Kind: captions
Language: en
the following is a conversation with
rodney brooks one of the greatest
roboticists in history he led the
computer science and artificial
intelligence laboratory at mit then
co-founded irobot which is one of the
most successful robotics companies ever
then he co-founded rethink robotics that
created some amazing collaborative
robots like baxter and sawyer
finally he co-founded robust.ai
whose mission is to teach robots common
sense which is a lot harder than it
sounds
to support this podcast please check out
our sponsors in the description
as a side note let me say that rodney is
someone i've looked up to for many years
in my now over two decade journey in
robotics because
one he's a legit great engineer of real
world systems and two he's not afraid to
state controversial opinions that
challenge the way we see the ai world
but of course while i agree with him on
some of his critical views of ai i don't
agree with some others
and he's fully supportive of such
disagreement
nobody ever built anything great by
being fully agreeable
there's always respect and love behind
our interactions and when a conversation
is recorded like it was for this podcast
i think a little bit of disagreement
is fun
this is the lex friedman podcast and
here is my conversation with rodney
brooks
what is the most amazing or beautiful
robot that you've ever had the chance to
work with
i think it was domo
which was made by one of my grad
students aaron ed singer
it now sits in daniela roose's office uh
director of sea sale
and it was just a beautiful robot and
aaron was really clever
he didn't give me a budget ahead of time
he didn't tell me what he was going to
do
he just started spending money he spent
a lot of money he and jeff webber who um
as a mechanical engineer who
aaron insisted he bring with him when he
became a grad student built this
beautiful gorgeous robot domo which is a
upper torso humanoid
two two arms uh
with things three fingered hands um
and
face eyeballs um all uh not the not the
eyeballs but everything else series
elastic actuators uh you can interact
with it um
cable driven all the motors are inside
and it's just
gorgeous the eyeballs are actuated too
or no oh yeah the eyeballs are actuated
with cameras and you know so it had a
visual attention mechanism you know wow
looking when people came in and looking
in their face and
talking with them
why was it amazing
the beauty of it you said you said what
was the most beauty what is the most
beautiful it's just mechanically
gorgeous as as everything aaron builds
has always been mechanically gorgeous
it's just exquisite in the detail
we're talking about mechanically like
literally the amount of actuators
the actuators the cables he
anodizes different parts different
colors and it's just
looks like a work of art what about the
face is that do you find the face
beautiful in robots um when you make a
robot
it's making a promise for how well it
will be able to interact so i always
encourage my students
not to over promise you know even with
its essence like the thing it presents
it should not over promise yeah so i the
the joke i make which i think you'll get
is if your robot looks like albert
einstein it should be as smart as albert
einstein so
the only thing in in
domo's face is
the eyeballs um and because that's all
it can do it can look at you and pay
attention um and
so there is no
it's not like one of those
um japanese robots that looks exactly
like a person at all
but see the thing is us humans and dogs
too
don't just use eyes for it as
attentional mechanisms they also use it
to communicate it's part of the
communication like a dog can look at you
look at another thing and look back at
you and that designates that we're going
to be looking at that thing yeah or
intent you know in on on both baxter and
sawyer at rethink robotics they had a
screen with you know graphic eyes so it
wasn't actually where the cameras were
pointing but it
the the eyes would look in the direction
it was about to move its arm so people
in the factory nearby were not surprised
by its motions because it gave that
intent away
before we talk about baxter which i
think is a beautiful robot
let's go back to the beginning when did
you first
fall in love with robotics we're talking
about beauty and love to open the
conversation this is great
i've got these i was born in the end of
1954 and i grew up in adelaide south
australia and i have these two books
that are dated 1961
so i'm guessing my mother found them in
a store in 62 or 63.
how and why wonder books um
how am i on the book of electricity
and how i won the book of
giant brains and robots
and
i learned how to build circuits you know
when i was eight or nine simple circuits
and i and i read you know learned the
binary system and
um saw all these
drawings mostly uh of robots and um then
i tried to build them for the rest of my
childhood wait
61 you said this was when the two books
i've still got the at home what what
does the robot mean in that context no
they they were some of the robots that
they had were
arms you know big arms to move nuclear
material around but they had pictures of
welding robots that looked like humans
under the sea welding stuff underwater
um so they weren't real robots uh but
they were you know what people were
thinking about for robots
what were you thinking about were you
thinking about humanoids were you
thinking about arms with fingers were
you thinking about faces or cars no
actually to be honest i realized my
limitation on building mechanical stuff
so i just built um the brains
mostly i out of
different technologies as i got older
i built a
learning system which
was chemical based and i had this ice
cube tray each
well was a cell
and by applying voltage to the two
electrodes it would build up a copper
bridge so over time it would it would
learn
a simple network um so i could teach it
stuff and that was
mostly things were driven by my budget
and nails as electrodes and a an ice
cream i mean an ice cube tray was was
about my budget at that stage later i
managed to buy transistors and then i
could build gates and flip-flops and
stuff so so one one of your first robots
was an ice cube tray yeah
and it was very cerebral because it
learned to add
very nice uh well
just a decade or so before in 1950 alan
turing
wrote the paper that formulated the
touring test and he opened that paper
with the question
can machines
think
so let me ask you this question can
machines
think can your ice cube tray
one day think
um certainly machines can think because
i believe you're a machine and i'm a
machine and i believe we both think um i
think it's a big fear i think any other
philosophical position is sort of a
little ludicrous what does think mean if
if it's not something that we do
um and we and we are machines
so yes machines can but do we have a
clue how to build such machines that's a
very different question are we capable
of building such machines uh you know
are we smart enough we think we're smart
enough to do anything but
maybe we're not maybe you know we're
just not bad enough to build
stuff like us the kind of computer that
uh alan turing was thinking about do you
think there is something
fundamentally
uh or significantly different between
the computer between our ears the the
biological computer that humans use
and uh the computer that he was thinking
about from a from a sort of high level
philosophical yeah i believe it uh
that's very wrong in fact i'm
halfway through a
i think it'll be about a 480 page book
um
titled the working title is not even
wrong
and if i may i'll tell you a bit about
that book
so there's two two well three thrusts to
it
um
one is the history of computation what
we call computation it goes all the way
back to uh uh
some manuscripts in latin from 1614 and
1620 by napier and kepler through
babbage and lovelace and then
turing's 1936 paper uh is you know where
what we think of as the invention of of
of modern computation
and that paper by the way did not set
out to
you know invent computation it set out
to uh negatively answer one of uh
hilbert's three later set of problems he
called it um
as an effective way of of of getting
answers and and hilbert
hilbert really worked with rewriting
rules as did um
um
a church who also at the same time a
month earlier insuring disproved
hilbert's one of these three hypotheses
the other two had been already been
disproved by godel so turing set out to
disprove it because that's it's always
easier to disprove these things than to
prove that there is an answer and um
so he needed um
and it really came from his his uh
professor well as an undergrad at
cambridge who said who turned it into is
there a mechanical process so he wanted
to have a show a mechanical process that
could
calculate numbers
because that was a mechanical process
that people used to generate tables they
were called computers the people at the
time and they followed a set of rules
where they had paper and they would
write numbers down and based on the
numbers that keep writing other numbers
and they would produce
numbers for these tables engineering
tables
that the more the more iterations they
did the more significant digits came out
and so turing
in that paper set out to
define what sort of machine could do
that mechanical machine
where it could produce an arbitrary
number of digits
in the same way a human computer did
and he came up with a very simple
set of constraints where there was an
infinite supply of paper this is the
tape of the turing machine
and each turing machine had a
set of it came with a set of
instructions that
as a person could do with pencil and
paper write down things on the tape and
erase them and put new things there
and
he was able to show that that system was
not able to do something that hilbert
hypothesized so he disproved it but he
had to show this was this this system
was
good enough to do
whatever could be done but couldn't do
this other thing yeah and there he said
and he says in the paper i don't have
any real arguments for this but based on
intuition
so that's how he defined computation
and then if you look over the next from
1936 up until really around 1975
you see people struggling with
is this really what computation is and
so marvin minsky very well known in ai
but also
a fantastic mathematician
uh in his book finite and infinite
machines from the mid 60s which is a
beautiful beautiful mathematical book
um says says at the start of the book
well what is computation turing says
it's this and yeah i sort of think it's
that it doesn't really matter whether
stuff's made of wood or plastic it's
just you know that relatively cheap
stuff can do this stuff and so yeah
seems like computation
and donald knuth uh in his first volume
of his
you know art of computer programming
in around 1968
says well what's computation
it's this stuff like turing says that a
person could do each step without too
much trouble
and so one of his examples of what would
be too much trouble was
a step which required knowing whether
fermat's last theorem was true or not
because it was not known at the time and
that's too much trouble for a person to
do as a step
and um
hop craft and allman
sort of said a similar thing
later that year and by nineteen seventy
five
in the a ho up crop the norman book
they're saying
well you know we don't really know what
computation is but intuition says this
is sort of about right and this is what
it is
that's computation it's a sort of
agreed-upon thing which happens to be
really easy to implement in silicon and
then we had moore's law which took off
and it's been an incredibly powerful
tool
i certainly wouldn't argue with that the
version we have a computation incredibly
powerful can we just take a pause so
what we're talking about is there's an
infinite tape with some simple rules on
how to write and on that tape and that's
that's what we're kind of thinking about
this is computation yeah and it's
modeled after humans how humans do stuff
and
i think it's a
curing says in the 36 paper one of the
critical facts here is that a human has
a limited amount of memory
so that's what we're going to put onto
our
mechanical computers so
so you know unlike mass
unlike mass or charge or yeah it's not
it's it's not given by the universe it
was
this is what we're going to call
computation yeah and then it has this
really
you know it had this really good
implementation which has completely
changed our technological world that's
computation
second part of the book
i uh or argument in the book i have this
two by two matrix
with
um science
in the top row engineering in the bottom
row
left column is intelligence right column
is
life so in the bottom row the
engineering there's artificial
intelligence and there's artificial life
in the top row there's neuroscience
and abiogenesis how does living matter
turn in how does non-living matter
become living better yes four
disciplines these four disciplines all
um
came into the current form in the period
1945 to 1965.
