Kind: captions
Language: en
The XAI goes 50 million in units of H00
equivalent AI compute, but much better
power efficiency online within five
years.
>> If it takes whatever, five minutes to
install, 5 minutes times 50 million.
>> That's a lot. Like, how many hours is
that? This is wild, man. Anybody else
made that claim and I wouldn't buy it
for a second, but the fact that Elon is
saying it, he really might get there.
That's crazy. Five minutes times 50
million is 475 years. Jesus.
>> So they're gonna have to scale it.
They're probably going to assemble them
all fly and just bring the whole box.
>> But still, somebody has to do it. I'm
just saying like you're going to have to
throw so many people at this problem or
if he thinks within 5 years you'd be
able to have robots assemble them. 50
million is a lot. It does not matter
what object you are talking about. 50
million grains of sand is a lot. Just to
benchmark it, OpenAI is saying they will
have over 1 million GPUs by the end of
2025. So that's a 50x multiple.
>> Well, okay, but already a million in
2025. That's huge. So the superclusters
that Gro 4 is rocking that allowed them
to leaprog everybody else is like
250,000 by 320,000 or something like
that. So it's like that we're going to
have uh a million by the end of this
year. Like, damn. 300% year-over-year
growth with no signs of letting up. We
just we're not drinking in what that
really means.
>> Tennessee is starting to raise their
hand about XAI's consumption and their
like gas output because they have
temporary turbines running right now and
there's only supposed to be like 10 of
them, but he's been running like 35 of
them. So, there are environmental
concerns. Is there something we should
be concerned about? Because 50 million
GPUs, that's a nuclear power plant.
>> There there is a thing you should be
deathly afraid of. China winning the AI
race. You can't stop. Game theoretics
tell you you cannot stop. Whatever
you're worried about, sorry, this is the
Manhattan project. You can't play
around. If the Nazis might develop
nuclear, you have to beat them to it.
You have to beat them to it. Did I
mention you have to beat them to it? You
are in that kind of race with AI. AI has
the potential of being a winner take all
technology. It probably won't be winner
take all as long as you're roughly
paralleling each other. But boy oh boy,
you can't stand still and you cannot
slow yourself down. You just can't.
>> And it is, this is the first model. I
remember you were telling me this that
it's actually showing signs of like
understanding like the physical world
and
>> and like reasoning and stuff like that.
So now when you get to 50 million times
that output, that's AGI, right? Like
that's what we're talking about. If you
get that big of a cluster and you can't
achieve AGI, then Yan Lun is right and
LLM just won't yield that result. The
distance that AI has already traveled is
so crazy. And when you think about,
okay, remember, you're not only going to
make these units bigger, you're going to
make them more efficient. And so, so
much of AI is about the algorithm, and
we're so early. Oh my god, when I think
about the video compression algorithms
that we use now versus the ones that we
used to use and what a big difference it
makes. Oh my god, like it's you're going
to get the same thing with AI, like far
less GPUs to get the same output, but
you actually have these gigantic GPUs.
So now you're just able to do that much
more. And that's where this really gets
extraordinary. If people have not
experimented with OpenAI's agents, do
they are incredible.
>> Is there a a limit though? Because there
is something to the Manhattan project of
it all.
>> Is there a limit to what we ought to do
or
>> Yes. What what we ought to do. It's
>> there are of course limits to what you
ought to do. You shouldn't be
programming something that you know will
enslave humanity. Blah blah blah. You
shouldn't be um reckless about security.
You should definitely be taking steps to
make sure that these things are aligned.
Should be going way out of your way to
protect yourself. But what you shouldn't
be doing is putting a regulatory burden
on this that makes it impossible to
build. You shouldn't be saying, "Oh,
we're worried about emissions." It's
like, hold on. you've got a way bigger
fish to fry right now. Do the cleanest
energy you can, but don't stall this
stuff out. What people always lose sight
of is there are regulations that stop it
and then there are regulations that give
a clear framework that people can work
within. So, for instance, if we don't
want to see all the gas and coal being
used, great. Then make it possible for
them to spin up nuclear energy, hydro
energy, whatever else we're going to let
them use. But don't just shackle them
and say like you can't use that at all.
You've got to put them on a path where
they can leverage that now, but be
incentivized to move towards something
cleaner.
>> AI will cement classes of wealth. I'm
really like focused.
>> AI will cement classes of wealth.
>> Cement classes of wealth.
>> No, it won't. We'll do the exact
opposite.
