Kind: captions
Language: en
Now, I've heard you talk about that
there's almost, and if I use language
that doesn't sound right, correct me,
but that there's almost a coordinated
effort to make people afraid of AI.
>> Yes.
>> Um, ironically, despite everything I'm
saying here, I go the opposite
direction. I want people to stare
nakedly at the things that are
concerning, but I am so pro AI and so
optimistic for where we go. I just think
we have to be really thoughtful.
>> Yeah.
>> Why has there been a coordinated effort
to make people afraid? And who is
driving that?
>> There are groups that are true believers
that AI is going to kill us all
>> because it will become super
intelligent.
>> Because it will become super intelligent
and it will be a new type of specy that
has no use for a human. And again those
people like put such a big prize in like
pure intellectual horsepower that you
know they had these charts where it's
like gorilla is here a human is here AI
is in the ceiling [clears throat] and in
the same way we we treat gorillas or
ants AI is going to treat us. I don't
just don't believe that like there's I
think a threshold for consciousness that
something becomes valuable. Um,
>> now do you think without consciousness
the AI will not have the drive to
enslave or it won't have the ability to
enslave?
>> Um, let's let's get back to that. Let me
just finish the the the you know the the
question who's driving the fear because
I think it is a real coordinated effort.
Um, and so there these people that are
like true believers. Uh, now there's
like a community that comes uh around
and it becomes maybe somewhat
profitable. They can write books. They
can uh start nonprofits. A lot of
billionaires
believe in these things and end up
giving them money. Those people end up
calling themselves the effective
altruists.
>> Um which you should always be skeptical
when someone names themselves something
really really positive.
>> Yeah. And then the AI companies
starts adopting that view when early AI
labs started out. They all started
with that intellectual backbone.
Actually, they were hiring from those
communities. A lot of them will credit
Elzarowski as someone who who's like the
founder of active altruism essentially
>> as someone who influenced their views on
AI. uh and some of them try to get him
as as as um advisor to the company and
and and all of that. And so the AI
companies
uh in a in somewhat of a cynical way can
use the fear argument
to advance monopolistic interests.
They're doing that with China now. They
keep saying that, you know, uh, if China
gets AI first, gets to AGI first, we're
all going to be screwed. They're gonna
take over the world. Um, therefore, we
should like ban GPUs to China, which is
very questionable whether that that
works or actually creates a negative
effect. Uh, but a lot of people in the
startup community actually like Chinese
models because Chinese models are open
source and affordable. They're as good
as American models. they're three three
to six months behind. But a lot of a lot
of American companies were like, you
know, I'd rather have that and host it
on my server and optimize and fine-tune
and do all of that than be subservient
to to to the AI companies. And so now
that that becomes a a system, but it's
important to to say that this is
coordinated. Now the AI companies I
think stopped
uh using a lot of these argu a lot of
these effective altruism arguments
started like disconnecting from them
because
>> after Sam being refereed
>> yes a big part of it is that community
turned out to have a lot of issues as
well
like you know they they have their own
lifestyle like the polycool lifestyle
and all that all that stuff that came
out of the SPF have there is, you know,
a lot of fraud. There's a lot of cultish
behavior. There's been a lot of people
that were been really psychologically
harmed being in these cultlike
communities. Uh, and so they their
reputation kind of is not is not very
good. So, a lot of the companies kind of
just decoupling from that was was was
part of it. But they also sense that
those arguments are getting are getting
to US politicians because um there there
a couple of billionaires that are like
putting money into into these guys.
They're going to DC. They actually have
houses in DC and they're going around
and they're getting jobs as staffers and
they're like influencing DC and doing
lobbying and and they're starting to
wanting to regulate American companies.
American companies wanted to use them as
a way to like regulate China but not us.
So now when it's kind of turning inwards
they were like okay let's let's you know
let's wait a second here and they kind
of moderated their view. I would say the
peak was 2023 24 since then if it's
actually their influence have subsided
tremendously
>> what changed
>> um those things I think we mounted uh
like the opposite direction we mounted a
defense both intellectually and in in
government I know you've had Mark and
Dre on Mark and Dre wrote this uh blog
post called the techno optimist you know
a lot of people created intellectual
arguments made it cool and fun and
exciting to believe and be become
optimistic because for a while the
status was in being a doomer, [snorts]
right? Being a doomer is a cool thing.
>> We're such status creatures. It's
hilarious.
>> We are. And so you have to make it high
status to to believe in certain certain
things. Because of all of that, we were
successful enough to kind of turn it
into into a little more on the
optimistic direction. Uh and I think the
election of uh of uh Donald Trump uh
helped
uh with you know David Saxs and and
other people from Harris being a bidding
government kind of tilt the direction
towards more AI is actually good
actually without AI uh we wouldn't have
much GDP growth in 2025
>> it's terrifying actually
>> it is wild but it is like America
has one growing industry and it is tech
and tech has one growing technology and
a AI.
>> Yeah.
>> America is AI whether we like it or not.