um
that's interesting there was
neuroscience before but it wasn't
effective neuroscience it was you know
these ganglia and there's electrical
charges but no one knew knows what to do
with it
and furthermore
there were a lot of players who are
common across them
i've
identified common players except for
artificial intelligence and habiogenesis
i don't have but for any other pair i
can point to people who work them and a
whole bunch of them by the way we're at
the research lab for electronics at mit
um where uh warren mcculloch uh held
held held forth
and in fact mcculloch pitts um letven
and maturana wrote the first paper on
functional neuroscience called what the
frog's eye tells the frog's brain where
instead of it just being this bunch of
nerves they sort of showed what
different
anatomical components were doing and
telling other anatomical components
and
you know generating behavior in the
front would you put them as basically
the fathers or the one of the early
pioneers of what are now called
artificial neural networks
yeah i mean mcculloch and pitts
pitts was
much younger than him in 1943 had
written a paper inspired by bertrand
russell
on
a calculus for the ideas imminent in
neural systems
where they had tried to without any real
proof
they had tried to give a formalism for
neurons
basically in terms of logic and gates or
gates and not gates with with no
real
evidence that that was what was going on
but they they talked about it and that
that was picked up by minsky for his
uh dissertation on which was a was a
neural network we would call it today
it was picked up by um
john von neumann when he was designing
the edvac computer in 1945 he talked
about its components being neurons
based on and in references he's only got
three references and one of them is the
mcculloch pitts paper
so all these people and then the ai
people and the artificial live people
which was john von neumann originally it
was like overlap because you know
they're all going around the same time
and three of these four disciplines turn
to computation as their primary metaphor
so i i've got a couple of chapters in
the book one is titled wait
computers are people because that's
where our computers came from yeah
and you know from people who are
computing stuff and then i've got
another chapter
wait people are computers which is about
computational neuroscience yes so
there's this whole circle here and that
competition is it and you know i have
talked to to people about well maybe
it's not computation that goes on in the
head of course it is yeah okay well
when elon musk's rocket goes up is it
computing
is that how it gets into orbit by
computing
but we've got this idea if you want to
build an ai system you write a computer
program
yeah in a sense so the word computation
very quickly starts doing a lot of work
that it was not initially intended to to
do it's the second saying if you talk
about the universe as essentially
performing a computation yeah right
wolfram does this he turns it into
computation
you don't turn rockets into computation
yeah by the way when you say computation
in our conversation do you tend to think
of computation narrowly in the way
touring thought of computation
it's it's gotten very
okay
you know
squishy yeah squishy okay
um
but
computation in the way turing thinks
about it and the way most people think
about it
actually fits very well with
thinking like a hunter-gatherer
there are places and there can be stuff
in places and the stuff in places can
change and it stays there until someone
changes it
and it's this
metaphor of place and container which
you know is a combination of our
place cells in our hippocampus and uh
cortex
but this is this is how we use metaphors
for mostly to think about and when we
get outside of our metaphor range we
have to invent tools which we can sort
of
switch on to you so calculus is an
example of a tool it can do stuff that
our raw reasoning can't do and we've got
conventions of when you can use it or
not
but sometimes um
you know people try to all the time we
always try to get physical metaphors for
things which is why
quantum mechanics has been such a
problem for a hundred years because it's
a particle no it's a wave it's got to be
something we understand and i say no
it's some weird mathematical logic
that's different from those but we want
that metaphor
well you know i i suspect that that you
know 100 years or 200 years from now
neither quantum mechanics nor nor dark
matter will be talked about in the same
terms you know in the same way that um
lodges theory eventually went away
because
it just wasn't an adequate explanatory
metaphor you know that metaphor was the
stuff
there is stuff in
the burning the burning is in the matter
because it turns out the burning was
outside the matter it was the oxygen
so our desire for metaphor and combined
with our limited cognitive capabilities
gets us into trouble that's my argument
in this book now and people say well
what is it then and i say well i wish i
knew that i tried to talk about that but
i you know give some ideas but so so
this is the three things computation is
sort of a particular thing we use
um
uh
oh can i tell you one beautiful thing
one yes
so you know i used an example of a thing
that's different from computation you
hit a drum and it vibrates and there are
some some stationary points on the drum
surface you know because the waves are
going up and down the stationary points
now
you could compute them
to arbitrary position
um but the drum just knows them the drum
doesn't have to compute
what was the very first computer program
ever written by ada lovelace
to compute bernoulli numbers and
bernoulli numbers are exactly what you
need to find those stable points in the
drum's surface wow anyway and there was
a bug in her program
the arguments to divide were reversed in
one place
and it still worked well she never got
to run it they never built the
analytical engine she wrote the program
without without it you know
uh so so computation computation is sort
of you know a thing that's
become dominant as a metaphor but yeah
is it the right metaphor
um
all three of these four fields adopted
computation and
you know the
a lot of it swirls around warren
mcculloch and his all his students and
he funded a lot of people
um
and uh
and and our human metaphors our
limitations to human thinking will play
into this
the
three themes of the book
so i have a little to say about
computation
so
uh
so you're saying that there is a gap
between the computer or the the the
machine that performs computation
and this machine
that appears to have consciousness
and intelligence yeah can we um that
piece of meat in your head piece of meat
and maybe it's not just the meat in your
head it's the rest of you too i mean you
have you have you actually have a neural
system in your gut um i tend to also
believe
not believe but
we're now dancing around things we don't
know but
i tend to believe other humans are
important
like so we're almost like
i i just don't think we would ever have
achieved the level of intelligence we
have with other humans
i'm not saying so confidently but i have
an intuition that some of the
intelligence is in the interaction yeah
and and i think you know i think it it
seems to be very likely again we you
know this is speculation but
we our species and probably um
probably in the end those to some extent
because you can find uh old bones where
they seem to be counting on them by
putting notches
um that when the in the neanderthals are
done we are able to put um
some of our stuff outside our body into
the world and then other people can
share it
and then we get these tools that become
shared tools and so there's a whole
coupling that
would not occur in you know
the single
deep learning network which was fed you
know all of literature or something
yeah the the the neural network can't um
step outside of itself but is there
is there some um
can we explore this dark room
a little bit and try to get at something
what what is the magic where does the
magic come from in the human brain that
creates the mind
what's your sense
as scientists
that try to understand it and try to
build it what are the directions that
if followed might be productive is it
creative interactive robots
is it creating large deep neural
networks that
do like self-supervised learning and
just like we'll we'll
we'll discover that when you make
something large enough some interesting
things will emerge is it through physics
and chemistry biology like artificial
life angle like we'll sneak up in this
four quadrant matrix that you mentioned
is there anything
you're you're most if you had to bet all
your money
financial advice i wouldn't
okay so
every intelligence we know and includes
you know animal intelligence dog
intelligence you know
octopus intelligence which is very
different sort of architecture from from
us
all the intelligences we know
perceive the world
in some way
and then have action
in the world
but they're able to
perceive objects
in a way which is actually pretty damn
for not phenomenal
and surprising
you know we tend to think you know that
that that uh
the box over here between us which is a
sound box i think is a blue box but
blueness
is something that we construct with with
um
color constancy it's not a it's not a
it's not the blueness is not a direct
function of the
photons we're receiving it's actually
context you know which is why
um
you can turn
you know
maybe seen the examples where
um
someone turns a stop sign into a
some other sort of sign by just putting
a couple of marks on them and the deep
learning system gets it wrong everyone
says but the stop sign is red you know
why is it why is it think it's the other
sort of science because redness is not
intrinsic in just the photons it's
actually a construction of an
understanding of the whole world and the
relationship between objects to get
color constancy
um but our tendency in order that we get
an archive paper really quickly is you
just show a lot of data and give the
labels and hope it figures it out but
it's not figuring it out in the same way
we do we have a very complex perceptual
understanding of the world dogs have a
very different perceptual understanding
based on smell they go smell smell a
post they can tell
how many
you know different dogs have visited in
the last 10 hours and how long ago
there's all sorts of stuff that we just
don't perceive about the world and just
taking a single snapshot is not
perceiving about the world it's not
seeing that the registration between us
and the object
and registration is a
a philosophical concept brian cantwell
smith talks about a lot very difficult
squirmy thing to understand
but i think none of our systems do that
we've always talked about nai about the
symbol grounding problem how our symbols
that we talk about are grounded in the
world
and when deep learning came along and
started labeling images people said ah
the grounding problem has been solved no
the labeling problem was solved with
some percentage accuracy which is
different from the grounding problem
so you uh
there's uh you agree with hans marvik
and
what's called the marvex paradox
that highlights this counterintuitive
notion that
reasoning
is easy
but perception
and mobility are hard
yeah we shared an office when um when i
was working on computer vision and he
was working on his first mobile robot
what was those conversations like they
were great
so do you still kind of maybe you can
elaborate and do you still believe this
kind of notion that
perception is really uh hard and like
can you make sense of why we humans have
this poor intuition about what's hard
and not well well let me let me give us
sort of a
an another another story sure
if you go back to you know the original
um you know teams working on ai um
from the late 50s into the 60s you know
and you go to the ai lab at mit
um who was it that was doing that was a
bunch of really smart kids who got into
mit
and they were intelligent
so what's intelligence about well the
stuff they were good at playing chess
doing integrals
that was that was hard stuff
yeah but you know a baby could see stuff
that wasn't that wasn't intelligent
anyone could do that that's not
intelligence
and so it you know this there was this
intuition that the hard stuff is the
things they were good at
and the easy stuff was the stuff that
everyone could do
yeah and maybe i'm overplaying it a
little bit but i think there's an
element of that yeah i mean there
i don't know how much truth there is to
uh like chess for
example has was for the longest time
seen as the highest um
level of intellect
right
until we got computer so we're better at
it than people and then
we realized you know if you go back to
the 90s you'll see you know the stories
in the press around when when
kasparov was beaten by deep blue oh this
is the end of all sorts of things
computers are going to be able to do
anything from now on and we saw exactly
the same stories with alpha zero the
go playing program yeah
but still to me
reasoning is a special thing
and perhaps no way we actually we're
really bad at reasoning we just use
these analogies based on our how to
gather intuitions but why is that not
don't you think the ability to construct
metaphor is a really powerful thing oh
yeah it is stories it is that's it's the
constructing the metaphor and
registering that yeah something
complicated
is not what we're doing with
with vision too and
we're telling our stories we're
constructing good models of the world
yeah yeah but but um
i think we we jumped between what we're
capable of and how we're doing it right
there there was a little confusion that
went on um
uh as we were telling each other stories
yes exactly
trying to dilute each other no i i just
think uh i'm not exactly so i'm trying
to pull apart this marvex paradox
i don't view it as a paradox
what did evolution what did evolution
spend its time on yes it spent its time
on getting us to perceive and move in
the world that was you know 600 million
years as multi-cell creatures doing that
and then it was you know relatively
recent that we that we you know were
able to
um hunt or or gather or you know even
even animals hunting that's much more
recent and then and then anything that
we you know speech uh language those
things are you know a couple of hundred
thousand years probably if if if that
long and then agriculture 10 000 years
you know
all that stuff was built on top of those
earlier things which took a long time to
develop so if you then look at the
engineering of these things so
building it into robots
what's the hardest part of robotics do
you think
as uh through the decades that you
worked on robots
in the context of what we're talking
about vision you know perception
the actual sort of the
the biomechanics of movement
i'm kind of drawing parallels here
between humans and machines always like
uh what do you think is the hardest part
of robotics
i sort of think all of them
there are no easy parts to do well um
we we sort of go reductionist and we
reduce it to if only we had all the
location of all the points in 3d yeah
things would be great
you know if only we had labels on the on
the images you know
things would be great but you know as as
we see that's not good enough
some deeper
understanding but if you if i came to
you and i could solve
one category of problems in robotics
uh instantly
what would give you uh the greatest
pleasure
[Laughter]
i mean is it uh
you know you you look at robots that
manipulate objects
uh
what's hard about that
you know is it uh the perception
is it the uh the reasoning about the
world like common sense reasoning
