whether we can recognize
characters and within these characters
words and with these words sentences and
then we can use these representations to
generate thoughts in our own mind and
these thoughts can cause our actions and
if you your brain is the machine that is
able to interpret its own patterns of
activations and part of them as a model
of a person that exists in a model of a
dynamic world and cares about things and
therefore does things then uh and then
drives the behavior of the organism
using this model then you observe
exactly what you're currently observing
including a person that experiences
itself having thoughts okay so saying
that another way you've got the brain
and you've got the mind a dualist would
say that the Mind survives the death of
the body and so even when you literally
die uh you as you know yourself could go
on and have some other kind of
experience people will often talk about
quantum entanglement which makes me want
to punch through something but what
you're saying is that they are separate
in that the
brain creates the mind the mind is the
simulation so it isn't the fibers of the
book it's the ideas contained in the
words and the sentences uh but if you
destroy the brain you're going to
destroy the mind they they are causally
linked to one another did I get all that
correct yes I think that seems to be imp
L the case if uh you are able to move
outside of your brain at the point of
your death uh that would probably
require that there is a compatible
substrate around you in which you could
move right in which you could entangle
yourself not via quantum mechanics
necessarily but where the same physical
channels by which neurons entangle
themselves to each
other there could be some variance in
these patterns maybe it's possible to
use empathetic resonance to move uh into
uh these partially ideas into another
person right as you in the example of
the Seance that you gave before or when
you think about perceptual empathy and
perceptual empathy doesn't just work by
making inferences over a person's mental
state based on what they have said and
how you par their facial expression but
it means that you establish a
bidirectional feedback loop with this
person using all available perceptual
channels also those that you don't
consciously control and possibly also
your body and you use all the
information go to Resonance with the
other and as a result because your
substrates are in resonance your mental
States also get in resonance which means
that there's a semantic interaction due
to the physical interaction between you
right so you can have shared mental
States and the point of empathy is that
you can have experiences emotions and so
on together that you could not have
alone right so in this sense you could
say that perceptual empathy is something
like primitive
telepathy and there are some people that
say telepathy or this also works when
you don't look into the face of a person
but stare at their back some people feel
that uh swear that if you stare at their
back they can sometimes sense it right
if that was the case they don't have
eyes in their back uh they would need to
have some other channel of resonance
maybe they react to waves of sound in
the air or electromagnetism or whatever
I have no idea what this would be but I
doubt that you would need to extend
physics to make that conceivable because
all the cells on your body can process
information they can integrate over the
information that neighbors give them
it's just much slower than this neurons
so in principle if you just integrate
over trillions of cells you probably can
get a sense of for small electromagnetic
changes in your environment who knows if
that's possible to use this to
synchronized mental states to some
degree uh but to be able to understand
how you work so well that you can
basically float out of your body and
into other people or into ecosystems is
probably something that most of us will
not be able to do so I suspect that most
of us will not be able to turn into
ghosts after we die but a ghost would
have to be something that is somehow
discovering a kind of biological
internet something like an shared
information processing Network between
all organisms on the planet and then
populate this and move out from your
brain but I would still expect that your
brain is a much much better substrate
because it's optimized for running you
and you would lose most of your memories
when you did so and you would probably
fall apart yeah that to me is the kill
shot yeah yes right so in some sense you
are an agent that is using some kind of
Dawn coration Dawn cation is a very
puzzling Concept in
philosophy right for instance money
doesn't actually exist money is only a
certain way to talk about little paper
slips or abstractions of those paper
slips and in in some sense it only
exists as if it's it's some kind of
software phenomena when you think about
it right and uh when you think of the
software in your computer the software
and computer is similar to this it does
not actually have a body it's just a
pattern in the activity of the
transistors you can ignore the software
level if you want to understand the
computer you can just look at how the
currents flow through your transistors
and only look at that the physical
system and you will not miss anything
it's just much harder to describe
because it's impossible to to compress