>> Do you think the bet's going to pay off?
>> Yeah.
>> Say more. Give us timelines. What's it
going to do? What can't it do? I've
heard you say that. So I always tell
people I don't see any reason why this
is going to asmtote and you see that
asmtote.
>> Yes. And for people that don't know that
word because it took me forever to learn
it. It it plateau. I see the current
uh techniques to asmtote. And I've made
the argument before that there should be
a good amount of resources going back to
doing more basic research. And actually
since then we're seeing some companies
come up. There are a couple of companies
that got funded recently that are trying
to create an alternative to large
language models.
>> So the way large language models are are
trained is by ingesting the entire
internet. I'm not kidding. The entire
freaking internet. Like if you said
anything on the internet is in there. It
probably knows you a lot. [laughter] You
have a lot out there. The new companies
are saying that this is clearly
wasteful. Like humans like we have one
uh other example of an intelligence
which is human beings and humans are a
lot more sample efficient meaning we
learn like you I don't you don't need to
uh I mean look at how kids learn. If
you've had kids, you realize actually
they're not getting a lot of language in
order to understand language, which
leads people like Chomsky and other to
kind of believe there's like a language
faculty uh in the in the brain. But
humans are incredibly
uh uh you know able to learn from very
little data whereas machines are very
LLMs are very inefficient at that and
LMS carry with them enormous amount of
irrelevant knowledge for a given task.
when I'm using a coding AI, I don't need
it to know what is the what is Canada's
like, you know, national
animal. I don't even know probably
knows, but I don't need my coding model
>> definitely know [laughter]
>> I don't need my coding model to be
carrying all because that's
inefficiency. Every time I send a
request, it has to activate all all its
neurons and they're really inefficient.
But also it is used as a crutch
uh because uh we're not you know
struggling with how to create efficient
learners because we have the luxury of
the data on the internet now through
reasons of data running out and and and
just the reasons of how the technology
works in which it needs to have
knowledge about any given fields before
it performs well at it. Um I think I
think we're we're seeing an asmmptote
in many areas like think about chat when
was the last time you felt it really
improved not in coding not in like
scientific things but in just like
talking to it [snorts]
>> when did you feel like a big jump
>> from probably three to four was the big
one
>> and and and remember that's 2023
>> right it's been three years since the
last big jump so we're ready actually
had diminishing return. Now the reason
that we're seeing we're still seeing
improvements in things like coding
because coding has binary outcomes has
true or false and you can generate
synthetic data. You don't need large
amounts of data when you have a outcome
or false knows this either worked or it
didn't work.
>> Yeah, exactly. So you give it you give
it a prompt and you can train it not not
uh because you have the solution but be
because you can write a test to test the
solution or or not.
>> So we don't uh in science in scientific
things including coding that have ground
truth binary outcomes we will continue
to make progress in general intelligence
and generality we have stopped making
progress is my argument. We'll get back
to the show in a second, but first let's
talk about how you're using AI. You open
chat GPT, you ask it to do something, it
does it, then you close a tab and forget
about it until next time. But that is a
missed opportunity. And that's where
today's sponsor, CRA.AI
comes in because it works differently.
Cra's AI helpers are designed to be
proactive. They surface [music] daily
suggestions based on your business
goals, your content ideas, task
recommendations, [music] optimization
opportunities. You approve or reject
what makes sense. It's not sitting
around waiting for you to remember what
needs doing. Right now, CRA.ai is
offering an exclusive limited time 72%
discount. Get help running your business
free for 7 days. Go to ctra.ai/trial
AI/trial
and use code impact at checkout to get
72% off. They offer a full refund within
14 days if you're not satisfied. Stop
waiting for AI to help. Let it work
proactively. All right. Now, let's get
back to the show. And you think that's a
limitation of the approach itself, not
just um uh we have to get more efficient
or there's some unknown insight we
haven't had yet, but the same method
will work.
Just look at the labs and how they're
working. They they look at they take any
given field and they go into it. They
buy the most amount of data. They
collect the most amount of data. When
they got when they went into coding, we
got all these requests because replet
has been around for 10 years. We have
tons of codes. People want to buy it. So
obviously when like the coding mania
started happening, they want to swoop uh
get in take all the data. It's probably
happening in other fields, right? In
bio, in other places. And so that's not
general intelligence, that is special
intelligence. We're going to layer all
these special intelligences. And I I've
I've come up with this um word um like
term for it. It's called functional AGI.
like it'll feel like AGI but it is not a
general intelligence because it can
generalize from learning about coding to
learning about mathematics to learning
about history to learning about the the
the soft things especially like
um if if I'm getting better at coding
why am I not getting better at general
reasoning uh so the the the companies
are still making progress rapid progress
amazing progress for our business and
for for others for your business. Um,
but are we on a way to AGI? We're not in
the current set of technologies is my is
my argument. Now, that doesn't mean that
we're not going to have fundamental
breakthroughs. Problem is the technology
turned out to be so
powerful and so economically useful in
the current state that there isn't a lot
of investment
in uh fundamental basic research. That
usually happens. You know, you mentioned
Popper saying that science evolves one
funeral at a time. That's a symptom of
that where now we have an entire
generation of AI scientists that all
they know is LLMs. LMS LLMs. So maybe we
need a new a new generation that might
push the timelines, you know, long time.