is it the actual
building a robot that's able to interact
with the world
is it like human aspects of a robot
that's interacting with humans and that
that game theory of how they work well
together well let's talk about
manipulation for a second because i had
this really blinding
moment
uh uh you know i'm a grandfather so
grandfather's had blinding moments yes
just a
three or four miles from here
last year my
16 month old grandson was in his new
house first time right
first time in this house and he'd never
been able to get to a window before but
this had some low windows and he goes up
to this window with a handle on it that
he's never seen before and he's got one
hand pushing the window and the other
hand turning the handle to open the
window
he he knew he two different hands
two different things he knew how to how
to put together yeah
and he's 16 months old and there you are
watching an awesome
in an environment environment he'd never
seen before mechanism how did he do that
yes that's a good question how did he do
that that's why
it's like okay like you could see the
the leap of genius from using one hand
to perform a task to combining
doing i mean first of all in
manipulation that's really difficult
it's like two hands
both necessary to complete the action
and completely different and he'd never
seen a window open before
but in third somehow handle opened
something yeah there may have been a lot
of
slightly different failure cases that
you didn't see
yeah not with a window but with other
objects of turning and twisting in
handles
you know there's a there's a great
counter to
um you know
reinforced reinforcement learning will
just give you know the robot um
or you give the robot plenty of time to
try everything yes actually
can i tell a little side story here
so i'm in um
deep mind in london
is
three four years ago where
um you know there's a big google
building
and then you go inside and you go
through there's more security and then
you get to deep mind where the other
google employees can't go yeah and i'm
in a
i'm in a conference room bayer
conference room with some of the people
and
they tell me about their reinforcement
learning experiment with uh robots um
um which um
are just trying stuff out and they're my
robots they're they're sawyers that we
sold them um
uh and they really like them because
sawyers are compliant and can sense
forces so they don't break when it's
bashing into walls they they stop and
they do stuff
and you know so you just let the robot
do stuff and eventually it figures stuff
out by the way so
we're talking about robot manipulation
so robot arms and so on yeah so he's a
robot yeah um just to go what's sawyer
so here's a robot arm that my company
rethink robotics yeah built thank you
for the context
sorry okay cool so we're indeed mine
and the you know it's in the next room
these robots are just bashing around to
try and use reinforcement learning to
learn how to act and can i go see them
oh no they're secret they're all my
robots they're a secret
that's hilarious okay anyway the point
is you know this idea that you just let
uh reinforcement learning figure
everything out is so counter to how a
kid does stuff
so
again story about my grandson i gave him
this this uh
box that had lots of different lock
mechanisms he didn't randomly you know
when he was 18 months old he didn't
randomly try to touch every surface or
push everything he found he could see
what where the mechanism was and he
started exploring the mechanism for each
of these different lock mechanisms
and there was reinforcement no doubt of
some sort going on there but he applied
a pre-filter which cut down the search
space dramatically
i i
i wonder to what level we're able to
introspect what's going on
because what's also possible is you have
something like reinforcement learning
going on in the mind in the space of
imagination so like you have a good
model of the world you're predicting and
you may be running those tens of
thousands of like
loops but you're like as a human you're
just looking at yourself trying to tell
a story of what happened and it might
seem simple but maybe there's a lot of
computation going on
whatever it is but there's also a
mechanism that's being built up it's not
just
random search
mechanism prunes it
dramatically yeah that that pruning
uh that pruning stuff but it doesn't
it's possible that that's
so you don't think that's akin to a
neural network inside of reinforcement
learning algorithm
is it possible
it's
yeah until it's possible i uh but but i
you know
um
i
i i'll be incredibly surprised if that
happens i'll also be incredibly
surprised that you know after all the
decades that i've been doing this where
every every few years someone thinks now
we've got it
now we've got it
you know
four or five years ago i was saying i
don't think we've got it yet and
everyone was saying no you don't
understand how powerful hey i had people
tell me you don't understand how
powerful it is
um i you know i i i sort of had a a
a track record of what the world had
done to think well this is no different
from before oh we have bigger computers
we had bigger computers in the 90s and
we could do more ship stuff
but okay so let me let me let me push
back because i i'm i'm generally sort of
optimistic and tried to find the beauty
in things
i think there's a lot of uh
surprising and beautiful things that
neural networks this new generation of
deep learning revolution has revealed to
me it has continually been very
surprising the kind of things it's able
to do now generalizing that over saying
like this we've solved intelligence
that's another
uh big leap but is there something
surprising and beautiful to you about
neural networks that where actually you
set back and said
i i did not expect this
oh i think i think their performance
their performance on imagenet was
shocking the computer vision those early
days was just very like wow okay
that doesn't mean that they're solving
everything in computer vision
we need to solve or in
vision for robots what about alpha zero
and self-play mechanisms and
reinforcement learning isn't that yeah
that was all in in donald mickey's 1961
paper um
everything that was there which
introduced reinforcement learning um no
but come on so no you're talking about
the actual techniques but isn't it
surprising to you the level it's able to
achieve with no human supervision
of chess play like
to me there's a big big difference maybe
blue and maybe what that's saying
is how overblown our view of ourselves
is
you know we that chest is easy
yeah i mean i i
i came across this
1946 report
that um and i'd seen this as a kid in
one of those books that my mother had
given me actually
um 1946 report which pitted uh
someone with an abacus against an
electronic calculator
and he beat the electronic calculator
you know so there at that point was well
humans are still better than
machines are calculating are you
surprised today that a machine can you
know do a billion floating point
operations a second and you know you're
you're puzzling for for minutes through
one
so you know
i i am i mean i i don't know but i am
certainly surprised
there's something
uh to me different about learning
so system that's able to learn learning
now see now you get into one of the
deadly sins
because
of using terms uh overly broadly yeah i
mean there's so many different forms of
learning yeah and so many different
forms you know i learned my way around
the city i learned to play chess i
learnt latin um i learned to ride a
bicycle all of those are
you know are very different capabilities
yeah and
if someone
you know has a
in you know in the old days people would
write a paper about learning something
now the corporate
press office puts out a press release
about how
company x has
has
is leading the world because they have a
system that can
yeah but here's the thing okay so what
is learning
what i'm referring to there's many
things but
a suitcase would it's a suitcase word
but
loosely
there's a dumb system
and over time
it becomes smart
well it becomes less dumb at the thing
that it's doing yeah smart is a
different mass is a loaded word yes less
less dominant thing is again it gets
better performance under some measure
yeah and some set of conditions at that
thing and and most of these learning
algorithms
um
uh learning systems
fail when you change the conditions just
a little bit in a way that humans don't
so right i was at deepmind
um
the alphago had just come out
and i said what would have happened if
you'd given a 21 by 21 board instead of
a 19 by 19 board they said fail totally
but a human player would actually you
know well
would actually have to play again and
actually funny enough if you look at
deepmind's work uh since then
uh they are pers
uh
they're presenting a lot of algorithms
that would do uh
that would do well at the at the bigger
board so they're slowly expanding this
generalization i mean to me there's a
core element there i think it is very
surprising to me that even in a
constrained game of chess or go
that through self-play by a system
playing itself
that can it can achieve
super human level performance
through learning alone so like okay so
so you know you you don't find them as
you did it in search of that you didn't
you didn't like it when i referred to
donald mickey's 1961 paper
there in the second
part of it it came a year later they had
self-play on an electronic computer
at
it's tic-tac-toe okay
but it learned to play tic-tac-toe
through self-play that's not and i've
learned to play optimally what i'm
saying is
uh i
okay i have a little bit of a bias but
i i find ideas beautiful but only when
they actually
realize the promise that's another level
of beauty like for example
uh what the
bezos and elon musk are doing with
rockets we've had rockets for a long
time but doing reusable cheap rockets
it's very impressive
in the same way
i okay yeah i would have not predicted
first of all when i was uh
started and fell in love with ai
the game of go was seemed to be
impossible to solve okay so i thought
maybe you know i
maybe it'd be possible to maybe have big
leaps in a moore's law style of way in
computation i'll be able to solve it but
i would never have guessed that you
could learn your way
however
i mean
in the narrow sense of learning learn
your way to
to to beat the best people in the world
at the game of go without human
supervision not studying the game of
experts
but okay so so that's just
using a different learning technique yes
arthur samuel
in the early 60s and he was the first
person to use machine learning
got had a program that could beat the
world champion at checkers
now yes so and that at the time was
considered amazing by the way arthur
samuel had some fantastic advantages
yeah do you want to hear arthur samuel's
advantage
two things one he was at the 1956 um ai
conference i knew arthur later in life
he was at stanford when i was scratching
there he wore a tie and a jacket every
day the rest of us didn't
um he's a
delightful man delightful man um
it turns out claude shannon in a 1950
scientific american article
uh outlined
on chess playing outline the learning
mechanism that arthur samuel used
and they had met in 1956. i assume there
was some communication but i don't know
that for sure
but arthur samuel
had been a vacuum tube engineer on
getting reliability of vacuum tubes and
then had overseen
the first transistorized computers at
ibm and in those days
before you shipped a computer you ran it
for a week to see to get early failures
so he had this whole farm of computers
running random code
for hours and hours
a week for each computer we had a whole
bunch of them so he ran his chess
learning program
with self play
on on ibm's production line he had more
computation available to him than anyone
else in the world and then he was able
to produce a chess playing program i
mean a checkers playing program that
could beat the world champion
so that's amazing the question is
what i mean surprised i don't just mean
it's nice to have that accomplishment
is there is
a stepping
towards something that feels
uh
more intelligent
than before yeah and the question is
that's in your view of the world okay
well let me then
it doesn't mean i'm wrong
no no it doesn't
so the question is if we keep taking
steps like that how far that takes us
are we going to build a better
recommender systems are we going to
build better robots or will we solve
intelligence so
you know i'm putting my bet on
um
but still missing a whole lot a lot um
and and why would i say that well in
these games they're all um you know 100
information games
but again but
each of these systems is a very uh short
description of the current state
um
which is different from registering and
perception in the world okay which gets
back to my rfx paradox i'm definitely
not saying that uh
chess is somehow harder than uh
perception
or
any kind of even any kind of robotics in
the physical world i i definitely think
is is way harder than the game of chess
so i was always much more impressed by
like the workings of the human mind
that's incredible the human mind is
incredible i've i believe that from the
very beginning i want to be a
psychiatrist for the longest time i
always thought that's way more
incredible in the game of chess i think
the game of chess is uh
i love the olympics it's just another
example of us humans picking a task and
then agreeing that a million humans will
dedicate their whole life to that task
and that's the cool thing that the human
mind is able to focus on one task and
then compete against each other and
achieve like weirdly incredible levels
of performance
that's the aspect of chess that's super
cool not that chess in itself uh is
really difficult it's it's it's like the
framazlast theorem is not in itself to
me that interesting the fact that uh
thousands of people have been struggling
to solve that particular problem is
fascinating so can i tell you my disease
in this way sure
which actually is closer to what you're
saying so as a child you know i was
building various i called them computers
they weren't general purpose computers
ice cube tray the ice cube tray was one
but i built other machines and what i
like to build was machines that could
beat adults at a game and they couldn't
they adults couldn't beat my machine
yeah so you were like uh that's powerful
like that's uh that's a way to rebel
yeah i i by the way um
did you
when was the first time you built
something that outperformed you do you
remember like well
i knew how it worked i was probably nine
years old and i built a thing that
it was a game where you you take turns
in taking
matches from a pile and
either the one who takes the last one or
the one who doesn't take the last one
wins i forget and so it was pretty easy
to build that out of wires and nails and
little coils that were like plugging in
the number and a few light bulbs
um the one the one i was prouder of i
was 12 when i built a
a thing out of old
telephone switchboard switches that
could uh always uh win at tic-tac-toe
and that was a much harder circuit
to design but again it was just
it was no active components it was just
three position switches empty x
zero
and um
and nine of them and and a light bulb
one which which move it wanted next and
then yeah the human would go and move
that see there's magic in that creation
yeah yeah i tend to uh i tend to see
magic in robots that
like i i also think that intelligence is
uh is a little bit overrated i think we
can have deep connections with robots
very soon
and well we'll come back to connections
robot sure but but i do want to say
i i don't i i think people too many