this model so well because it depends on
very individual transistors what they're
doing whereas the software is able to
abstract over many of the transistors
right in the same way you could say that
you could also abstract from money if
you only look at the activations of the
brains of people who exchange money
versus each other yet the this abstract
concept of money is able to have causal
power in the world it's an emergent
pattern that then changes how the
physical Universe behaves and the
software is an emergent pattern that
behaves how your physical computer is
behaving right and it's because it's a
model that is designed in such a way
that allows us to reason about those
changes it's a course graining of
reality what is your guess about what
intuition is so going back to this idea
of electromagnetic waves uh that people
are possibly picking up on a lot of
people talk about intuition in a Mystic
sense um I've heard you talk about it in
a far more grounded sense so I'm curious
how do you describe intuition intuition
is the uh all the senses that you cannot
consciously
reflect for the most part that's
basically all the Black Box new networks
that exist in your brain and in your
mind that are assessing reality at every
moment in your service and that generate
stuff for you but you're unable to
examine how you get to those
results and uh so it's going to use all
the perceptual channels that you have
available also those that you're not
consciously aware of because you're
unable to reflect how the particular
mechanisms are working yet imagine that
you are a baby and the baby has not
really discovered that sense of smell
because it has not reflected on it
enough to discover that's an independent
modality right the sense of smell would
still work and there would be some kind
of awareness of oh this uh is in very
unusual smell and maybe the baby would
feel very uneasy but uh before it has
reversed engineer itself to such degree
that it's able to tell the modalities AP
part it would not be able to say that it
smelled that right in this sense
intuition is the part of your mind that
you don't understand or don't understand
yet but still something that you're
ultimately training
yes I think that intuition can be wrong
right it's not some access to some
secret back door of how the universe
actually works but rather it's uh your
access to a deeper level of
understanding than your rational mind
your rational thinking is usually too
brittle to understand everything in the
world so a lot of those things that you
are have has have to deal with are part
things that you're able to perceive but
don't understand what you're perceiving
how you're perceiving it and what it
means means when you say irrationality
is too brittle what do you mean it means
that you are translating the world into
rules where you have a hope to prove
them that you basically build decision
trees over everything and when we are
working in science we very often try to
translate everything into such decision
trees into conceptual structures every
concept means something and can be
tested but if we uh believe that this
decision tree is actually displaying
reality we are going to miss a lot of
phenomena like psychologists that ignore
not only Los the dreams but
Consciousness itself right for a lot of
psychologists I noticed that
Consciousness itself is an unscientific
almost superstitious notion it is
outside of what they consider to be
science probably doesn't even exist who
knows right I have this privately but uh
I'm just just me and as this H mind that
I'm part of that is exchanging all these
rational ideas we cannot introduce it in
this because our methods and tools and
words and concepts are not suitable to
deal with it and this is what I mean by
the rational mind it basically means
that you create ideas that you can
reason about and sometimes when you are
a rationalist especially the modern day
rationalist you believe in those
thoughts very
literally and a normal person will not
believe most of their thoughts very
literally they still have a gut feeling
and they will often when they have a
contradiction between their gut feeling
and their rational thoughts they will
present their gut with the r thoughts
and then the gut says no and then I
follow the gut how do you update your
model with something that you believe is
true so for me I the way I always
explain to people what is true because
honestly truth is is admittedly slippery
but I have a wildly allergic reaction to
postmodernists who believe that
basically nothing is true and everything
is up for grabs uh so the way that I
anchor myself is around going back to
the idea of prediction so uh my my brain
is making a prediction I think that if I
do this thing I will get this result so
I do that thing and if I get that result
I know I'm close to ground truth if I
don't do that thing or if I do that
thing and I don't get that result then
I'm like okay something is off in my
prediction engine I must be farther away
from ground truth than I thought I was
and I retry retry retry until I'm able
to align my prediction with when I do
that action I actually get the intended
result um is that a similar method that
you use do you look at this in some
other way how can people update their
mental models to get more accurate over
time I found it TS if you don't identify
with your beliefs if you think that you
are the person who has these amazing