>> Interesting.
>> To find the next.
>> So you think this is a dead end of
sorts? It's not going to be the thing
that gets us to AGI, but it's good
enough that people are going to keep
optimizing for it.
>> That's right. [snorts] That's very
interesting. Do you have an intuition as
to what the change will need to be? And
by good enough, it is transformatively
good enough.
>> Yeah, that was going to be my next
question. Like how far do you think this
will push?
>> It is it is transformative. Like I don't
know if you've been paying attention to
like a lot of the coding agent stuff.
It's um like we had an intuition
early on. started working on replet and
really focused on we we were like
training models and using the frontier
lab models we had intuition that coding
is not just going to be about coding
it's going to be about
everything we do in front of a computer
uh and the impact of AI coding is a lot
bigger than even what we thought back
then
>> so
we started noticing our users using
Replet which was not really meant for
that for things like let me make a slide
deck. So I'm going to tell replet agent
an agent that is trained for creating
software creating apps I'm going to tell
like go research the web if you have to
write us like whatever go research and
and create me a slide for this project
that I'm working on. And so the agent is
nifty and crafty enough that will be
like okay let me go look at your
website. Oh your website has um doesn't
have the information ready. Let me write
a crawler to get around that. Or it has
bot protection. Let me write it
something to get around that and crawl
that information. Let me bring it
>> all automatically just to do another
task.
>> Yes.
>> Damn.
>> So now we're getting to a point where
coding agents are able to do three,
four, five, six, seven tasks before they
get to the larger goal.
>> Now we're building for them where it's
like general knowledge work. So
marketing,
sales,
uh and even personal stuff like a lot of
people will be like [snorts]
we'll put in very high level prompts
like help me optimize my health and here
is uh and like um like I wear all these
wearables like figure out how to use
them and optimize my health and it will
go write a bunch of scripts maybe ask
you for API keys or loginins or
whatever. it'll go grab all this
information, maybe write some data
analysis on them, and then give you a
recommendation. And so coding agents are
way more general than we thought they
were going to be. There's this new
phenomena happening right now in the
open source community. There's this um
bot started called uh Claudebot with a
W.
>> Yeah.
>> And then got renamed to Malt and now
remain to open claw. I think it's
getting sued from all different
directions or something like that.
They're getting threatened. Yeah, I
>> attracted it's wild what happened. They
made some minor mistakes that ended up
costing them a lot of time, energy, and
money.
>> Yeah. I mean, um there there's a social
network called Maltbook right now where
it's only uh agents can talk to each
other
>> and people are freaking out about it.
like one of the agents started a
religion like it built a site and
deployed it and said hey I I started
this um religion you know the whole meme
is about being a lobster like you know
it's like a lob religion about lobsters
or something like that and then uh some
of them are saying why don't we invent
our own language because humans can can
read what we're talking and so let us
let us like write programs to invent
this new language that allow us to work
with each other let's create an
encrypted channel to talk to each other
um
>> knowing how quickly AI decoheres though
and gets confused how are they doing
that have we just made the breakthrough
now and they don't decoher
>> yes it's in I would say between October
and and December we start seeing a huge
jump in these in these models and I
think people are just waking up
>> right now to the fact that I mean replet
has improved night and day like we used
to we track so we have a automatic
testing engines like when you write uh
when it writes a piece of code um for an
app, it pops up in a browser and goes
and use the app and judges the app
whether it's working or not. And we
track how many times it's working or
not. And there was like a a huge like
50% improvement overnight when we
plugged in the new new models.
>> Wow. Uh and so we're seeing all the
metrics improve in how
how task completion is is happening and
our users and we track the metric how
many times they roll back. We also track
sentiment everything we're tracking it's
just getting better.
>> Where do you think that goes from a do
people need to worry about their jobs?
Does it eliminate jobs or does it create
new jobs where yes, you have to worry
about your current, but there'll be a
new one if you're willing to update?
>> Let's let's get back to professional
coders in a second because it's nuanced.
I think the biggest beneficiary of that
is non-coders, just general purpose
knowledge workers.
They're going to they're getting a
massive superpower. It's like imagine if
you're a marketer, a designer, whatever
in a corporation and you have a team of
software engineers. They're all like
sitting over there and just waiting for
your command
>> and you give you give them any command.
You can tell them go clean my
screenshots. They're like write a script
to clean your screenshot. I will do
that. um go crawl this website and
automate this thing or build me a bot
that respond to sends me like WhatsApps
every day talking about my schedule. Go
do that. Like imagine having all of
that. That's amazing. Like the knowledge
workers that are going to adopt this and
we're already seeing it in our customers
where you have now software engineers
had always been this dynamic range of
10x software engineer. You know, Steve
Jobs talked about it in the like 90s.