people make the mistake of seeing that
magic and thinking well we'll just
continue you know but each each one of
those is a hard-fought battle for the
next step the next step yes i mean the
open question here is and this is why
i'm playing devil's advocate but i often
do when i read your blog post in my mind
because i have like this eternal
optimism is it's not clear to me
so i don't do what obviously the
journalists do or they give into the
hype but it's not obvious to me how many
steps away we are from
from uh
truly
transformational
understanding of what it means um
to build intelligence systems like or
how to build intelligence systems
i'm also aware of the whole history of
artificial intelligence which is where
your deep grounding of this is
is there's been an optimism for decades
and that optimism just like reading old
optimism is absurd because people were
like
this is they were saying things are
trivial for decades since the 60s
they're saying everything is true
computer vision is trivial but
i think my mind is
working crisply enough to where i mean
we can dig into if if you want i'm
really surprised by
the things deepmind has done i don't
think they're so they're
yet
um
close to solving intelligence but i'm
not sure it's not 10 to 10 years away
what i'm referring to is interesting to
see when the engineering
um
it takes that idea to scale
and this and the idea what and no it
fools people
okay honestly rodney if it was you me
and demis inside a room forget the press
forget all those things just as a
scientist as a roboticist you know that
wasn't surprising to you that at scale
so we're talking about very large now
okay okay let's pick one that's the most
surprising to you okay please don't yell
at me
gpt3
okay
i was honestly gonna bring that out okay
thank you
alpha zero alpha go alpha go zero alpha
zero and then alpha fold one and two
so are any do any of these kind of have
this core of uh
no forget usefulness or application or
so on which you could argue for alpha
fold like as a scientist
was doors surprising to you that it
worked
as well as it did
okay so if we're going to make the
distinction between
surprise and usefulness and and and
i'll i have to explain
this i would say alpha fold
and
one of the problems at the moment with
alpha fold is
you know it gets a lot of them right
which is a surprise to me because
they're a really complex thing uh but
you don't know which ones it gets right
which
then is a bit of a problem now they've
come out with a reason you mean the
structure of the protein it gets a lot
of those right yeah it's a it's
surprising right yeah it's been a really
hard problem so that was a surprise how
many it gets right
so far the usefulness is limited because
you don't know which ones are right or
not and now they've come out with a
thing in the last few weeks which is
trying to get a useful tool out of it
and they may well do it um in that sense
the least alpha fold is different
because your
alpha fold two is different
because now it's producing data sets
that are actually uh
you know potentially revolutionizing
computational biology like they will
actually help a lot of people
but you would say
uh potentially revolutionizing we don't
know yet but yeah that's true yeah but
they're you know but i i i got you i
mean this is okay so you know what this
is gonna be so fun so let's go
right into it speaking of robots that
operate in the real world
let's talk about self-driving cars
oh
okay
because you do you have built robotics
companies you're one of the greatest
roboticists in history and that's not in
space
just in the space of ideas we would also
probably talk about that but in the
actual
building and execution of businesses
that make robots that are useful for
people and that actually work in the
real world and make money
you also sometimes are critical
of mr elon musk or less more
specifically focused on this particular
technology which is autopilot inside
teslas
what are your thoughts about tesla
autopilot or more generally vision-based
machine learning approach to
semi-autonomous driving
uh these are robots they're being used
in the real world by hundreds of
thousands of people
and
if you want to go there i can go there
but that's not too much which there is
let's say they're on par safety wise as
humans currently meaning
human alone versus human plus robot okay
so first let me say i really like the
car i came here in here today which is
um a
2021 um
model uh mercedes e450
i am impressed by the um
machine vision
sonar other things i'm impressed by what
it can do i'm really impressed um
with
many aspects of it and and i'm
um it's able to stay in lane is it uh oh
yeah it does it does the lane stuff um
it uh
you know it's it's looking up on either
side of me it's telling me about nearby
cars or blind spots and so on yeah when
i when i when i'm when i'm
going in close to something in the park
i get this beautiful gorgeous top-down
view of the world i am impressed
up the wazoo of how you know
registered and metrical so it's like
multiple cameras and it's all right
together to produce the 360 view kind of
360 view you know synthesized as though
it's above the car and it is
unbelievable
i got this car in january it's the
longest i've ever owned a car without
digging it um so it's better than me
for me and it together uh better so i'm
not saying technology's um
uh bad or not useful
but here's my point yes
it's
it's a replay of the same movie
okay
so maybe you've seen me ask this
question before but um
when um
when when did the first car go over 55
miles an hour
for over
um
over 10 miles
on a public freeway with other traffic
around driving completely autonomously
when did that happen
was it cmu in the 80s or something it
was a long time ago it was actually in
1987. in in munich um munich uh at the
bundesvar yeah um
so they they had it running in 1987.
when do you think and elon has said he's
going to do this
when do you think we'll have the first
car drive coast to coast in the us
hands off the wheel hands off the wheel
feet off the pedals coast to coast
as far as i know a few people have
claimed to do it 1995. that was comedy i
didn't know about oh that was the code
yeah uh they didn't claim did they claim
a hundred percent not a hundred percent
not a hundred percent but
and then there's a few marketing people
who have claimed 100 percent since but
my point is that
you know i i what i see happening again
is someone sees a demo
and they over generalize and say we must
be almost there well we've been we've
been working on it for 35 years because
that's demos but this is going to take
us back the same conversation with alpha
zero are you not
okay i'll just say what i am
because i thought okay when i first
started interacting with the
with the mobile eye implementation tesla
autopilot
i've
driven a lot of car you know i have been
in the google self-driving car since the
beginning
um
i thought there was no way before i sat
and used mobileye i thought they're just
knowing computer vision i thought
there's no way it could work as well as
it was working so
my model of the limits of computer
vision
uh was
was way more limited than the actual
implementation of mobile eye i was so
that's one example i was really
surprised i was like wow that was that
was incredible
the second surprise came
when tesla threw away mobileye
and started from scratch
uh i thought there's no way they can
catch up to mobile eye i thought what
mobile i was doing was kind of
incredible like the amount of work and
the annotation yeah well mobile i was
started by amazon structure and and used
a lot of traditional you know
hard-fought computer vision techniques
but they also did a lot of good sort of
uh like non-research stuff like actual
like
uh just good like what you do to make a
successful product right scaled all that
kind of stuff and so i was very
surprised when they from scratch were
able to catch up to that
uh that's very impressive and i've
talked to a lot of engineers though i
was involved this is
that was impressive uh and
the recent progress especially under um
well with the involvement of andre
kupathi
the what they were
what they're doing with the data engine
which is converting into the driving
task into these multiple tasks and then
doing this edge case discovery when
they're pulling back like the level of
engineering
made me rethink what's possible i don't
i still you know um
i don't know to that intensity but i
always thought it was very difficult to
solve anton was driving with all the
sensors with all the computation i just
thought it was a very difficult problem
but i've been
continuously surprised
how much you can engineer first of all
the data acquisition problem because i
thought
you know just because i worked with a
lot of car companies
and
they're
they're so a little
a little bit old school to where i
didn't think they could do this at scale
like aws style
data collection so when tesla was able
to do that
i started to think okay so what are the
limits of this
i still believe
that um
driver like sensing and the interaction
with the driver and like studying the
human factors psychology problem is
essential it's it's always going to be
there
it's always going to be there even with
fully autonomous driving
but i've been surprised what is the
limit especially of vision based alone
how far that can take us
um so that's my level is a surprise
now
okay uh can can you explain
in the same way you said like alpha zero
that's a homework problem that scaled
large in his chest like who cares go
with here's actual people using an
actual car and driving many of them
drive more than half their miles
using the system right
so and
yeah they're doing well with with pure
vision without your vision yeah and you
know
and now no radar which is i suspect that
can't go all the way and one reason is
without without new cameras that have a
dynamic range closer to the human eye
because human eye has incredible dynamic
range and we make use of that dynamic
range in its uh 11 orders of magnitude
or some crazy number like that
the cameras don't have that which is why
you see the the the the bad cases where
the sun on a white thing and the blinds
are
in a way it wouldn't apply in the person
i think there's
a bunch of things
to think about before you say this is so
good it's just gonna work
okay
um
and i'll come at it from multiple angles
and i know you've got a lot of time yeah
okay let's
have thought about these things yeah i
know you've been writing a lot of great
blog posts about it for a while before
tesla had autopilot right so you've been
thinking about autonomous driving for a
while from every angle
so so a few things
you know in the us um i think that the
the death rate from
motor vehicle accidents
is about 35
a year
which is an outrageous number
not outrageous compared to covert deaths
but you know there is no rationality
and that's part of the thing people have
said engineers say to me well if we cut
down the number of deaths by 10 by
having autonomous driving that's going
to be great everyone will love it
and
my prediction is
that if autonomous vehicles kill more
than 10 people a year they'll be
screaming and hollering even though 35
000 people a year have been killed by
human drivers
it's not rational
it's a different set of expectations
and that will probably continue
so there's that aspect of it
the other aspect of it is that
when we introduce new technology
we often change the rules of the game
so when we introduced
cars
first
you know into our daily lives we
completely rebuilt our cities and we
changed all the laws yeah jaywalking was
not an offense
that was pushed by the car company so
that people would stay off the road so
there wouldn't be deaths from
pedestrians getting hit
we completely changed the structure of
our cities and had these foul smelling
things you know everywhere around us and
you know and now you see pushback in
cities like barcelona is really trying
to exclude cars
etc
um
so
i think that
to get to self-driving
we will
um
large adoption
it's not going to be just take the
current situation take out the driver
and put the same car doing the same
stuff because the end case is too many
um
here's an interesting question
how many
fully autonomous
train systems do we have in the us
i mean do you count them as fully
autonomous i don't know because they're
usually as a driver but they're kind of
autonomous right no well
let's get rid of the driver
okay i don't know
it's either 15 or 16. most most of them
are in airports
okay um there's a few that go about five
two that go about five kilometers out of
airports yeah
um
uh
when do when is the first
fully autonomous train system for mass
transit expected to operate fully
autonomously
with no driver uh in the u.s city
it's expected to operate in 2017
in honolulu
it's delayed but they will get there but
by the way it was originally going to be
autonomous uh here in the bay area i
mean they're all
very close to fully autonomous right
yeah but
getting the clues is the thing and i
have i have i've often gone on a fully
autonomous train in japan um one that
goes uh out to that fake island in the
middle of tokyo bay i forget the name of
the
and
and what do you what do you see when you
look at that what do you see when you go
to a fully autonomous train in a in a in
a
um an airport
it's not like regular trains there's
at every station there's a double set of
doors um so that there's the door of the
train and this door off the the um
off the the platform yeah um
and it's really visible in this japanese
one because it goes out in in amongst
buildings
the whole track is built so that people
can't climb onto it yeah
so there's engineering that then makes
the system safe and makes them
acceptable i think we'll see
similar sorts of things happen
in the u.s what surprised me
i thought
wrongly
that we would have
special purpose lanes on 101
in the bay area the the leftmost lane
so that it would be normal for
teslas or other cars to move into that
lane and then say okay now it's
autonomous and have that dedicated lane
i was expecting
movement to that you know five years ago
i was expecting we'd have a lot more
movement towards that we haven't and it
may be because
tesla's been over promising by saying
this you know calling their system fully
self-driving um i think they may have
been gotten there quicker by
collaborating
to change the infrastructure
this is one of the problems
with
long-haul trucking
being autonomous i think it makes sense
on freeways
at night for the trucks to go
autonomously um
but then is that how to get onto and off
of the freeway what sort of
infrastructure do you need for that
um do you need to have the human in
there to do that
or can you get rid of the human so i
think there's ways to get there but it's
an infrastructure
argument
because the long tail of cases
is very long
and the acceptance of it will not be at
the same level as human drivers
so i'm
i'm with you still and i was with you
for a long time but i am surprised how
how well how many edge cases
of machine learning and vision based
methods can cover
this this is what i'm trying to get get
at is um
i think there's something fundamentally
different with vision based methods and
tesla autopilot and any company that's
trying to do the same
okay well i'm not i'm not going to argue
with you because you know i
i
was speculating yes but i
you know my gut feeling tells me
it's going to be
things will things will speed up when
there is engineering of the environment
because that's what happened with every
other technology
i'm a bit i don't know about you but i'm
a bit cynical that