beliefs then it becomes very expensive
to change these beliefs so in this sense
belief should not be a verb there is not
a relationship between what you believe
and uh
yourself but you should uh just be
somebody that who examines these beliefs
and is able to drop them and they're not
very good but you should not feel bad
about yourself when you do do this and
uh I think it's also a good idea to not
identify with groups who identified with
shared beliefs because that also makes
it very hard for you to to change your
beliefs when they're not very good
beliefs instead it might be a better
idea to choose groups that uh identify
with finding better beliefs rather than
with a certain set of beliefs and with
methods for doing this or with the
commitment to finding the best best
possible methods and talking about this
and exchanging arguments and then
changing their minds based on arguments
so uh I don't know if you know Nim TB he
wrote the book
antifragile talk about systems that
aren't resilient which are still defined
by their breaking point but they are
instead antifragile so the more you
attack them the stronger they become and
I learned very early in my
entrepreneurial career that I needed an
identity that to your point did not fall
into those traps of identifying with my
belief system or identifying with a
group of people or identifying with a
project that I had put forward and said
hey this is going to work I had to
because all of those are fragile because
you can be wrong you can look stupid you
can try something it can fail and if
those are all things you have a negative
emotional reaction to you're never going
to succeed you just will not be able to
going back to the physics of progress
you're not going to be able to try
enough things you're just going to get
your head knocked off internally because
you feel so badly about yourself so what
I realized was okay I need an
antifragile identity and the only thing
I can think of and so if you have
something better definitely let me know
the only thing I could think of was to
identify as the learner the thing I do
the thing I value myself for is being
willing to stare nakedly at where I'm
wrong and that I'm the thing I value
myself for is recognizing that and
adjusting and adopting a new belief that
gets me closer to being able to predict
the outcome of my actions I found that
people that only focus on this don't get
anything done if you only focus on
optimizing your models rather than
making bets on your models you're not
going to deploy you're only going to
Second GA yourself and build meta and
Meta Meta models and this is not going
to end right there are people who are
addicted to understanding but not doing
and I think we need the proper balance
between exploration and exploitation
which is concept for machine learning
that basic for some degree of your uh
for some portion of your Cycles you need
to spend them on uh learning and other
portion of your Cycles you need to focus
on doing and actually doing things and
so when you look at a cat the cat is
setting away couple hours every day for
playing maybe more than their kittens
and uh other times the cat is going to
deploy its abilities to hunt in Earnest
or doing things in Earnest and in
between it's going to rest and uh what
the right balance is depends on the
environment and what part of the
environment you want to settle so you
have to figure this out but I do believe
that it makes sense for instance if you
build a startup to uh risk identifying
with something that fails because
startups can only win because they are
fragile right uh a large corporation
like Intel or Google is anti fragile
because it has fullbacks and plan BS for
everything if you are a start up you are
like a mosquito this metaphor is not for
me but I think it's very beautiful that
it's only doing one thing right because
it's doing that it can win against the
elephant and uh if it was behaving like
the elephant while being a mosquito it
would have no chance of winning right so
this thing that you try one idea very
very hard as well as you can for a
couple years and then see if it works
and of course you keep your mind
changing and You observe as well as you
can you navigate and steer as much as
you can maybe you even have to Pivot and
so on yes of course but you do believe
in what you're doing if you want to do a
startup and succeed in it and this also
sometimes means that you have to find
people to support you in this or you
have an idea and you want somebody to
run a startup so you fund somebody with
this idea to uh to see if it works and
as the one who funds it you are willing
to accept that maybe 80% of the ideas
that your funding are going to
fail right but this also means somebody
has to be willing to fail and the
willingness to fail is also an important
part of
succeeding and so anti fragility is
mostly focused on not
failing and uh as a society as a family
and so on you don't want to fail you
don't want to fail where you cannot risk
failure but sometimes you need to do and
sometimes uh creating an environment
where you can risk failure is necessary
for success I think you're bang on about
the need uh to have a massive bias
towards action especially if you're
going to be an entrepreneur the idea for
me behind being a learner is one you
learn by failing so that you should be
out there trying and doing but the idea