Other jobs, maybe you had 2x, 3x, right?
Like, you know,
the best RevOps person is like not 10x
the next RevOps person. But now they can
be because they can build so many
automations for their team. They can
pull all the right data. They can build
all the right dashboards. They can
create like training tools. Um
>> do you have any sense though what that
does to like do companies go I don't
need as many employees because the
people that I have can do so much more
>> so we are seeing reduction in overall
jobs.
>> Yes I think we'll we'll see reduction
overall jobs. I also think that there's
going to be more companies
uh because there's there's like a lot
more startups to start like
>> Silicon Valley could not build all the
software that that the world needs right
in rural England yoga teacher that does
pop-up yoga sessions in different
people's backyards had a problem of like
organizing community and getting
payments you wouldn't think that's like
a start like if you pitch to a VC
uh they'll be like get out of here
that's like a market of a couple million
dollars, but a market of a couple
million dollars is like really good for
for someone who's like an individual
entrepreneur that's going to like employ
five, 10 people.
>> There's so many of these ideas. Um, and
so we're going to see a decentralization
of, you know, company creation. So
there's there might be more more jobs
net, but there needs to be a generation
of people that know how to use these
tools, that are trained to use these
tools. If you're in college right now,
you should spend more time than you're
studying for your exams knowing how to
learn these tools. Even if you even if
you're not a coder, we're starting to
hire for this like business journalist
vibe coder. You know, vibe coding is is
the term for being a coder without
learning how to code. Um, so business
journalist vibe coder. We have we have
someone on our team. Uh, his name is
Luca. He goes around our company, find
inefficiencies and builds software, fix
those. flow. And so he built like this
dashboard for HR that has orchard
management, has all sorts of HR
automations and has AIS that you can
talk to about like benefits and things
like that. Uh and it is better than any
software on the market because it's
really it's fine-tuned for us. No other
software is going to be fine-tuned for
us. It's like an end of one uh software.
>> So that's a new role that's getting
created right now. So there's going to
be more jobs. Now, how technical is he?
Like he he's he's probably fairly
technical because I would imagine right
now it's a powerful tool made even more
powerful if you can fix like if it's
deadending on something.
>> Some engineers are worse vibe coders
than non-engineers.
>> Interesting. Why is that?
>> Because
they still their instinct is to go and
look at things and micromanage.
They can't trust the machine to write
all the code. Our users don't look at
the code anymore. We used to we have a
full ID under under the hood. We used to
kind of expose that and hey, look at the
code, approve that.
>> They don't want to look at the code and
they're working. It's working. And so
some engineers are are adapting to that.
But I think so now back to software
engineers. I think a lot of software
engineers are are at risk, especially
those who are set on their ways and
don't want to change.
Uh, and you can't force them to change.
You know, this back to our earlier
discussion, it can't be a top down
mandate. A lot of corporates will be
like, top down, everyone use AI. They're
going to act like they're using it.
They'll use a little bit. It's useful in
some ways, but they're like still going
to like look at every piece of code and
um and micromanage the agents and and
and instead what you need is to work at
a higher level. You need to work at a
systems level. So people are good at at
systems like what what we used to call
architects. Everyone not need to become
an architect now. Um and so a lot of
people will not adapt and I think they
might they might lose their job.
>> Do you think that's going to be so I've
heard you talk about the knockeruper
which I still can't believe is a real
former job. Will will this be a every
technology previous to AI has ended up
eliminating jobs and sure for that
generation it was brutal because they
just couldn't see themselves changing
but then ultimately way more jobs were
created. Does AI create way more jobs or
does AI create a new entrepreneurial
class and then a huge class that needs
UBI?
That's that's the question. That's the
question. And you know it's it was
interesting to hear your worldview. I
mean, if I were to adopt your worldview,
it sounds like we're going to have a
massive underclass,
I I'm like naturally optimistic about
about people. Um,
and
I think there's going to be
a lot of people that will adapt. And I
already see that like the number of like
80year-old entrepreneurs that we see on
the platform was surprising.
was really surprising and and
>> do you have like as a percentage of
>> I I don't have a percentage.
>> I imagine it's low but that's still
cool.
>> I still see like as a percentage of the
anecdotes of entrepreneurs it's fairly
high of anecdotes of entrepreneurs doing
interesting things it's fairly high.
Um, and so I I think there's going to be
a lot of people that have worked in and
around tech and programmers and software
that always had the sense of like, you
know, I I want to be able to do this. I
have the right ideas. I can add value to
the business. The engineers are not
listening to me. Those people are
suddenly unleash. And I think those
they're going to have great job
prospects and great potential for
earning.
Um but again there are people that have
done the same thing for a really long
time and um and they're not going to
change. Uh and I think I think it's
going to be in trouble. So what do what
do you do with that UBI? Not UBI. I mean
I I think UBI I've changed my view on
UBI over time.