>> Rich people will have access to it and
then it falls away. What's the counter
argument to that? The reality is ei
techchnology ever ever starts with the
rich have access to it because they can
afford it and then it gets cheaper and
cheaper and cheaper and cheaper and
cheaper because humans are selfish. They
are selfish and so they want to be like
I want to be the one to bring this to
the masses. I want to make this thing
cheaper so that more people use mine. I
don't want that guy to have it. I want
them to use my thing. It always happens.
Okay. First, the railroads made a very
small number of people extremely wealthy
and then it completely turned America
into a superpower because we had this
vast space with all these different
resources and we could ship it around
really fast. People need to again look
at history. These patterns repeat all
the time. You can always trust humans to
be selfish and the most selfish thing
people can do is make technology more
available to more people. They want to
hoard the money possibly. They want to
hoard the assets. Again, that's not how
the system works, but people try to do
it. So they're always trying to get
richer and richer and richer and make
more money. But what do they end up
actually doing? They make life better
for everybody because the only way to
get money is to make it available to
everybody.
>> I feel like Daniel Presley said
something about that where he was like,
"This is going to far benefit the rich
first and then once they get cemented,
then they're going to disseminate the
information down."
>> Oh god. People should have to run a
company, Drew. People should have to run
a company and they are suddenly going to
go, "Oh, wait. You've got to spend
dollars sequentially. You got to invest
in one thing. It has to yield a return
otherwise you can't invest in this other
thing." People just literally they
literally think that building appear out
of nowhere. Companies kick out products
just automatically. And I think because
so few people understand what they do on
a daily basis, how it actually impacts
the company that they're like, they just
think work is a thing you have to do
because you have to make money, but it
doesn't actually do anything in the real
world. Instead of understanding, uh, we
make all this wealth, we make all these
material things that make our lives
better because people have an innovative
idea and they work really hard to create
an engine that outputs something that's
more valuable than its inputs. And that
is very difficult. There are two really
hard things that any great entrepreneur
has to do. You have to attract and
retain talent. And you have to allocate
capital wisely. And if you can't do
those, and most people just do not
understand money. They don't understand
capital allocation. They don't
understand how you have to sequence
things. They don't understand the
importance of profits. They just don't
understand it.
>> Ooh, this super chat is a banger. Let's
go.
>> In the age of AI, Kissinger warns the
future won't make things cheaper. It'll
split society in two. Those who wield AI
and those ruled by it. The cost isn't
dollars, it's your agency. That reminds
me a lot of Einstein's dreams.
>> Yeah, they have their finger on a very
important thing. So, if you know who
Emad Most is, this is why he is just
theish about making sure that AI does
not get trapped behind pay walls. This
is why Elon Musk originally wanted Open
AI to actually be open and available to
everybody. This is why Elon Musk, every
time one of their new models comes out,
they release the previous one and make
it open and available. People are going
to make AI open and available. You do
not have to panic. It won't be the
bleeding edge one. That just is what it
is because they do need to make money to
convince all of you to work. That's the
hard truth. None of us will work for
free. We're the problem. Wrap your head
around that.
>> Even money off the table, it's
affordable to everybody. It's
achievable, obtainable for everybody.
There are people who still in 2025 of
July reject AI. I was talking to a buddy
of mine. He was talking about how he has
like this idea. It's a super good
YouTube idea. I don't want to ruin it,
but it's so good about just how like
kitchens aren't made how you think
they're made. He's like, "Yeah, but that
takes so much data and I just have to
aggreate all this information." And I
was like, "Yeah, why don't you just use
AI? You can kind of like separate like
you can organize all your thoughts so
that way you can kind of take it chunk
by chunk." He's like, "Nah, I'm a
tactile person. I much rather have a pen
and pencil." and like he took pride in
that and it's almost like yeah I don't
want to I don't want to use AI. I'm
gonna do it the oldfashioned way. Yeah.
And it's kind of like the people that
are like I don't want a car. I want to
ride a horse. And it's like yeah that's
cool but then you're going to go not go
twice as far. And I'm worried that with
the rapid adoption of technology those
people are going to be left behind. And
it's not even a rich poor thing. It's a
you think you don't need it versus this
is actually a tool that you're not
utilizing the right way.
>> I you can't help people that don't want
to be helped. And I think that there are
going to be people that are for
religious reasons, they will completely
reject AI. They will completely reject
technology. The Amish already do it.
There's going to be more people. Yep.