infrastructure which relies on on
government
to help out
in these cases um
if you just look at infrastructure in
all domains it's just a government
always drags behind on infrastructure
there's like there's so many just well
in this country in the future sorry yes
in the in this country and and of course
there's many many countries that are
actually much worse on infrastructure oh
yes there's nothing many are much worse
and there's some that you know like
high-speed rail the
other countries have done much better i
guess uh my question is like
which is at the core what i was trying
to think through here and ask is like
how hard is the driving problem
as it currently stands so you mentioned
like we don't want to just take the
human out and duplicate whatever the
human was doing but if we were to try to
do that
what
how hard is that problem
because i used to think is way harder
like i i used to think it's uh
with vision alone it it it would be
three decades four decades okay so i i
don't know the answer to this thing i'm
about to pose
but i do notice that on highway 280 here
in the bay area which largely has
concrete
surface rather than blacktop surface
the white lines that are painted there
now have black boundaries around them
and
my
lane drift system in my car would not
work without those black boundaries
interesting so i don't know whether
they've started doing it to help the
lane drift whether it is an instance of
infrastructure
following the technology but but it
my car would not perform as well without
that change in the way they paint the
line unfortunately really good lane
keeping
is not as valuable
like
it's orders of magnitude more valuable
to have a fully autonomous system like
yeah but but for me
lane keeping is really helpful because
i'm busy at it but you wouldn't pay
10 times like
um
the problem is there's not financial
like it doesn't make sense to to to uh
revamp the infrastructure to make lane
keeping easier
it does make sense to prevent the
infrastructure if you have a large fleet
of autonomous vehicles now you change
what it means to own cars you change the
nature of transportation i mean but that
that for that you need
uh autonomous vehicles let me ask you
about waymo then i've gotten a bunch of
chances to to ride in in a waymo um
self-driving car and they're
i don't know if you'd call them
self-driving but well i mean i i wrote
in one before they were called waymo
yeah still at x so there's currently
this was a big another surprising leap i
didn't think it would happen which is
they have no driver
currently yeah in chandler in chandler
arizona and i think they're thinking of
doing that in austin as well but they're
they're expanding
although although you know and i i do an
annual uh checkup on this so as of late
last year they were aiming for hundreds
of rides
a week
not thousands and
um there is still no one in the car but
there's certainly uh
um safety uh people in the loop and it's
not clear how many you know what the
ratio of cars to safety people is
i it wasn't
uh obviously they're not 100 transparent
about this no none of them are 100
transfers
but i'd
at least the way they're
i don't want to make definitely but
they're saying there's no tele operation
um
like they're i mean
okay and and and that sort of fits with
with um youtube videos i've seen of
people being trapped in the car yeah um
by a red cone on the on the street and
they do they do have rescue vehicles
that come yeah and then a person gets in
and drives it yeah
but isn't it incredible to you
it was to me to get in a car with no
driver and watch the steering wheel turn
like for somebody who has been studying
at least certainly the human side of
autonomous vehicles for many years and
you've been doing it for way longer like
it was incredible to me that this was
actually could happen i don't care if
that scale is a hundred cars this is not
a demo this is not this is me as a
regular the argument i have is that
people make interpolations from that
interpolation that you know it's here
it's done
um you know it's just you know we've
solved that no we haven't yet
and and that's my argument okay so i'd
like to go to
you uh you keep a list of predictions
on your amazing blog post it'd be fun to
go through them but before then let me
ask you about this
you have um
you have a
harshness to you sometimes in your
criticisms of what is
and so like because people extrapolate
like you said and they they kind of buy
into the hype and then they
they kind of start to think that um
uh the technology is way better than it
is but
let me ask you maybe a difficult
question
sure
do you think if you look at history of
progress
don't you think to achieve the quote
impossible you have to believe that it's
possible
absolutely yeah like his his his his
two great runs
great unbelievable
first human
um power human uh
you know heavier than their flight yeah
1969 we land on the moon that's 66 years
i'm 66 years old in my lifetime that
span of my lifetime
barely get you know flying i don't know
what it was 50 feet or the length of the
first flight or something to landing on
the moon
unbelievable
fantastic but that requires by the way
one of the wright brothers
both of them but one of them didn't
believe it's even possible like a year
before
right so
like not just possible soon but like
yeah ever so so so you know how
important is it to believe and be
optimistic is what i guess oh yeah it is
important it's when it goes crazy when
when i you know you said what was the
word you used for my bad harshness
harshness yes
i just get so frustrated yes when when
people make these leaps
and tell me that i'm that i don't
understand right i
you know yeah
there's
just from
irobot which i was co-founder of yeah i
don't know the exact numbers now because
i haven't it's 10 years since i stepped
off the board but i believe it's well
over 30 million
robots cleaning houses from that one
company and now there's lots of other
companies yesterday
was that a crazy idea that we had to
believe
uh
in 2002 when we released it
yeah that was we we had we had uh you
know
believed that it could be done let me
ask you about this so irobot one of the
greatest robotics companies ever
in terms of manufacturing
creating a robot that actually works in
the real world probably the greatest
robotics company ever
you're the co-founder of it
um
if if the rodney
brooks of today talked to the rodney of
back then
what would you tell him because i have a
sense that
would you pet him on the back and say
what you're doing is going to fail
but go at it anyway
that's what i'm referring to is with the
harshness you've accomplished
an incredible thing there one of several
things we'll talk about
what like that's what i'm trying to get
at that line
no it's it's when
my harshness is reserved for people who
are not doing it
who claim it's just well this shows that
it's just gonna happen but here here's
the thing
this shows but you have that harshness
for elon too
and no no it's a different harshness no
it's it's a
different um
argument with yuan you know i
i think spacex is an amazing company on
the other hand you know i in one of my
blog posts i said what's easy and what's
hard i said space x
vertical landing rockets it had been
done before
grid fins have been done since the 60s
every soyuz has them um
reusable space
dcx reused those rockets that landed
vertically
there's a whole insurance industry in
place for
rocket launchers so all sorts of
infrastructure
that
was doable it took a great entrepreneur
a great personal
expense he almost drove himself you know
bankrupt doing it
um a great belief to do it
whereas
hyperloop
there's a whole bunch more stuff that's
never been thought about never been
demonstrated so my estimation is
hyperloop is a long lot long a lot
further off
and and if i've got a criticism of of of
elon it's that he doesn't make
distinctions
between
when the technology's
coming along and ready and then he'll go
off and and
mouth off about other things which then
people go and compete about and try and
do
and
so
this is where i um i i understand what
you're saying i tend to draw a different
distinction i
i have
a similar kind of harshness towards
people who are not telling the truth
who are
basically fabricating stuff to make
money or to well he believes what he
says i just think to me that's a very
important difference yeah i'm not
because i think uh in order to fly
in order to get to the moon you have to
believe um even when uh most people tell
you you're wrong and most likely you're
wrong but sometimes you're right i mean
that's the same thing i have with tesla
autopilot i i think that's an
interesting one i was especially when i
was you know um at mit and just the
entire human factors in the robotics
community were very negative towards
elon it was very interesting for me to
observe colleagues at mit
i wasn't sure what to make of that that
was very upsetting to me
because i understood where that where
that's coming from
and i agreed with them and i kind of
almost felt the same thing in the
beginning until i kind of opened my eyes
and and
realized there's a lot of interesting
ideas here there might be over hype you
know
if if you focus yourself on the idea
that
you shouldn't call a system full
self-driving
when it's obviously not
autonomous fully autonomous
you're going to miss the magic oh
yeah you are going to miss the magic but
at the same time there are people who
buy it
literally pay money for it yeah and take
those words as given so it's that's uh
but i haven't so
that i
take words as given as one thing i
haven't actually seen people that use
autopilot that believe that the behavior
is really important like the actual
action so like this is like to push back
on the very thing that you're frustrated
about which is like journalists in
general people
uh
buying all the hype and going on
in the same way
i think there's a lot of hype about
the the negatives of this too that
people are buying without using people
used the way this is what this was this
opened my eyes actually
the way people use the product is very
different than the way
they talk about it this is true with
robotics with everything everybody has
dreams of how a particular product might
be used or so on this and then when it
meets reality there's a lot of fear of
robotics for example that robots are
somehow dangerous and all those kinds of
things but when you actually have robots
in your life whether it's in the factory
or in the home making your life better
that's going to be
that's way different your perceptions of
it are going to be way different and so
my
just tension was was like here's an
innovator
um
uh like uh
uh what is it sorry super cruise from
cadillac was super interesting too
that's a really interesting system
there's we should like be excited by
those innovations okay so let me can i
tell you something that's really annoyed
me recently
it's really annoyed me that
the press and friends of mine on
facebook are going these billionaires
and their space games you know why are
they doing that yeah that's been very
frustrating really pisses me off i i
must say i i applaud that yeah i applaud
it yeah it's the taking
and not necessarily the people who are
doing the things but
you know like that i keep having to push
back against
unrealistic expectations when
these things can become real yeah i
this was interesting ana because there's
been a particular focus for me is
autonomous driving
elon's prediction of when certain
milestones would be hit
there's several things to be said there
that i always i thought about because
whenever you said them it was obvious
that's not going to me as a person that
kind of
not inside
the system it was obvious it's unlikely
to hit those
there's two comments i want to make one
he legitimately believes it
and two much more importantly
i think
that
having ambitious deadlines drives people
to do the best work of their life even
when the odds of those deadlines are
very um
to a point and i'm not killed i'm not
talking about anyone yeah i'm just
saying so there's a line there right you
have to have a line because
you over extend and it's
it's demoralizing yeah
[Music]
but i will say that there's an
additional thing here
that those words
also
um
drive the stock market
yeah and you know we have because of the
way that rich people in the past have
manipulated
the rubes
through investment we have
um um we have
developed laws about what you know what
you're allowed to say and yeah i promise
and you know there's an area here which
is
i i i tend to be
maybe i'm naive but i i tend to believe
uh that
like engineers innovators people like
that they're not
they're my they don't think like that
like manipulating the price at the stock
price but
it's possible that i'm uh i'm certain
it's possible that i'm wrong
i it's a very cynical view of the world
because i don't i think most people that
run companies and build like especially
original founders
they um yeah i'm not saying that's the
intent i'm saying it's a eventually it's
kind of you uh yeah
you you fall into that kind of a
behavior pattern i don't know i i tend
to
i wasn't saying i wasn't saying it's
falling into that intent it's just a
you also have to protect investors in
this in this market yeah
okay so you have first of all you have
an amazing blog that people should check
out but you also have this in that blog
a set of predictions
it's such a cool idea i don't know how
long ago you started like three four
years ago it was um january 1st
2018 18. yeah and i made these
predictions and i said that every
january 1st i was going to check back on
how my predictions would be that's such
a great orthodontics for 32 years
oh you said 32 years i said 32 years
because it's still that'll be january
1st 2050 yeah i'll be
i will just turn 95
um
you know so
and so people know that your predictions
at least for now are in the space of
artificial intelligence yeah i didn't
say i was going to make new predictions
i was just going to measure this set of
predictions that i made because yeah it
was sort of i was sort of annoyed that
everyone could make predictions they
didn't come true and everyone forgot so
i should hold myself to a high standard
yeah but also just putting years and
like date rangers on things it's a good
thought exercise yeah like and like
reasoning your thoughts out and so the
topics are
uh artificial intelligence autonomous
vehicles and space yeah
um
i was wondering if we could just go
through some that stand out maybe from
memory i can just mention to you some
let's talk about self-driving cars like
some predictions that you're
particularly proud of or are
particularly interesting uh from flying
cars to the the other element here is
like how widespread the location
where the deployment of the autonomous
vehicles is
and there's also just a few fun ones is
there something that jumps to mind that
you remember from the predictions
well i did i think i did put in there
that there would be a
dedicated self-driving lane on 101 by
some year and i think i was over
optimistic on that one yeah actually
yeah i actually do remember that but you
uh i think you were mentioning like
difficulties in different cities
yeah yeah so
cambridge massachusetts i think was an
example yeah like in cambridge port you
know yeah i lived in cambridge port for
a number of years and
you know the roads are narrow and
getting getting anywhere as a human
driver is incredibly frustrating when
you start to put and people drive the
wrong way on one-way streets there it's
just
your prediction was
driverless taxi services operating on
all streets in cambridgeport
massachusetts
in uh
2035.