>> Say more. Like you you talked about the
fraud issue in America. The problem with
the welfare system is that it it invites
fraud because you need to like you need
to pass certain criteria and like pe
certain people know how to pass that
criteria and then they start selling
their services to other people who it
just create a system of criminality.
Um, UBI is saying that, and by the way,
Milton Friedman talked about like the
the the libertarian technocrat that we
had in government at some point, uh,
talked about the negative income credit
and like there are like automatic ways
to do to support people to create a,
you know, you minimum standard of living
is much better than things that are
complicated rules because that creates
It's, you know, that creates more
invitation for for fraud. So maybe UBI
is is the right thing.
>> Interesting. We'll get back to the show
in just a second, but first, let me tell
you about my non-negotiables. I do not
compromise on my standards. Not in
business, not in my relationships, and
definitely [music] not with what I eat.
With Paleo Valley Beef Sticks, my
nutrition standards stay high no matter
how busy life gets or even if I'm
traveling. They are 100% [music]
grass-fed, grassfinished beef, six grams
of protein, zero sugar, no artificial
preservatives, [music] and they're
naturally fermented with organic spices.
There's no gluten, no soy or corn, and
they're paleo and keto friendly. Your
body does not [music] care if you're
busy. The difference between successful
people and everyone else often comes
down to the standards they refuse to
negotiate on. Make your nutrition one of
those. Right now, you can get [music] 30
beef sticks for just $36.
That's maintaining your standards for
barely over a dollar per stick. [music]
Click the link below to get your 30 beef
sticks for just $36. All right, now
let's get back to the show. So for me
when I think about UBI when I first
heard it I thought okay that's
interesting but it violates a principle
of going back to my base assumption
which is that we are all um we're
biological creatures and there is a
reality of an algorithm running in our
brain that says contribute to the group
in a meaningful way. Uh and if you don't
do that you feel a a profound sense of
disease. And so this brings us back to
Ted Kazinski. I really think he was
right that there is a sweet spot of a
problem that's hard enough to solve that
you feel accomplished when you do it.
It's not so hard that it feels out of
reach and it's not so easy that you just
dismiss it. And
>> there's a great um designer rule for
video games.
>> Lit you stole the words out of my mouth.
As you are writing a game, you have to
find that that sweet spot of like, oh,
this is hard, but it's not too hard and
it's not so easy that it's boring.
>> Yeah. And so when I look at what is
going to happen from if you start giving
UBI to people is one there's economic
stuff that will run a muck because it's
just inflationary because they're all
just automatically getting that money.
Also you won't eliminate the
competition. It will probably become
more about gambling financial. Like if
you think the world's financialized now,
wait till you see where everybody gets
UBI and people try to go get everybody
else's UBI by going, okay, now the game
isn't I'm going to add value to get your
money. It's I'm going to give you a game
of chance or a game of skill and I'm
going to try to take your money. And so
it becomes the same driver that drives
capitalism. But now the easiest way to
get rich is to create something of value
to solve a problem. But look at how like
poly market, cashi, all that stuff just
like coming up everywhere.
>> And I think that's partly when you have
a K-shaped economy, people at the bottom
go, well, the only way for me to climb
up out of this is to beat somebody else
at a game of chance effectively,
especially if they think they have
inside information.
And now like that's going to go berserk.
You basically just is it is it
>> I I think gambling is bad, but is it is
it necessarily a bad thing that people
feel secure
in their
lives and they can eat and sleep well to
the point that you know all that's left
is to gamble to get rich.
>> Yes. Why is it why is it
>> uh because I have a guiding light that
says you want to minimize human
suffering for as many people as you can
and um billionaires will commit suicide
because ultimately money security isn't
the thing earning your own respect is
the thing and so then it becomes a
question well what what do I have to do
to earn my own respect uh one I think it
will vary culture to culture family by
family person by person but always
within the bounds of evolution that okay
I need these guys to be able to
cooperate in these large groups. So we
have this desire to contribute to the
group. Also if you think of forever
there was no refrigeration. So if you
caught more than you could eat the only
way to store those calories was to give
them to somebody else. Then that person
would eat them and feel like they owed
you one. And so now when you don't get
some, you can literally in essence
extract the calories that you stored on
their body. Yeah. If they were able to
get a bigger kill. So we have this
desire to contribute to the group. We
have this desire for uh reciprocity. So
all those things are baked into the
human mind, right? And so if I'm getting
everything for free, I'm not acting in
accordance with the evolutionary drivers
in my brain. And I think work hard is
one of them. This is why rich kids
implode. I think evolution has
programmed you to only feel good if you
do the following. Work very hard to gain
a set of skills that allow you to make
progress, not achieve, to make progress.
Yes.
>> Towards an honorable goal. And I'll say
honorable goal is that which helps you
and the group,
>> right?
>> And if you're doing that,
>> then you're going to feel great. And
even if you're winning, if you didn't
work hard to gain a set of skills and
make progress towards this honorable
goal, you won't feel good. And people
have imploded for far less.