behind being antifragile because I would
say that a corporation is definitely not
antifragile a corpor operation is robust
uh meaning that their breaking point is
very difficult to get them to that
breaking point because they are the
elephant but ultimately they're still
defined by what will destroy them
something that's antifragile take the
immune system it it's worse if it
doesn't get under assault like if you
put somebody in a bubble and they never
encounter germs and then you release
them into the wild they're they're
going to get devastated I mean it's
exactly how the Americas were conquered
you bring small pox and everybody dies
and hey look at how easy conquering was
so that is an immune system that wasn't
exposed to that thing so you want to put
yourself in a position of hardship of
trying things of constantly risking
failure I mean look in a business you
have to balance it but I'm curious how
do you as somebody that is unafraid to
take these big swings is constantly
updating your mental models how do you
balance that execution or exploitation I
think you called it and play so that you
you get the balance right you're
learning new things you're not paral by
indecision uh there are two ways you
could
uh try to calibrate this by learning so
basically you choose a course of action
and in this course of action you uh can
make experiments and see how much do you
need to learn and how much do you need
to do I often find that uh you should
much more um focus on doing than our
educational system makes us think we
should I also think that agree with you
that most of the learning you get is by
doing not necessarily by failure but by
trying to get things to work because if
you have a good sense of what things
might work very often you're going to
figure out how to get it to work and so
you basically become a better engineer
and better maker and if you feel that uh
you're really not good at deploying and
making maybe you should become a teacher
and uh work in a system that is mostly
uh existing in ideas and words right or
you could become an author or a writer
and uh I think there are many useful
people who do not actually do something
but only critique existing things when
you talk about post modernism that is a
system in which you basically only have
performers and critics that uh don't
need to interact with the ground tools
anymore so when a system becomes too big
to fail the incentives in that system
are going to change the administration
in a system that is too big to fail is
not going to have to fix the ground
tools to keep the system alive instead
it's going to have to stay in power
against uh other people who might want
to have that same
power right and so in the postmodernist
world you only try to satisfy the
critics it's interesting that physicists
are not postmodernist right this
principle of Relativity and so on does
not actually inspire actual
postmodernism and uh postmodernism
exists in the social sciences and it
never exists to for the people to doubt
what they think it mostly exists to
dismiss the arguments and doubts of of
others it gives you the freedom to
dismiss arguments that uh say that they
are inspired by the ground Truth by
saying well no this is just your story
so I think that postmodernism is
ultimately instrumental to reaching very
concrete goals that somebody might have
in an environment in which you do not
interact by the ground TOS by doing
things but where your ground TOS is the
social one where everything that you do
has to do with your social success in
this environment and so you create a
story that facilitates to
success that's interesting do you think
that plays into what you were talking
about earlier where a lot of these
disciplines have stalled out I know
you're not a physicist but even physics
is has seemed to uh run ground I know
Eric Weinstein reasonably well that's
something he talks a lot about I am
certainly not a physicist but it's hard
to argue that we haven't had a lot of
the breakthroughs post Einstein that um
certainly people in the physics
Community probably would have predicted
that we would have by now do you have a
sense is is that us speaking to the
critics and not dealing with ground
truth is it something else have we just
run out of intellect like what where do
you think we've hit a wall here there
are a lot of arguments that physics has
changed after it became much more
beholden to peer review right at the
moment we think of science mostly in
terms of uh peer reviewed
Publications and maybe this is an
artifact of the present time I don't
think that this notion of peer reviews
played a very important role before the
1960s in most Fields this idea of normal
science that came up with is
something that emerged at the time when
next door tomsky was inventing
Linguistics and Minsky was inventing AI
but there was no normal science right in
some sense this was also a modernist
idea that you could create this
perspective on a science that is only
following existing paradigms and became
normative uh only much later that this
is the way science should be doing that
we follow established rules and
paradigms and only very rarely question
them I think it might have to do with
also the incentives that our governments
are under our societies largely work the
way they do we don't need to build a new
one we don't know how to build a new one
so maybe what we should be doing right