yeah and
that may have been too optimistic you
think so you know i've gotten a little
more pessimistic since i made these
internally on some of these things so
what uh
can you put a year to a major milestone
of deployment of a taxi service
in um
in a few major cities like something
where you feel like yeah so autonomous
vehicles are here so let's let's take um
the grid streets of san francisco north
of market okay okay
um
relatively benign
um
uh environment the streets are wide
the major
problem is
delivery trucks stopping everywhere
which made things more complicated
a taxi system there with um
somewhat designated pickup and drop-offs
unlike with uber and lyft where you can
sort of get to any place and the drivers
will figure out how to
get in there um
we're still a few years away i you know
i live in that area so i see
you know the self-driving car companies
cars
multiple multiple ones every day i'll
say
cruise
zooks less often
way more all the time
different and different ones come and go
and there's always a driver
there's always a driver at the moment
although i have noticed
that um
sometimes the driver does not have the
authority to take over without talking
to the home office because they will sit
there waiting for a long time
and clearly something's going on where
the home office is making a decision um
so they're you know and and so you can
see whether they've got their hands on
the wheel or not and and it's the
incident resolution time that tells you
gives you some clues
so what year do you think what's your
intuition what date range are you
currently thinking
san francisco would be autonomous
uh
taxi service from
any point a to any point b without a
driver
are you are you still
are you thinking uh 10 years from now 20
years from now 30 years from now
certainly not 10 years from now
it's going to be longer if you're
allowed to go south of market way longer
um
and unless it's re-engineering of course
roads
by the way what's the biggest challenge
you can mention a few is that the the is
it the delivery trucks is it the edge
case is the computer perception
uh well
here's a case that i saw outside my
house a few weeks ago
um about 8 pm on a friday night it was
getting dark it was before the solstice
it was a
cruise vehicle come down the hill uh
turned right um
and
stopped dead
covering the crosswalk why did it stop
dead because there was a
human just two feet from it
now i just glanced i knew what was
happening the human
was it was a woman was at the door of
her car trying to unlock it with one of
those things that yeah you know when you
don't have a key yes
that car
thought
oh she could jump out in front of me any
second yeah as a human i could tell no
she's not going to jump out she's busy
trying to unlock her she's lost her keys
she's trying to get in the car
and it it stayed there for
until i got bored um yeah
and so the the human driver in there did
not take over
but here's the kicker to me
a guy comes down the hill
with a stroller i assume there's a baby
in there
and
now the crosswalk's blocked by this
cruise vehicle
what's he going to do
cleverly i think he decided not to go in
front of the car
he went but he had to go behind it he
had to get off the crosswalk out into
the intersection
to push his baby around this car which
was stopped there and no human driver
would have stopped there for that length
of time um they would have gotten out of
the way
and that's another one of
my pet peeves
that
safety has been compromised
for individuals who didn't sign up for
having this happen in their
neighborhood
yeah but
now you can say that's an edge case but
yeah well i'm in general
not a fan which
of uh anecdotal evidence for stuff like
this is one of my biggest problems
with the discussion of autonomous
vehicles in general people that
criticize them or support them by using
cases okay uh uh aren't using anything
so so let me but i got you you know you
know your question is when is it going
to happen in san francisco i say not
soon but now it's going to be one of
them but when where it is going to
happen
is in
limited domains
campuses of various sorts
gated communities
where the other drivers
are
are not arbitrary people
they're people who know about these
things they you know it's been warned
about them
and at velocities where it's always safe
to stop dead yeah um you can't do that
on the freeway that i think we're going
to start to see
and they may not be shaped like
you know current cars they may be you
know things like you know may mobility
has those
things and various companies have these
yeah i wonder if that's a compelling
experience to me it's always important
it's not just about automation it's
about creating a product that like
that makes your it's not just cheaper
but it makes your this fun to ride one
of the most
one of the least fun things is for a car
that stops
and like waits there's something deeply
frustrating for us humans
for the rest of the world to take
advantage of us as we wait but
um
think about uh you know not you as the
customer but someone who's in their
80s
in an uh you know a retirement village
whose kids have said you're not driving
anymore
and this gives you the freedom to go to
the market more that's a hugely
beneficial thing but
it's a very uh few orders of magnitude
less impact on the world it's not it's
just a few people in a small community
using cars as opposed to the entirety of
the world uh
i like that uh the first time that a car
equipped with some version of a solution
to the trolley problem is uh what's niml
stand for like not in my life not in my
life i define my lifetime as up 2050
2015.
yeah
uh you know and then i ask i ask you
when when have you had to decide which
person should i kill um no you put the
brakes on and you break as hard as you
can
i mean uh making that decision it is uh
you know i do think autonomous vehicles
or semi-autonomous vehicles do need to
solve the whole pedestrian problem that
has elements of the trolley problem
within it but it's not yeah well so
here's a and i talked about it in one of
the articles or blog posts that i wrote
his his and people have told me i one of
my co-workers has told me he does this
he
he tortures
autonomously driven vehicles and
pedestrians will will torture them
they'll you know once they realize that
you know putting one foot off the curb
makes the car think that they might walk
into the road kids teenagers will be
doing that all the time they will
i by the way one of my that's a whole
nother discussion because my main issue
with robotics is hri human robot
interaction
i believe that robots that interact with
humans will have to
push back
like they can't just be
bullied because that creates a very
uncompelling experience for the humans
yeah well you know waymo before it was
called waymo discovered that you know
they had to do that at four-way
intersections they had to they had to
nudge forward to give the cue that they
were going to go because otherwise the
other drivers would just
beat them all the time
so you co-founded irobot as we mentioned
uh one of the most successful
robotics companies ever
what are you most proud of with that
company and uh the approach
you uh
took to robotics well like there's
something i'm quite proud of there
which may be a surprise
but um
i was still on the board when this
happened it was march 2011
and we
sent robots to japan and they were used
to
uh shut help shut down
the fukushima fukushima daiichi nuclear
power plant
um which was everything i've been there
since i was there in 2014 to the robots
some of the robots were still there i
was i was proud that we were able to do
that
why were we able to do that
and and you know people have said well
you know japan is so good at robotics
it was because we had had
about 6 500 robots
uh deployed
in iraq and afghanistan
teleopt but with intelligence
dealing with
roadside bombs
so we had uh i think it was at that time
nine years of
in-field
experience with the robots in harsh
conditions
whereas the japanese robots which were
you know getting you know just goes back
to what
annoys me so much getting all the hype
look at that look at that honda robot it
can walk well the future's here
um couldn't do a thing uh because they
weren't deployed but we had deployed in
really harsh conditions for a long time
and so we're able to
to
do something very positive in a very bad
situation what about
just the simple
and for people who don't know one of the
things that irobot has created is the
roomba
uh vacuum cleaner
what about the simple robot
that that is the room bus quote-unquote
simple
that's deployed in
tens of millions of in tens of millions
of homes
what do you think about that
well i make the joke that i started out
life as a
pure mathematician and turned into a
vacuum cleaner salesman so
if you're going to be an entrepreneur be
ready for that
be ready
um but i was you know
there was a there was a
wacky uh lawsuit that i got posed for uh
not too many years ago
and i was the only one who had emailed
from the 1990s
and
no one in the company had it so i went
and went through my email and and it
reminded me of
you know the joy of what we were doing
and and what what was i doing what was i
doing at the time we were building
um
building
uh the roomba
one of the things was we had this
incredible incredibly tight budget
because we wanted to to put it on the
shelves at
there was another home cleaning robot at
the time it was the
electrolux
trilobite
which sold for 2 000 euros
and to us that was not going to be a
consumer product
so we
had reason to believe that 200 was a was
a thing that people would buy at
that was our aim but that meant we had
you know that's that's
on the shelf making profit
uh that means the cost of goods has to
be minimal
so i find all these emails of me
going you know i'd be in um
taipei for a mit meeting and i'd stay a
few extra days and go down to shinshu
and talk to these little tiny companies
lots of little tiny companies outside of
uh tsmc taiwan semiconductor
taiwan semiconductor manufacturing
corporation which
let all these little companies be
fabulous they didn't have to have their
own fab so they could innovate
and then
um
they were building their innovations
were built stripped down 6802s
1682 was what was in an apple one get
rid of half the silicon still have it be
viable
and i'd i'd previously got
some of those for some earlier failed
products of of a robot and um and that
was um in hong kong
going to all these um
companies that built you know
they weren't gaming in the current sense
there were these handheld games that you
would play
um or
or birthday cards because we had about a
50 cent budget for computation
so i'm trekking from place to place
looking at their chips
looking at what they'd removed
their interrupt their interrupt handling
is too weak for a general purpose so i
was going deep technical detail and then
i found this one from a company called
winbond which had
and i'd forgotten it had this much ram
it had 512 bytes of ram and it was in
our budget and it had all the
capabilities we needed yeah so and
you're excited yeah and i i was reading
all these emails calling i found this
so did you think did you ever think that
you guys could be so successful
like eventually this company would be so
successful did you could you possibly
have imagined
um no we never did think that we had 14
failed business models up till 2002 and
then we had two winners the same year um
uh
no and then you know
we i remember the board
um because by this time we had some uh
venture capital in
the board went along with us building um
some robots for you know aiming at the
christmas 2002
market and
we went three times over what they
authorized and built 70 000 of them and
sold them all
in that first because we released on
september 18th
and uh
all sold by christmas so it was uh
so we were gutsy but
but yeah you didn't think this will take
over the world well this is uh
so a lot of amazing robotics companies
have gone under over the past few
decades
why do you think it's so damn hard
to uh run a successful well there's a
robotics company there's a few things um
one is
expectations of capabilities by the
founders that are
off base
the founders not the consumer and the
founders yeah expectations what what can
be delivered sure
mispricing
and
what a customer thinks is a valid price
is not rational
necessarily yeah
and expectations of customers
and
just the
sheer hardness of getting people
to adopt a new technology and i've
suffered from all three moons
uh you know i've had
more failures and successes
in terms of companies i've suffered from
all three
um
so
do you think
one day there will be a robotics company
and by robotics company i mean where
your primary source of income is
from robots
that will be a trillion plus dollar
company
and it's so what come what would that
company do
i can't you know because i'm still
starting robot companies yeah
i'm not making any such predictions in
my own mind i'm not thinking about a
trillion dollar company and by the way i
don't think you know in the 90s anyone
was thinking that apple would ever be a
trillion-dollar company so these are
these are very hard to to predict but
sorry to interrupt but don't you because
i kind of have a vision
in this in a small way and it's a big
vision in a small way that
i see that there would be robots in the
home
at scale like roomba but more
and that's trillion dollar
right and i and i think there's a
there's a real market pool for them
because of the um um demographic
inversion you know who's who's going to
do all the stuff for the older people um
there's too many i'm you know i'm i'm
leading here
it's going to be too many of us
and
um
but
we don't have capable enough robots to
to make that economic argument at this
point
do i expect that that will happen yes i
expect it will happen but i gotta tell
you we introduced the roomba in 2002 and
i stayed another
nine years we were always trying to find
what the next home robot would be and
still today the primary product of 20
years almost 20 years later 19 years
later the primary product is still the
roomba so irobot hasn't found the next
one do you think it's possible for one
person in the garage to build it versus
like google
launching google self-driving car that
turns into waymo you think it's pos
this is almost like what it takes to
build a successful robotics company do
you think it's possible to go from the
ground up or is it just too much capital
investment
yeah so
it's very hard to get there
um
without a lot of capital and we're
starting to see
you know a fair chunks of capital uh for
some robotics companies um you know
series b's because i saw one yesterday
for 80 million dollars i think it was
for co-variant um
[Music]
but it can take real money to
to get into these things and you may
fail along the way i've certainly failed
at rethink robotics um and we've all
lost 150 million dollars in capital
there so okay so rethink robotics is
another amazing robotics company you
co-founded
so what was the vision there
what was
the dream and
what what are you most proud of with
rethink robotics i'm most proud of the
fact that we got um
robots out of the cage in factories
that were safe absolutely safe for
people and robots to be next to each
other so these are robotic arms robotics
arms they're able to pick up stuff and
interact with humans yeah and that
