Limitless Podcast

Josh:
It's been a big couple of weeks for elon we had a few pretty hit

Josh:
episodes last week talking about starlink talking about the ai5 chip

Josh:
and this week it's just another big breakthrough ejaz this week we're coming

Josh:
out with a lot of new grok and xai news which is pretty exciting i mean one

Josh:
of the leading headlines he said i now think xai has a chance of reaching agi

Josh:
with grok 5 never thought that before and now there's two things that kind of

Josh:
spawn this one which we'll get into a little bit later, which is the Grok Fast model.

Josh:
It is remarkable. It is a full order of magnitude better than anything else for its size.

Josh:
And it is really, really impressive. But the thing we're going to start with,

Josh:
Ejaz, is this chart that we're showing on screen right here,

Josh:
which is the single thing that convinced Elon, wait a second,

Josh:
maybe, maybe just maybe, Grok 5 could actually lead to AGI.

Josh:
And it's because we're seeing this crazy anomaly on the chart where Grok 4 was

Josh:
kind of ahead, but somehow without any new major release, Grok 4 is now way ahead.

Josh:
So Ejaz, can you explain to us like what's going on in this chart?

Josh:
How did they get so good so fast without a major new model release?

Josh:
I mean, this didn't even come from XAI, did it?

Ejaaz:
It's a good question. And no, it didn't come directly from XAI.

Ejaaz:
It actually came from two random AI researchers, one called Jeremy Berman and

Ejaaz:
the other one called Eric Pang, who tweaked Grok 4's model,

Ejaaz:
also known as fine tuning, to basically make it a hell of a lot smarter.

Ejaaz:
And so they put it to the ultimate test, Josh.

Ejaaz:
It's this thing called the Arc AGI benchmark.

Ejaaz:
And for those of you who have not been spending all your time researching benchmarks,

Ejaaz:
the Arc AGI benchmark tests how good your AI model is at being successful.

Ejaaz:
Intelligently human. What I mean by that is it presents the AI model with puzzles

Ejaaz:
that it's never seen before, that it can't possibly have been trained to solve

Ejaaz:
and sees how good it does. Now, Josh, let me ask you this question.

Ejaaz:
Before Grok 4 itself was released, what do you think the highest score was on this benchmark?

Josh:
Lower, but I'm not sure how much lower. I don't know this particular numbers,

Josh:
but maybe I'll guess five to 10% lower than what the best is now,

Josh:
kind of like an incremental improvement.

Ejaaz:
Nope nope nope it was way way

Ejaaz:
lower in fact it only scored between five to eight

Ejaaz:
percent from the top models from open ai google and all those kinds of things

Ejaaz:
and then grok 4 came yeah grok 4 came along and it broke that frontier and scored

Ejaaz:
22 percent guess how much these two models that these two random ai researchers um scored on wait

Josh:
So you're I'm looking at the screen. I'm seeing 79.6%. Is that right?

Josh:
Is this a 4X multiple on base Grokfor?

Ejaaz:
80%. And this had nothing to do with the XAI team at all. I want you to focus

Ejaaz:
on this chart that I'm showing you right now.

Ejaaz:
And look at my cursor circling around these two orange dots that are off into the distance.

Ejaaz:
You see Grokfor thinking over here, which was basically the heaviest,

Ejaaz:
most expensive model that Elon and the XAI team released.

Ejaaz:
Um when they launched croc 4 and they were just completely beaten by these two

Ejaaz:
models but i'm sure you're probably thinking josh how the hell did these two

Ejaaz:
researchers do that and um you know why aren't they being hired by elon immediately they

Josh:
Don't have the resources of a giant lab.

Ejaaz:
Like they're

Josh:
Competing against i mean if you remember these people are getting billion dollar

Josh:
offers to come work for a single employer and there's a collection of these

Josh:
so how is it that one individual it's being a collection of these people.

Ejaaz:
So these two researchers introduced two novel

Ejaaz:
ways of training their models one is called

Ejaaz:
open source program synthesis and the

Ejaaz:
other is called test time adaptations before

Ejaaz:
i get into an explanation as to how these work i

Ejaaz:
want to remind the audience that what really makes a model really intelligent

Ejaaz:
um is largely part in due to the data that it's trained on people spend so much

Ejaaz:
money i'm talking hundreds of millions to billions of to acquire the best data to train their models.

Ejaaz:
And the reason why this is so important is the model, when it's trying to answer

Ejaaz:
a question, draws on the data that it's been trained on, right? So it's hoping that

Ejaaz:
it can look back on the data that it's been trained on and find the right answer

Ejaaz:
somewhere in all of these tokens and characters, right, Josh?