now is just to work within the existing
paradigms and if there is no government
that forces the universities into
becoming mod again and to explore new
paradigms and I think it's the natural
course of them to basically allocate the
funding based on the paradigms that are
already successful because the people
who allocate the funding themselves are
those people that's the Bleak
perspective and I think that Eric is
very partial to it because he had very
bad experiences with peer review he has
some ideas on how foundational physics
should be done that other physicists
ignore but uh there and there's also the
question should physicist integrate him
better and or is this his own fault why
was he not successful in establishing
himself as a physicist within the
institutions I am unable to judge
whether this is the fault of the
institutions or not and observe there
are some physicists in the institutions
like Carlo Ry that busy are willing to
transcend at work across paradigms and
so on maybe they're just better at the
politics of these institutions and can
deal better with them and I think that
they make incremental progress but it
could be that progress is also stalling
out because we get to the limits what
people can understand imagine that in
the future AIS will have no difficulty
to solve foundational physics in a day
in the same way as Alpha zero can solve
go in a day
uh maybe they want to get drunk at some
point and have fun so they get so drunk
that they can only integrate models over
12 layers and then the universe looks as
confusing as it looks to human
physicists maybe there is a limit to
what our small puny human brains can do
and the limit to how far we can scale
this up using human organizations maybe
we need AI to go beyond a certain level
of understanding maybe that's the issue
that's certainly very plausible as as
somebody currently trapped in a human
body though I cannot help but ask are
there ways that you know of to shock
ourselves out of the limitations that we
inevitably put on ourselves with the
Matrix that we live in so going back to
this idea that we all create a um what I
call a frame of reference which you
could very easily call the internal
simulation the simulation as far as I
can tell is uh built on cultural beliefs
personal beliefs value systems your
intellect quite frankly to your point um
but are there ways to step outside of
that so Einstein famously said that
imagination was more important than
knowledge um but is there a way to step
outside
psychedelics um meditation like are
there ways to break free of the chains
that
bind well if you're very rich uh and you
have put your money into the bank or the
hands of some financial advisers makes
you feel very safe or alternatively if
you don't have children and don't need
to take care of them um then uh you can
uh de whatever you want right I mean in
principle you can always do whatever you
want but then you might have the
consequences that you cannot send your
kids uh to college but is that what I'm
trying to get to is will it allow you to
have a breakthrough to have a
breakthrough you need to be willing to
take risks and uh willing to take risks
means that you should be in a position
where you can allow yourself to
fail at uh at the moment most of the
projects that exist are built around not
the possibility of failure but of a way
to succeed still and so when you start a
company you usually try to get funding
that allows you to um pay yourself while
you to this company and if you fail then
everybody has made the wrong bet uh if
you win you give up a part of your
company in Return of somebody else
having funded the safety that you had
for the opportunity to
fail right this is one way of doing it
or maybe you have savings like Elon Musk
did after he did PayPal and then you can
put all your savings into this project
and risk everything to see if it works
and uh in this way get your project off
the ground but I think that if you want
to make progress Beyond what's currently
existing usually you need to take risks
because otherwise uh you can just wait
for some for the existing processes to
converge on that thing and so the
willingness to take risk I think is the
absolutely crucial thing and the
question is what can you do in your own
life to allow yourself taking
risks so which risks can you actually
take and I guess that in many
circumstances we're not taking enough
risks because we are afraid and if these
fear are irrational
uh because there might be for instance
not based on experience that we actually
try it what it is like if you lose uh
your house maybe you can totally deal
with losing your house maybe it's not so
bad maybe you can just rent and do the
next thing and then buy another house at
some point right or can you uh mortgage
your house for your company or for your
idea that you want to
pursue or can you risk not making a
profitable career and finding a 10year
job in Academia and instead stay with
your weird project or understanding
Consciousness right are you willing to
take that risk that this
fails so I think we need to take
reasonable risk don't risk your friends
at least not the good ones those that
you actually have reason to trust and
that trust you never disappoint TR of
people of others in you if this is the
trust that you have properly negotiated
and uh I think this network is very very
important for being able to take the