humans could re-task them
without writing code
right and and now that sort of become an
expectation for a lot of other little
companies and big companies are
advertising they're doing that's both an
interface problem and also safety
problem yeah
yeah
so i'm most proud of that
i completely i let
myself be talked out of
what i wanted to do and you know you've
always got you know i can't replay the
tape
you know i can't replay it maybe
maybe i you know if i'd been stronger um
and i remember the day i remember the
exact meeting
um can you take me through that meeting
yeah
um
so i said that i'd set as a target for
the company that we were going to build
three thousand dollar robots with force
feedback
um that was safe for people to be around
wow that was my goal
and we built uh so we started in 2008
and we had prototypes built of plastic
plastic gearboxes
and at a three thousand dollar you know
uh
lifetime of three thousand dollar
i was
saying we're going to go after not the
people who already have robot arms in
factories the people who never have a
robot arm we're going to go after a
different market so we don't have to
meet their expectations
um
and and so we're going to build it out
of plastic it doesn't have to have a 35
000 lifetime it's going to be so cheap
that it's opex not capex
and so we had we had a prototype that
worked reasonably well
but
the control engineers were complaining
about these plastic gearboxes with a
beautiful little planetary gearbox um
but
we could
use something called serious elastic
actuators we embedded them in there we
could measure forces we knew when we hit
something et cetera
the control engineers were saying yeah
but this is torque ripple because these
plastic gears they're not great gears
and there's this ripple and trying to do
force control around this ripple is so
hard
and i'm not going to name names but i
remember one of the mechanical engineers
saying we'll just build a metal gearbox
with spur gears
and it'll take six weeks we'll be done
problem solved
two years later we got to get the spur
gearbox working yeah um we we cost
reduced it every possible way we could
um yeah but
now the price went up to and then the
ceo at the time said well we have to
have two arms not one arm so our first
robot product baxter now cost 25 000
and the only people who are going to
look at that were people who had arms in
factories because that was somewhat
cheaper for two arms than arms and
factories but they were used to
0.1 millimeter reproduce
reproducibility of motion
and certain velocities
and we i kept thinking but that's not
what we're giving you you don't need
position repeatability use force control
like a human does no
no but we want we want that
repeatability we want that repeatability
yes all the other robots have that
repeatability why don't you have that
repeatability so
you clarify force controls you can grab
the arm and you can move it or you can
move it around but but suppose you um
can you see that yes suppose you want to
yes
suppose this this thing is a you know
precise thing that's going to fit here
in this right angle
um under position control you sent your
you you have fixtured where this is you
know where this is precisely and you
just move it open you know and it goes
there if force control you would do
something like
slidell over here till we feel that and
slide it in there and that's how a human
gets precise precision yeah they use
force feedback yes and get the things to
mate rather than just
go straight to it yeah
couldn't convince couldn't convince our
customers who were in factories and were
used to thinking about things a certain
way
and they wanted that one wonder so then
we said okay we're going to build an arm
that gives you that
so now we ended up building a 35 000
robot with one arm with um
um
oh what are they called um
um
a certain sort of gearbox made by a
company whose name i can't remember
right now but it's the name of the
gearbox
um
and um but it's it's
got torque ripple in it
so now there was an extra two years of
solving the problem of doing the force
with the talk ripple so we had to do the
the thing we had avoided
and
for the plastic gearboxes we ended up
having to do the robot was now
overpriced and
um they and that was your intuition from
the very beginning kind of that this is
not
you're opening a door to to solve a lot
of problems that you're you're
eventually going to have to solve this
problem anyway yeah and also i was
aiming at a low price to go into a
different market price that that didn't
have a thousand dollars would be amazing
yeah i think we could have done it for
five um but
you know you said talked about
setting the goal a little too far for
the engineers exactly
so why would you say that company um
not failed but went under
we had buyers and um
there's this thing called the committee
on foreign investment in the us cyphus
and um that had previously been invoked
twice
around where the government could stop
foreign money coming into a u.s company
based on
defense requirements
went through due diligence multiple
times we were going to get acquired
um but every consortium had chinese
money in it
and all the bankers would say at the
last minute you know this isn't going to
get past cyphus
and the investors would go away
and then we had two buyers
once we were about to run out of money
two buyers
and one used heavy-handed legal stuff
with the other one
said they were going to take it
and pay more
dropped out when we were out of cash and
then bought the assets
at 1 30th of the price they had offered
a week before
it was a tough week
do you um does it hurt to think about
like an amazing company that didn't
you know like a robot didn't find a way
yeah it was tough um i said i was never
going to start another company i was
pleased that everyone liked what we did
so much that the teams
the
team was hired by um
three companies within a week everyone
had a job in one of these three
companies some stayed in their same
desks because the com another company
came in and rented the space
so i felt good about people not being
out on the street
so baxter has a screen with a face
what uh
that's the revolutionary idea for a uh
robot manipulation like for a robotic
arm
uh what
proposition did you get well first the
screen was also used during um
codeless programming where you taught by
demonstration it showed you what its
understanding of the task was
so it had two roles
um
some customers hated it
and so we made it so that when the robot
was running it could be showing graphs
of what was happening i'm not sure the
eyes other comp and other
people and some of them surprised me who
they were saying
well this one doesn't look as human as
the old one we like the human looking
yeah so there was a mixed bag
but do you think that's uh
i don't know i i'm kind of disappointed
whenever i talk to um
to roboticists
like the best robotics people in the
world they seem to not want to do the
eyes type of thing like they seem to see
it as a machine
as opposed to a machine that can also
have a human connection
i'm not sure what to do with that it
seems like a lost opportunity i think
the trillion dollar company
will have to do the human connection
very well no matter what it does yeah i
agree
can i ask you a ridiculous question sure
i want to give a ridiculous answer
uh do you think uh
well maybe by way of asking the question
let me first mention that you're kind of
critical of the idea of the touring test
as a test of intelligence
let me first ask this question
do you think we'll be able to build
an ai system
that humans fall in love with and it
falls in love with the human
like romantic love
but we've had that with humans falling
in love with cars even back in the 50s
it's a different love right well i think
i think there's a lifelong partnership
where you uh
can communicate and grow like
i think we're a long way from that
i think we're a long long way i think
uh
blade runner was you know had the time
scale totally wrong
um
yeah but
do you so uh
to me honestly the most difficult part
is the thing that you said with the
marvex paradox is to create a human form
that interacts and perceives the world
but if we just look at a voice
like the movie her or just like an alexa
type voice
i tend to think we're not that far away
well for some for some people maybe not
but i i
you know i i
you know as humans as we think about the
future we always try to
and this is the premise of most science
fiction movies you've got the world
justices today and you change one thing
right but that's not how and it's the
same with a self-driving car you change
one thing no you everything changes yes
everything
grows together so surprisingly i might
be surprising to you or might not i
think the best movie about this stuff
was bicentennial man
and what was happening there um it was
schmaltzy and you know but what was
happening there
as the robot was trying to become more
human
the humans were adopting the technology
of the robot and changing their bodies
yeah so there was a convergence
happening
in
so we will not be the same you know
we're already talking about uh
genetically modifying our babies you
know there's a
there's
more and more stuff happening around
that
we will we will want to modify ourselves
even more for all sorts of of things
we put
all sorts of technology in our bodies um
to improve it you know
i've got
i've got things in my ears so that i can
sort of hear you yeah
yeah so we're always modifying our
bodies so
so you know i think it's hard to imagine
exactly what it would be like in the
future
but on the turin test side
do you think uh so forget about love for
a second
let's talk about just uh like the elect
surprise actually i was invited to be uh
what is the interviewer for the alexa
prize or whatever um
that's in two days
their idea
is uh
success
looks like a person wanting to talk to
an ai system for a prolonged period of
time like 20 minutes
how far away are we
and why is it difficult to build an ai
system with which you'd want to have a
beer and talk for an hour or two hours
like
not for to check the weather or to check
music but just like to
uh to talk as friends yeah well you know
we saw we saw um weisenbaum uh back in
the 60s with his programmer eliza yeah
um
being shocked at how much people would
talk to eliza and i i remember you know
in the 70s typing you know stuff to
eliza see what it would come back with
um
you know i think right now
and
this is a thing that um
uh
amazon's been trying to improve with the
like so there is no continuity of of of
topic
there's not
you can't refer to what we talked about
yesterday
it's not the same as talking to a person
where there seems to be an ongoing
existence where it changes
we share moments together and they last
in our memory together yeah and there's
none of that and there's no
um
sort of intention of these systems that
they have any
goal in life even if it's to be happy
you know they don't they don't
even have a semblance of that now i'm
not saying this can't be done i'm just
saying i think this is why we don't feel
that way
about them well that's a that's a a i'm
sort of a minimal requirement if you
want the sort of
interaction you're talking about it's a
minimal requirement whether it's going
to be sufficient
i don't know
we haven't seen it yet we don't know
what it
feels like i tend to be i tend to think
it's uh it's not
as difficult as solving intelligence for
example
and i think it's achievable in the near
term
but on the touring test
why don't you think the turing test is a
good test of intelligence oh i i because
you know
again the turing if you read the paper
turing wasn't saying this is a good test
he was using as a rhetorical device to
argue
that if you can't tell the difference
between a computer and a person
you must say that the computer's
thinking because
you can't tell the difference you know
when it's thinking you can't you can't
say something different
um what it has become as this sort of
weird game of fooling people um
so
back at the
uh ai lab in the late 80s we had this
thing that still goes on called the ai
olympics and one of the events we had
one year
was um
the original imitation game as turing
talked about because he starts by saying
can you tell whether it's a man or a
woman
so we did that at the at the lab we had
you know you'd go and type and
the thing would
come back and you had to tell whether it
was a man or a woman um
and um
the
the uh
one of the one of the one of the
uh
one man
came up with a question that he could
ask
which was
always a dead giveaway over whether the
other person was really a man or a woman
you know what he would ask them did you
have um
green plastic toy soldiers as a kid yeah
what do you do with them and a woman a
woman trying to be a man would say i
lined them up we had wars we had battles
and the man just bringing them out and
said i stomped on them
so you know
that's what that's what the turing test
the turing test with
computers has become what's the trick
question that's that's that's why that's
right it's sort of that's right devolved
into this
weirdness
nevertheless conversation not formulated
as a test is a pretty
it's a fascinatingly challenging dance
uh that's a really hard problem to me
conversation when nan poses a test
is a
is a more intuitive illustration how far
away we are from solving intelligence
than like computer vision it's hard
computer vision is harder for me to pull
apart
but with language with conversation you
could see because language is so human
we can so we can still clearly
uh
see it
you mentioned something i was gonna
go on off on okay um
i mean i have to ask you because you
you were the
head of csail ai left for a long time
you're i don't know uh to me when i came
to mit you're like one of the greats at
mit so what was that time like what
and and plus you uh
you're
i don't know friends with but you knew
minsky and all the all the folks they're
all the legendary ai uh people of which
you're one
so what was that time like what what are
memories that um stand out to you from
that time
from your time at mit from the ai lab
from the dreams that they are lab
represented to the actual like
revolutionary work well let me tell you
first a disappointment in myself
you know as i've been researching this
book um and so many of the players you
know were active in the 50s and 60s i
knew many of them when they were older
and i didn't ask them all the questions
now i wish
i had asked
i'd sit with them at our thursday
lunches which we had a faculty lunch and
and i didn't ask him so many questions
that now i wish i had he asked you that
question because because you wrote that
you wrote that you were fortunate to
know and rub shoulders with many of the
greats
those who founded ai robotics and
computer science and the world wide web
and you wrote that your big regret
nowadays is that often i have questions
for those who have passed on yeah and i
didn't think to ask them any of these
questions right
even as i saw them and said hello to
them on a daily basis so
maybe also another question i want to
ask
if you could talk to them today what
question would you ask what questions
would you ask well rick lyder i i would
ask him you know he had the vision for
humans and and
computers working together and he really
founded that at darpa and he
gave the money to mit which started
project mac in 1963.