Ejaaz:
These researchers decided to flip that completely on its head.

Ejaaz:
It's this thing called open source program synthesis where the model designs

Ejaaz:
its own bespoke solutions in real time.

Ejaaz:
So it doesn't even look at the data that it was trained on.

Ejaaz:
It just looks at the puzzle that it's presented with and it tries to break it

Ejaaz:
down into smaller components. So let's say the puzzle has 10 different steps

Ejaaz:
to reach to the end goal, the correct answer.

Ejaaz:
It'll break it down into 10 different little steps, whereas normally a model

Ejaaz:
would just look at the complete set of 10 steps and think, hmm,

Ejaaz:
how do I get from step one to step 10?

Ejaaz:
It just solves each step one at a time.

Ejaaz:
And that was like the massive breakthrough that they made.

Ejaaz:
And if this sounds familiar, you're probably thinking of this technique known

Ejaaz:
as reinforcement learning,

Ejaaz:
which basically has like the model like repeatedly go at a problem over and

Ejaaz:
over again this is pretty similar but it's the next step up in that field

Josh:
Okay got it yeah this this news kind of really annoyed me because of how seemingly

Josh:
simple it was i mean jeremy burman in the case of this i got some examples of

Josh:
specifically how he did it and he he was originally writing in python code but

Josh:
then he switched to just writing instructions in plain english and i think this

Josh:
is such an important thing that a lot of people forget.

Josh:
I mean, myself included, I'm speaking for myself here, that a lot of this really

Josh:
challenging, difficult work with engaging with LLMs is really just done in plain English.

Josh:
You're just writing sentences to a model in hopes that it produces a better

Josh:
output for you. It's not this crazy complex code base, although that exists deep down.

Josh:
But the way that they achieve this is actually just by writing plain English.

Josh:
So I did a little bit of digging. I have a few notes on how it works.

Josh:
And his system, it basically starts by having Grok 4. He chose Grok

Josh:
for is this model of choice it produces 30 english descriptions

Josh:
of rules to transform inputs into outputs so it

Josh:
takes that and then it tests these descriptions on training examples by

Josh:
pretending each is a test and scoring how well

Josh:
they match the correct outputs and then the top five descriptions get

Josh:
revised individually with feedback on mistakes like highlighting the

Josh:
wrong cells and stuff like that and then it combines the elements into

Josh:
the top one to create these pool descriptions so it kind of has this iterative loop

Josh:
where it tests itself it creates more examples it

Josh:
gets better data it confirms that it's the right output and that's

Josh:
generally why you see the actual outputs of this model are

Josh:
a little more expensive but the quality of it is amazing because it

Josh:
just continues to do this like self-iterative loop on itself and

Josh:
get better and better and better again all in plain english so if you are listening

Josh:
to this podcast in english you are fully capable of doing this because you speak

Josh:
the language and this isn't anything crazy it's just very refined um prompts

Josh:
that you're feeding to a model that result in these unbelievable outputs that

Josh:
are now best in the world.

Josh:
That's the coolest part to me, EJS. I don't know about EJS.

Ejaaz:
No, no, I agree. And it reminds me of Andrew Carpathy's hit tweet three months

Ejaaz:
ago where he goes, the new number one programming language

Ejaaz:
Turned out to be English. It's English. Right?

Ejaaz:
And kind of like to emphasize, again, how important this is.

Ejaaz:
This isn't just another frontier breakthrough of another benchmark.

Ejaaz:
I'm talking about the hardest benchmark that has just been 3x'd by two random researchers.

Ejaaz:
Right? This is, again, puzzles that are problem sets that an AI model has never

Ejaaz:
seen before. Typically, when you put an AI model up against a benchmark,

Ejaaz:
it has some kind of context.

Ejaaz:
Kind of think of yourself taking an exam at school or at university.

Ejaaz:
You can look at past papers. You can look at books. You kind of know what topics

Ejaaz:
they're going to talk about.

Ejaaz:
This is completely foreign to an AI model. And therefore, it is the hardest test.

Ejaaz:
So to have something achieve this almost feels like, and Josh,

Ejaaz:
I hate to say it, but I have to say it, like AGI.

Ejaaz:
And I think the fact that none other than Elon himself was taken completely aback by this.

Ejaaz:
I mean, again, to reiterate the tweet, I now think XAI has a chance of achieving

Ejaaz:
AGI with Grok5, never thought that before.

Ejaaz:
And the fact that he is now saying, hey, by the way, Grok5 starts to train in

Ejaaz:
a few weeks and you know what?

Ejaaz:
I think it's going to be out by the end of this year.

Ejaaz:
I think just speaks to the importance of this development.