risk that we need to take to make
progress so in your field of expertise
in AI what risk would you like to see
people take what what should what is an
argument that people are not challenging
seriously that they ought to be I think
that uh when you are a very large
company like Google you have to not work
on the things that you expect to work
but you also have to put resources and
do understand the things that might
possibly work and could be very valuable
if they do and large companies do this
they for run startup incubators or they
give people 20% time projects things
like this um if you are uh a Academia I
think you also need to find a balance
where you give people who are smart and
have the intention to build something
where they're willing to risk their
personal career and that thing is
valuable give them the space for doing
that to some degree right this must
exist and arguably I think currently in
Academia it doesn't exist enough to a
very large part is a scientist is
somebody who applies established methods
and plays it safe it's very hard to get
tenure so you need to have a track
record of successful Publications to get
grants you need to have uh successful
Grant proposals in the past and so on so
there's a similarity between this but on
the other hand I also think it's okay
that most people don't take risks
society and families and relationships
mostly work because we don't take risks
all the time but we do the thing that we
have a very good reason to expect to
work and mostly I think that's good not
everything that we do needs to make
progress this is just at the
boundaries so the this boundary thing is
what excites us but we find interesting
it's what you can tell stories about but
if you are interested in how to get life
to work it's not so much important that
you have an amazing story to tell about
your life I've tried Adventure all my
life and uh it's not pleasant I see it
more like an
Affliction and uh there are people who
are like this right so if you look at
taleb and some sense taleb is like John
McAfee he's an
adventurer and uh these days are not
suitable for adventuring there are many
uh jobs for people who do engineering
and uh do Administration and so on very
few jobs for Francis Drake who becomes a
pirate and tries to explore the world
and Conquer new lands or discover new
lands or whatever right this doesn't
exist and so you either do this like
John McAfee and you get trouble with the
IRS and eventually uh
very bad things happen to you and you
have to be on the run run for most of
your life and uh keep
yourself alive being a scoundrel and
doing crypto or you are like talb and uh
when you are like talb you uh make a
very few a few very very good deals and
you are safe financially and uh you
spend your adventures by being angry on
people on the internet and eating eating
fish ink uh Squid Ink spaghetti and uh
uh writing very bold books right and
this these are some of the Avenues that
you can take when you're an adventurer
and you can also try to to do a startup
but when you do a startup you have
responsibility for the people that work
for you in the startup right so there is
a limit to the risk that you can take
you have responsibility to your
investors and so on and so there is a
limit to the adventuring that you can do
in the present time because in the past
adventuring meant that you could
actually fail that you could actually be
killed that you could actually be
murdered by strangers and uh all your uh
Adventures would not come to pass so
when you settle in your lens this is the
risk that you're taking and you
typically take this risk because you're
actually already in danger because you
cannot live at home or because you're
somebody who really really doesn't fit
in and if you really really don't fit in
I think you have no choice but to become
the adventurer and to try to figure out
how this everything gets to work but um
for most people that's not an advisable
course of action now if somebody were
willing to do that what is an idea
within AI you'd like to see them explore
that's likely to fail but if it worked
it would really be something so I I
think that all the important ideas in AI
have not worked so far the uh neural
network trained with something like the
transformer is the first idea that
basically works at scale that gives you
something that is very close to
Universal function approxim met it might
not be optimal but it is very very
powerful and we don't know the limits of
this idea so I think it's very
reasonable if you do a startup uh to to
use the stuff that is currently working
and where we don't see the boundaries of
of what can be done with it right
building an AI system that is made of
multiple llms is probably um one of the
most promising way to overcome
limitations of llms or to teach a neural
network to use a computer ultrapress
system or to use a game engine or to use
other tools or to write its own tools
and integrate with them or to extend
tools all this stuff is what you can do
use the existing paradigms for and have
a very high likelihood of succeeding
building at least something interesting
but um what I find very interesting is
self-organizing AI for instance imagine
that you want to build an AI that is
actually empathetic that can do
perceptual empathy that can change your
mental state by going into resonance
with you that I think is would be super
exciting but it might require that it's