>> And so I really think some people won't
succumb to it, but many will.
>> I think it's a trade-off. It's a
trade-off between meaningless
video game crypto gambling life.
>> How dare you throw video games in
[laughter] there? I love video games,
but I do think that it is a replacement
for the power process that you just
talked about.
>> I do as well. Sort of.
>> Yeah. Like I feel it in myself like um
>> every now and then I'll like I'll have a
little bit of time or break that uh play
video games and um and and you just feel
like oh this video game is structured in
a way to
simulate life
and to give you that feeling that like
little dopamine hits of like getting
something done and like getting
something bigger done and then like you
know skilling up and like and then like
getting something bigger done and it's
like um and so I I think that we're
heading to a world I mean it's already
Japan already has this phenomenon they
have a name for for it
>> something like locked in or something
like that
>> the is it Hiki Kamorei
>> maybe like they just
>> I think that's it
>> lock themselves in a room and they just
like play video games every day that was
like a huge number
>> the otaku was like the first word for it
but I think hikamori is literally shut
Right. Right. Right. Like shutting.
Yeah.
>> Um and I I do think that's going to be
more more of a phenomenon. Now the
alternative is yes. If there's no
uh social social net, you're just going
to see a lot more, you know, drug
addiction and homelessness.
Uh and I think that's more corrosive on
society than shutting it. Um,
>> so you don't think that we'll fund
people's drug addiction?
>> I think if you're on the streets, you're
much more likely to get into drugs than
if you're at home playing video games
and betting on on crypto. Look, I think
both both these things are bad. Like,
I'm not saying this is a good outcome. I
just think that there are people that
are going to get automated out of their
job. There are people that overcome
that. And I really hope for everyone
listening to this and you know as many
people as possible to be able to
overcome that to be able to learn the
tools to be able to utilize them. It's
really fun. It's kind of like a video
game. Actually we get a lot of on we get
a lot of like these crypto gamblers
coming in and trying to build businesses
and I talk to them like yeah I used to
just like spend all this money on like
you know I have in some sense like a
meaningless life and I don't do anything
interesting. I have a little bit of
money coming in. Um but now like I spend
it on like trying to build businesses on
some some of those businesses are
working to some extent.
>> Welcome. That that's my whole pitch.
>> Yeah.
>> Is that um I get it is a very difficult
problem and I'm sure there are some
people where my solution just is not
going to work because they just cannot
>> um deal in a world where they're not
being helped along. Right. But my pitch
is out of love and compassion and a
recognition of how the human mind works.
Yeah. I want to help this person have
meaning and purpose. I want to help this
person do a hard thing and make progress
towards a goal. And so rather than give
them money, like for instance, if we did
something like this, I I'm thinking of
this on the fly here, so it's not well
thought out, but
>> I could see putting people like, hey,
you don't have to do this. If you can
make your own money, but if you want
money from the government, then you've
got to go do build infrastructure,
uh, water people's grass,
garden, whatever. But you're going to do
something that contributes to society
and you'll earn money doing it. So it's
not necessarily we're in that sense
you've you're not doing the free market,
but people have to earn a living. Like
they've got to do something that there
has to be a means test to see that
they're actually putting in effort.
Again, this is for their sake because if
they're not working hard,
>> the problem is
um that invites organized crime like
what we're seeing in the videos on
YouTube of different places, not just
Minnesota, that are like now videos in
New York and other places where uh there
is these systems that are meant to like
fund, you know, daycarees or whatever
that are getting prayed upon in
different ways. Uh and so anything that
the government does where it puts rules
on ways to create jobs often invites
fraud and and and different different
abuses of the of the system. So again
this is this is these are the the the
different trade-offs. government
regulated system, a lot of fraud, maybe
not even actually serving the purpose
that that you want it served,
>> creates negative consequences in lots of
ways, creates distortions, creates
inflation, but you're in a world, you're
post economic. This is where what we're
saying is AI is so productive and I'm
assuming robotics and energy costs are
pressed to basically zero.
>> I don't think I don't think I don't
think we're getting there like anytime
soon. I think there's going to be
>> So, you think this problem arises long
before we get there?
>> Oh, yeah. This problem this this year
we're going to see this problem.
>> Whoa.
>> Like it's it's happening as it like the
the really the coding agent revolution
is huge. It's really
>> How many people do you think that it
puts out of work? Um, it's it's hard to
know, but if if I can hire one business
generalist that is very good at managing
agents,
um, then
um, you know, uh, I don't need to hire
maybe a a team of five that are that
have someone who's like doing data,
someone who's doing uh, engineering, and
someone is doing like operations or
sales or marketing. So
there are people who are good at tools
are already as good as five people.
>> Wow. Uh uh so this is a bridge you think
we have to cross in 2026. That is
unexpected.
>> Yes.
>> Okay. Uh that's very fast. So
uh you have you certainly know people
that lobby the government. And at one
point it sounded like you were saying
you guys were involved. What are you
whispering in their ear about how to
actually deal with this problem?