And I'm sure a lot of them live
wonderful lives. They just will have no
political influence whatever. They will
have to basically do like the Amish do
and you create a community and you can
cycle within your own community. It is
what it is. So, people should have the
right to choose that life if they want
and some will and some of them will
maybe be even happier. I think there are
some crazy stats in the Amish and
they're actually pretty happy. When
you're choosing that life by choice,
that's awesome. People should be able to
choose the life they want.
>> Should I still continue college if
everything will change so much?
>> I wouldn't. Not if you have to take on
debt. If I had kids right now, I would
say under no circumstances should you go
to college if you're going to take on
debt. Unless the job that you're going
to get on the other side of that is
going to be like 3 to 5x more lucrative
in the first 5 years so that you can
begin paying it off like easily so that
you're not stressing out about it.
>> AI productivity is going to be
deflationary which will make anyone in
debt in one uh chat and then the
trillions of debt owed um by the US will
[ __ ] us hard. Don't say ah
okay PSA there are two t types of
deflation one is terrifying and one is
awesome the terrifying one is crisisled
deflation crisisled deflation is all bad
all the time innovationled deflation is
all around you and is
amazing by the way it's happening right
now. It's just that the government
consumes it all rather than letting you
take advantage of it. The reason people
are worried about deflation is they it
does do what you're talking about, which
is whatever debt you have, that debt is
going to effectively slowly go up in
value. It is going to require people to
change their strategy. Trust me when I
say uh unfortunately, this is not
something that's going to happen
overnight. And remember, if you've got
economic growth that outpaces that, then
you can go in and you can pay off that
debt, which is exactly what certainly
the government will be incentivized to
do, and individuals would be wise to
follow the same. But as the economy is
growing, as long as a meaningful
percentage of that is getting to the
people actually working in the system,
you're going to be fine. And people will
be wise to pay off their debt because
they realize they're in a deflationary
environment. But hey, no problem because
the economy is growing so much from
everything that's happening. Because
remember, it's not just that the cost of
the things is going down. It's that
there's more things coming onto the
market. You have you have just literally
an embarrassment of riches. There's so
many amazing things out there. This is
why they're getting cheaper over time.
You want to be in that position. It will
change the economy. It will force people
to change their behavior. But it is not
going to be a crisis. It is not people
hoarding their money because they're in
fear. Everything feels great. It feels
like the roaring 20s. Velocity of money
comes from the way that people feel.
They feel good. Remember, this is all a
confidence game. In a world where we
have AIEL deflation because things are
basically becoming free, energy costs
are falling out of the sky. Labor costs
are falling out of the sky. We can get
virtually anything we want for free.
Your problem becomes not deflation. Your
problem becomes meaning and purpose. And
by the way, if everybody's getting
everything for free, is anybody going to
worry that some like even if we had to
say, "All right, listen guys. Deflation
is causing a problem with the debt, so
we're just going to wipe out the debt."
And everybody's like, "Well, I live like
a king. I can literally get a car built
3D printed or, you know, put together by
a robot in an afternoon for $9.95."
Like, it it's just a totally different
game. So you don't want to imagine like
all the bad things that will come from
how amazing this is but not realize it
will be offset by the fact it's so
[ __ ] amazing that prices are falling.
Okay, there is a difference between
crisisled deflation and innovationled
deflation.
>> When has innovation ever led to actual
deflation? Price go down is not
deflation. Price of TV go down is not
deflation. When prices go down, they
sell you on the next thing or the
upgrade. Planned obsolescence is real.
That is true for the people that are
buying the latest and greatest. But for
the people who can now get a TV for
$200, like that's unbelievable when a TV
used to cost I mean
>> $2,000.
>> Yeah. The equivalent of right in
inflationadjusted dollars. When you
compare that people now have cheaper
options that are so much cheaper than
the cheapest option when that technology
first came on. We're so surrounded by
deflation that we don't even think about
it. So sure, the person that can afford
the latest and greatest will buy the
latest and greatest. But the person that
can't now has access to something that
kings would have killed for a long time
ago. This is the hyonic adaptation.
People do not realize how people that
are poor live better than somebody who
was wealthy 4,000 years ago. They don't
even have access to refrigeration. Like,
bro, you can be homeless and live on the
street and get access to somebody giving
you a couple bucks and you go buy
something out of a refrigerator in a
grocery store. So, you may not own the
refrigerator, but you still have access
to refrigerated food. The frame of
reference that they have is distorted
compared to what's actually there.
>> Poor people in America live better than
the three million people below that make
less than a dollar a day in the world.
So, it's like we're all on a scale to
ground us.
>> Ain't that the truth, Drew? Yep. Ain't
that the truth?