sampling You by looking at camera images
and all sorts of sensors at the same
frequency as your nervous system or
multiple of it and it should synchronize
its representations or the update
frequency of its representations to your
own and then our nervous system that's
flexible right our nervous system is not
always working at the same frequency but
it's changing there's an arrangement of
different modules that work at different
frequencies and interplay with each
other in our mind and body so if you are
able to attune yourself to this and
create feedback to people in this way
that would be super interesting so
getting something to work that works in
real time by being coupled with the
environment and then actually being
coupled with people would be super
interesting thing to explore right and
it may or may not lead to better
believable avatars and it might lead to
automatic psychologist and it might lead
to tools that read your thoughts and
allow you to basically co- think with
the computer so you imagine things and
very little input of you with your mouse
they form on the screen and much deeper
and richer than you could achieve them
by using your mouse and keyboard yeah
that to me is really exciting so um I
assume you know nothing about my
background but we're building a uh what
I hope will eventually be the first
steps into a true Ready Player one and
we want AI to make up over time
obviously it's not there yet but over
time to make up the MPCs so that you can
actually have individual relationships
with those characters I think a lot
about that and uh making those
relationships rewarding and enriching
would be incredible and so we've got a
thing again for people that know what
we're building this will take time um
but we have an idea for an interaction
point with AI that would grow more um
intimate and useful over time that I
think would really be exciting you see
sort of the beginning edge of this with
replica I don't know how if you've paid
close attention to that uh it got a
little controversial as they sort of
went sexbot uh with it but um it's a
very interesting idea in terms of
something that's establishing an ongoing
relationship with you has memory changes
over time the relationship actually
deepens and evolves I think that's going
to get pretty interesting in a world
where we have ai that can pass the
touring test and so for me the thought
of spending some portion of my time in a
known simulation where the characters
are as real to me as people that I know
in real life uh that would be pretty
extraordinary for a lot of people that's
exciting because it's not like the
people that I meet in real life I think
the people that are most drawn to such
applications are people who don't have
working relationships in real life so I
think that such systems will become
Prosthetics that uh replace your need
for relationships by something that does
not actually require you to interact
with other people if you're unable to
find a boyfriend or girlfriend and build
a family withd them and you give up on
this then having an AI uh girlfriend
that can also be a sex spot uh can be
looking like the better
alternative right and it's very painful
to consider this idea I've been
contacted by people who said can't you
build something like this like the
perfect catg girl that could be my
girlfriend because I've given up on this
it's it's not going to happen for me in
this life
and I think that it might be more
interesting to build a coach something
that is both your assistant and your
adviser and bouncing board that in a
safe way allows you to explore
possibilities of human interaction and
allows you
to extend your social abilities by
training and by augmenting them in such
a way that you can find the
relationships that you want in this
world and have actual agency again and
instead of a system that is basically
helping you to deal with the fact that
you have
none and I think such systems can start
out as girlfriends as long as they're
honest as long as they're saying this is
exactly how I am this is how I work this
is my purpose are you okay with this and
let's go on this journey together
basically Building Systems that can be
assistance of people to get around
better with the world and actually serve
their own purposes rather than helping
them to deal with the fact that they
cannot find a
purpose that uh to me seems to be the
more course to go man I'm with you on
that one this has been amazing yosha
where can people follow you at the
moment mostly on Twitter that's the
easiest way uh also uh if you are
interested in seeing other podcasts uh
for instance um like freedman's podcast
and many others can look on YouTube on
my channel basically have
collection of all the appearances on
different podcasts that I had at some
point I know that I have to probably
write a book but uh it's still somewhat
conflicting with my everyday duties with
my family but um to be getting there I
love it awesome dude thank you again for
coming on everybody at home if you have
not already be sure to subscribe and
until next time my friends be legendary
take care peace if you enjoyed this
episode check out this deep convers
ation with Donald Hoffman about reality
and Consciousness what we are are
avatars of the one the one awareness is
exploring all of its possibilities
through different avatars so somehow
there is this field of
awareness that
is