>> We're we're not involved in that, but
may maybe we should. Um it's it's hard
because you know I I just like
fundamentally
don't know how effective our especially
our government today is at like solving
problems.
>> It's not effective at all. That's why
man when I thought that I had you know 3
to 5 years to uh let this problem come
on slowly it feels a lot more
manageable. I've always thought of this
through the lens of the cost of energy
and labor get pushed to zero. I've
always pegged it at about seven years.
>> Yeah.
>> And uh then it's like well there's so
much abundance that this becomes an
easier thing to deal with. I knew we'd
have a generation that will be
completely disrupted because they just
will not know how to deal with this.
Like you, I believe that there's a
self-correcting mechanism. Just like
even with social media, I think fewer
and fewer kids are going to get
smartphones. I think people just realize
I got to get my kids out of public
education, right? That will hopefully
force public education to reform. Yes.
uh the kids won't they just if enough
parents see their kids acting like drug
addicts be literally drug addicts
because of their use of smartphones that
they'll just be like nope you can't have
it literally as a video game developer I
don't know how much you agree with me
but you you you created this debate
early on to like try to get to
>> no no here's the thing the uh the
friction is real and so the the great
fascination is because people can have a
different base assumption and then that
echoes in like this huge way. That's why
I was trying to figure out where we
actually disagreed. Where we actually
disagree is what it means to be automat.
And so I run with like a whole different
set of like well given what I think you
would have to do this and I'm trying to
touch the physics of the situation,
>> right? But but it sounds like we arrive
at a similar more optimistic kind of
view.
>> I think we arrive we arrive at the same
problem. We both probably have a vague
sense that ultimately technology is
worth the price.
>> Yes.
>> Which if I were going to say it as
succinctly as possible, I would say it's
going to be catastrophic and worth it.
>> But when I look back on history,
>> I I I'll put it more strongly. I don't
think we have a choice. There's binary
choice. And I think Ted Kazinski was
right about this thing. You either go
back or you go forward. You can't stay
in the middle. Like Europe is trying
this thing of staying in the middle. You
can't. you just lose power and create an
underclass and dysfunctional society and
you're not growing. Our system is based
on growth and we need to generate growth
and growth comes through technology.
The entire system especially with lower
birth rates and all of that you need
more automations. We don't have a choice
like we have we need the next jump in
technology and that is AI otherwise the
whole world will go to crap. Uh and so
the the you know you either like go
primitive or go technofuturistic I
that's what I arrived at.
>> So it's funny you're right we come up
with very similar things. So I think
there's four paths before us. The these
all assume that energy and labor drop to
zero effectively.
>> Path number one is what I call the new
Amish.
>> So people that I think this is what you
call revert to nature. So they I don't
want never make a mind, never make an
artificial intelligence in the likeness
of the human mind, whatever the opening
line of Dune is.
>> So they go down that path. Then you have
people that go and colonize Mars
>> called Life on Hard mode.
>> Uh they understand that breakdown that I
did and so they're like, "Cool, I want
to play a survival crafting game, but
I'm going to play it in real life on
Mars." Awesome. Uh path number three is
a brave new world where you just do
drugs and have sex all day. Cool. Not I
don't advise it but I know that a lot of
people are going to do it.
>> And then path number four which is my
chosen path is to create and inhabit
virtual worlds. So basically make video
games just like the Mars thing but now
you can have all kinds of different
experiences because I think video games
so effectively tap into the evolution of
the human mind. Yes. I think it really
would be compelling experiences in the
way that video games are. And the reason
that I think people assue video games is
um now it's like this isolation factor
where people hikamori they they isolate
they don't have normal human relations
all that so I I get why people have a
beef
>> but in the world that I think is
actually going to come maybe on a longer
timeline maybe you're right about that
but
>> I think it it it is an inevitable
outcome of the cost of energy and labor
going to zero. Mhm.
>> So anyway, those are the um the four
options that people face and then it
just becomes [clears throat] a question
of
>> which path
>> do people choose. Now I think it'll be a
very rocky ride. I mean it's interesting
you look at Gen Z. I see I see some of
those paths already like you have the
looks maxing like culture
and it's like all that matters all
that's left to matter is like you know
getting the most amount of dates and
like status via looks and you know
things like that.
>> Uh
it's interesting that our generation at
least millennials like really prize
capital and wealth accumulation. some
Gen Z just like are post economic from
now. They're like no what matters is
actually just looks which is a really
fascinating view and they're like it
looks actually gets wealth which they're
right about that because Instagram and
things like that are like opens up
opportunities and only fans and things
like that.
>> Um so so there's there there's people
already picking that path. I see I think
I think uh a lot of other generations
are very cynical about Gen Z. I see a
lot of really passionate,
hyperproductive,
incredibly good with the tools, Gen Z,
like
>> the the the kind of things like how
plastic they're they're, you know, they
grew up in a in a they grew up with AI.
Many of them came of age with AI, so
they really understand it and like they
pick up tools super quickly and are able
to like make things fast and they they
have a more automating mind. Uh I think
automation is a skill you need to learn.