and i i would have talked about what
what the successes were what the
failures were what he saw as progress
et cetera
i would have asked him more more
questions about that because now i could
use it in my book
but i think it's lost it's lost forever
a lot of the motivations are lost
um
uh
i i should have asked marvin why he he
and seymour pappet
came down so hard on neural networks in
1968 in their book perceptrons because
marvin's phd thesis was on neural
networks yeah how do you make sense of
that that destroyed the field for he
probably do you think he knew that the
effect that book would have
all the theorems and negative theorems
yeah
um yeah
so
yeah that's just the way of
that's the way of life
yeah but still it's kind of tragic that
he was both the proponent and the
destroyer of neural networks yeah
um is there other memory standouts
from the
the robotics and the ai work at mit
well
yeah but you gotta be more specific
well i mean like it's such a magical
place i mean to me it's a little bit
also heartbreaking
that
you know with google and facebook like
deepmind and so on so much of the talent
you know doesn't stay necessarily for
prolonged periods of time in these
in these universities oh yeah i mean
some of the companies are more guilty
than others are paying
fabulous salaries to some of the highest
you know producers and then just
you never hear from them again they're
not allowed to give public talks they're
sort of locked away
and it's sort of like collecting
collecting you know hollywood stars or
something and i'm not allowed to make
movies anymore i own them um yeah that's
tragic because i mean the there's an
openness to the university setting where
you do research to uh both in the base
of ideas and space like publication all
those kinds of things yeah you know and
you know there's the publication and all
that and often you know although these
places say they publish
yeah there's pressure
and um
but i think for instance um
you know on net net i think
google buying those 809 robotics company
was bad for the field because it locked
those people away
they didn't have to make the company
succeed anymore locked them away for
years and then
sort of all threaded away yeah
so um
do you have hope for mit
for him for mit yeah why shouldn't i
well i could be harsh and say that
i'm not sure i would say mit is leading
the world in ai
or even stanford
or berkeley i would say
i would say um deepmind google ai
facebook ai say i would take a slightly
different approach
or a different answer
i'll leave i'll come back to facebook in
a minute but i think those other places
are
following a dream of
one of the founders uh and i'm not sure
that it's well founded
the dream
and i'm not sure that it's going to have
the impact that
he believes it is
um
you're talking about facebook and google
and so on i'm talking about google
google but the thing is those research
labs aren't
there's the big dream
and i'm i'm usually a fan of no matter
what the dream is a big dream is a
unifier because what happens is you have
a lot of bright minds
working together
uh on a dream what results is a lot of
like adjacent ideas i mean yeah so much
progress is made yeah i'm i'm so i'm not
saying they're actually leading i'm not
i'm not saying that the universities are
leading yeah but i don't think those
companies are leading in general because
they're
you know and we saw this
incredible spike in you know
attendees at europe's and
as i said in my january first review
this year for
2020 will not be remembered as a
watershed year for machine learning or
ai
you know there was nothing
surprising happened anyway unlike when
deep learning hit
imagenet
that was a that was a
shake
and there's a lot more people writing
papers but the papers are
fundamentally boring yeah and
uninteresting incremental work
is there a particular memories you have
with minsky or somebody else at mit that
stand out
my funny stories
i mean unfortunately he's another one
that's passed away
you've known some of the biggest minds
in ai
yeah and you know they they did amazing
things and some sometimes they were
grumpy um
[Laughter]
well he was uh he was interesting
because he was very grumpy but that
that was
uh
i remember him saying in an interview
that the key to success
or being to keep being productive is to
hate everything you've ever done in the
past
maybe maybe that explains the perceptron
book
and there it was he told you exactly
but he meaning like just like
i mean maybe that's the way to not treat
yourself too seriously just uh always be
moving forward
uh that was this idea i mean that
crankiness i mean doesn't
yeah
so let me let me let me tell you you
know what really
um
you know the joy memories are about
having access to technology before
anyone else has seen it so so you know i
i got to stanford in 1977
and we had um
you know we had terminals that could
show live video on them um
digital digital sound system we had um
as
xerox graphics printer we could print
um
uh it wasn't you know it wasn't like a
typewriter ball hitting in with
characters it could print arbitrary
things only in you know one bit you know
black or white but arbitrary pictures
this was science fiction sort of stuff
um at mit the uh
the list machines which
you know they were the first personal
computers and you know they cost a
hundred thousand dollars each and i
could you know i got there early enough
in the day i got one for the day
couldn't couldn't stand up let's keep
working
um having that like direct glimpse into
the future yeah and you know i've had
email every day since 1977 um
and uh you know
the the host field was only eight bits
you know that many places but we i could
send email to other people at a few
places so
that was that was pretty exciting to be
in that
world so different from what the rest of
the world knew um
and uh
let me ask you probably edit this out
but just in case you have a story uh
i'm hanging out with don knuth
uh for a while tomorrow
did you ever get a chance to such a
different world than yours
he's a very kind of theoretical computer
science the puzzle of
of uh computer science and mathematics
and you're so much about the magic of
robotics like the practice of it did you
mention him earlier for like
not you know about computation did your
worlds cross they did enough you know i
i know him now we talked you know but
let me tell you my my donald canoe story
okay
so um
you know besides you know analysis of
algorithms he's well known for
writing tech yeah which is in latex
which is the academic publishing system
so he did that at the ai lab
and he would do it he would work
overnight at the air lab
and one
one day the
one night the uh
the mainframe computer went down
and um
a guy named robert paul was there he
only did his phd at the media lab at mit
and he was a you know
engineer uh and so
i and he and i you know tracked down
what were the problem was it was one of
those big refrigerator size or washing
machine size disk drives unveiled and
that's what brought the whole system
down so we got panels pulled off and
we're pulling you know circuit cards out
and donald knuth who's a really tall guy
he walks in and he's looking down and
says when will it be fixed you know does
he want to get back to write his tax
system
we figured out you know it was a
particular chip 7400 series chip which
was socketed we popped
it out we put a replacement in put it
back in smoke comes out because we put
it in backwards because we're so nervous
that donald knuth was standing over us
anyway we eventually got it fixed and
got the mainframe running again
so that was your little when was that
again that well that must be before
october 79 because we moved out of that
building then so sometimes probably 78
sometime or early 79.
yeah those all those figures is just
fascinating all the people with past
pass through mit is really fascinating
is there uh
let me
ask you to put on your big wise
man hat
is there advice that you can give to
young people today whether in high
school or college who are thinking about
their career
or thinking about life
how how to live uh
a life they're proud of a successful
life
yeah so so many people ask me for advice
and have asked when i give i talk to a
lot of people all the time
and there is no one way
um
you know there's a lot of pressure
to
produce papers
um that would be acceptable and be
published
uh maybe i was maybe i come from an age
where i would i could be a rebel against
that and and still succeed maybe it's
harder today
but i think it's important not to get
too caught up
with
what everyone else is doing
and
if you if a lot depends on what you want
of life if you want to
have real impact
you have to be ready to fail
a lot of times so you have to make a lot
of unsafe decisions
and the only way to make that work is to
make keep doing it for a long time and
then one of them will be work out and so
that that that will make something
successful or not
or you may or you just may you know end
up you know not having a you know having
a lousy career i mean it's certainly
possible taking the risk is the thing
yeah so but it it
but there's no way to to
make all safe decisions and actually
really contribute
do you um
think about your death by your mortality
i gotta say when covert hit
i did because we did you know in the
early days we didn't know how bad it was
going to be and i that that made me work
on my book harder for a while but then
i'd started this company and now i'm
doing full-time more than full-time in
the company so the books on hold but i
do want to finish this book um when you
think about it are you afraid of it
i'm afraid of dribbling
you know
i'm losing it the details of okay yeah
yeah but the fact that the ride ends
i've known that for a long time um so
it's
yeah but there's knowing and knowing
it's such a
yeah and it really sucks it feels it
feels a lot closer
so my in in my my blog with my
predictions my sort of pushback against
that was
i said
i'm going to review these every year for
32 years and that puts me into my
mid-90s so you know it's my whole every
every time you write the blog post
you're getting closer and closer to your
own prediction that's that's true of
your death yeah
what do you hope your legacy is
you're one of the greatest roboticist ai
researchers of all time
um
what i hope is that i actually
finished writing this book
and that
there was a there's one person
who reads it
and sees something
about changing the way they're thinking
and that leads to
the next
big
and then there'll be
on a podcast a hundred years from now
saying i once read that book
and that changed everything
uh what do you think is the meaning of
life
this whole thing the existence the the
the all the hurried things we do on this
planet
what do you think is the meaning of it
all yeah well you know i think we're all
really bad at it
life or finding meaning or both yeah we
get caught up in in in the it's easier
to get easier to do the stuff that's
immediate and not through the stuff it's
not immediate
um so the big picture we're batting yeah
yeah do you have a sense of what that
big picture is like why you ever look up
to the stars and ask why the hell are we
here
you know my my
my my atheism tells me it's just random
but you know i want to understand the
way random in the and that's what i talk
about in this book how order comes from
disorder yeah um
[Music]
but it kind of sprung up like most of
the whole thing is random but this
little
pocket of complexity they would call
earth
that like why the hell does that happen
and and what we don't know
is how common that
pop those pockets of complexity are or
how often
um because they may not last
forever
which is uh more exciting slash sad to
you if we're alone or if there's
infinite number of oh i i i think i
think it's impossible for me to believe
that we're alone um
that would just be too horrible too
cruel
it could be like the sad thing it could
be like a graveyard of intelligent
civilizations oh everywhere yeah that
might be the most likely outcome
and for us too yeah exactly yeah and all
of this will be forgotten yeah
including all the robots you build
everything forgotten
well on average
everyone has been forgotten in history
yeah right
yeah most people are not remembered
beyond a generation or two
um
i mean yeah well not just on average
basically
very close to 100 percent of people
who've ever lived or forgotten yeah i
mean
no longer i don't know anyone alive who
remembers my great grandparents because
we didn't meet them
so
still this fun this
this life is pretty fun somehow
yeah
even the
immense absurdity and uh at times
meaninglessness of it all it's pretty
fun and one of the for me one of the
most fun things is
robots and i've looked up to your work
i've looked up to you for a long time
that's right
ron it it's it's an honor that uh
you would spend your valuable time with
me today talking it was an amazing
conversation thank you so much for being
here well thanks for thanks for talking
with me i've enjoyed it
thanks for listening to this
conversation with rodney brooks to
support this podcast please check out
our sponsors in the description
and now let me leave you with the three
laws of robotics from isaac asimov
one a robot may not injure a human being
or through inaction allow a human being
to come to harm
two a robot must obey the orders given
to it by human beings except when such
orders would conflict with the first law
and three a robot must protect its own
existence
as long as such protection does not
conflict
with the first or the second laws
thank you for listening i hope to see
you next time