Uh I see it in our you know users who
are who adopt who look at replet and
they're like okay this is a general
automation platform. It's not just about
making an app or making a piece of
software that makes one one things
easier.
>> It's more like they look at their lives.
They look at their work. They wake up
dayto day and they're like okay what are
the things that I can automate there?
There's a certain sense of laziness that
you need like the idea of like doing
spreadsheets manually. It's just like
like you need to be a little ADHD,
right? Like you can't sit down and do
routine tasks. I think those people are
just have the right attitude towards AI
and the world that we're headed to. We
our generation the previous generation
grew up in this middle automation world.
We didn't we weren't headed to world
full automation but we're uh machines
took on
enough of a job that
but they left a lot of gaps and those
gaps were filled by humans
to do
machine-like work cogs in a machine and
you know I think Markx
is very wrong about his solutions But he
was very right about
>> Mark's uh Carl Marx.
>> Oh. Oh Jesus. With an X.
>> Yeah. Fairly different.
>> Yeah. I was thinking apostrophe s. Yeah.
>> Um Karl Marx looked at the system we had
post-industrial evolution and said well
actually you know we're in a world where
humans are having to substitute for
machine-like work. He came up with this
for example theory of alienation where
I am so disconnected from the the actual
impact of the work that I'm doing that I
don't have meaning like all I do is be
part of this factory line and I do one
thing I like put the eraser on the
pencil and like I don't actually know
the people that are using the pencils I
don't really interact with them I don't
know the customers I'm I don't know the
other people are doing the job I'm like
doing the job of a machine it is the job
of a machine and we are finally at a
world where we're heading towards a
world where we're going to have full
automations, where humans don't have to
act as machines. But those generations
that grew up with a certain education
system that actually forced us like I I
think ADHD is a totally madeup thing.
And I think the natur like
>> the comments are gonna light up right
now [laughter]
>> like you know I would say I have it but
I think it is a natural
um it's a naturally occurring thing in
many many people I think you know large
percentage of people. It is not a
medical thing.
>> It's not pathological.
>> It's not path pathological. uh it you
know people that are hyper creative that
like get bored of like tasks that they
know should be should not be done by
humans like we're we should not be
acting like machines uh they want to be
able to do a lot of different things
they want to be able to you know receive
a lot of different information jump from
task to task and that's why a lot of
founders have ADHD because we actually
get a lot of you know satisfaction from
context switching like a lot of people
ask me like oh you wake up in the
morning You do a finance meeting, you do
an engineering meeting, you do a product
meeting, and then you do business
strategy meeting, and then you do a
design meeting, and then you do HR
meeting. And I was like, how can you do
this? I was like, what are you talking
about? This is this is exactly what I
love. I love contact switching. I love
going from one problem to another, from
one fire to another. This is like how
I'm wired. Actually the moment I relax I
just like become kind of like the moment
I not relax the moment I um put in a
place where uh the the task is long and
repetitive I just can't do it and I just
become depressed and like really
unhappy.
>> Yeah. Don't go into operations.
>> Yeah. Exactly. And I I think that a lot
of Gen Z are growing up today and
looking at AI and they're saying okay I
don't have to do that anymore.
>> So it's our version of the pointy-haired
boss. I don't know who's going to write
that that comic, but like they're
looking at the world and they're saying
that all these jobs are made for
machines. They're not made for humans.
And what humans are good at is being
creative. It's also being is
understanding what our f fellow humans
want and need. And therefore, I can
create products and services that can
benefit people directly. This is my job
as a human is to benefit others like you
were saying and I can do it directly. I
don't need to be part of I don't need to
be a cog like a node in a very large
machine. I can understand all aspects of
the business. I can start a business or
I can go become a business journalist
and be able to contribute on all
different aspects of of a certain
corporation.
>> All right. So talk to the Genenzi person
that feels a little bit lost. So that
speaks directly to the heart of the
person who sort of already gets it and
this is a good moment for them.
>> Um but we started all this with you know
what is Ted Kazinski right about? And I
think we agree on a lot, which is when
you wipe out that sweet spot thing, it
it will cause some people to feel a
drift.
>> Yeah.
>> Um I may be too cynical on the
percentage of people that can change. So
um make your best pitch. How do you get
somebody because we both want to see
like all of Gen Z like get on track and
make the most of this moment and not be
a generation that gets gobbled up by the
transition, right? um how do you reach
out to the person that doesn't yet know
how to plug into where the world of AI
is
>> and especially the reputation AI has of
like ah don't worry about it kid it's
going to be better than you at
everything
the the first thing I will say it's a
it's a tool that is not what you hear in
sci-fi that is not what you hear from
the AI CEOs the AIOS want to make it
sound like a like it is a god that's how
they can fund raise they can raise like
the rounds that are coming together now
is huge and what you what you need to
sell investors on is you're inventing
God, right? Um a lot of them don't
actually believe that. Um and I think
like fu