Josh:
Yeah, I think one of the things that was really startling for me was the realization

Josh:
of how little resources it takes to get so good. And then I was wondering, well, why?

Josh:
Clearly, this isn't anything super novel, although they did do some unique training frameworks.

Josh:
And I think the reason that I, the conclusion that I came to was just scale.

Josh:
I mean, the cost per query, the cost per token of these new super high end models

Josh:
that just came out is very high.

Josh:
And you can't really scale that to a lot of people because the companies are

Josh:
just resource constrained.

Josh:
So it leads me to believe and leads me to think, well, what happens when a company

Josh:
with a lot of resources dedicates all of their brainpower to this specific type

Josh:
of reinforcement learning, like we're going to see with Grok5,

Josh:
and they do so in a way that's compressed enough, that's efficient enough to

Josh:
actually run it at scale on the servers without melting everything down without

Josh:
charging $1,000 a month per membership.

Josh:
And I think that's probably what we see with Grok5 is this new juiced up reinforcement

Josh:
learning, but efficient and actually built for scale.

Josh:
I mean, even if it just launches at the specs of these two individual researchers,

Josh:
that's a huge win because that's an incredible model.

Ejaaz:
Yeah. And it's open source. It's open source and available for everyone.

Josh:
It's pretty remarkable. Yeah. So I think very interesting things coming.

Josh:
If I was a betting man, I would be betting big on Grok5. I think they very much

Josh:
see a solution that people really want.

Ejaaz:
I was just thinking about why

Ejaaz:
both of us are finding this development both amazing, but really annoying.

Ejaaz:
And I think it's because to some degree

Ejaaz:
we both believe that in order for AI models today

Ejaaz:
to get to AGI we would need to completely re-architect

Ejaaz:
how they're designed you know transformers was

Ejaaz:
the big breakthrough that's why models that we know and use today

Ejaaz:
are so smart but it's not as

Ejaaz:
smart as we expected it and there was this kind of like lag of improvement and

Ejaaz:
now we suddenly see a 3x improvement where this model is kind of breaking this

Ejaaz:
leading benchmark and so now I I think I'm starting to believe that maybe if

Ejaaz:
we invest hundreds of billions of dollars in the post-training part,

Ejaaz:
where typically we've been investing in the pre-training, in the compute,

Ejaaz:
but if we invest it in the post-training, we may clearly reach AGI before redesigning

Ejaaz:
the entire thing up front.

Ejaaz:
Does that resonate with you, Joshua? Or am I just, do I sound crazy?

Josh:
No, it does. It does. It's funny because we frequently record the show and you

Josh:
expect to be surprised and then something happens.

Josh:
You're like, oh my, I wasn't expected to be surprised in that way. and this

Josh:
is one of those things where i mean i wasn't expecting to see a new leader in

Josh:
between major model releases from an independent researcher so the fact that

Josh:
this is even possible really just blows the doors off of a lot of expectations

Josh:
i had and this isn't even the only interesting news this week from the xai team

Josh:
because they released new model alert grok for,

Josh:
fast let me tell you each as when i saw how this model worked i was like whoa

Josh:
this is um again blown away super.

Ejaaz:
Impressed can you run us through some of the highlights please

Josh:
This spec sheet yeah so first of all the leading headline

Josh:
two million token context window is outrageous i think the current leader is

Josh:
google with the gemini model they have 2.5 pro and flash both of them i believe

Josh:
have a million tokens uh this is two million tokens of context for those that

Josh:
aren't aware context is the basically active memory of a language model it's the more context you can,

Josh:
collect, the more clarity it has into the actual data that it's talking about

Josh:
and conversing with, you want that number to be bigger.

Josh:
This is the biggest by far, by a doubling.

Josh:
So that's a really important headliner. The second one, probably even more outrageous,

Josh:
47 times cheaper than ROC4.

Josh:
Which is crazy because when you look at it on

Josh:
the scale below if you can scroll down just a little

Josh:
bit grok 4 is right in line with every other great

Josh:
model it is a grok 4 fast is just beneath

Josh:
o3 it's above deep seek it's above cloud 4 sonnet it's a above cloud 4 opus

Josh:
it's just like this remarkable model that is better than a lot of the leading

Josh:
models but 47 cheaper than the base model and i think that's going to be a pretty

Josh:
interesting thing when we get into like scaling these models and using them for code.

Josh:
EJ, as we talked last week about how good the Grok model is for coding because

Josh:
it was so cheap and so effective.

Josh:
This is another case of that. And the way they did that, I was so interested

Josh:
in how they were able to come up with like the secret sauce to do it.

Josh:
And basically what they did is they taught to model to spend its brainpower

Josh:
only on tools when it helps.

Josh:
So they use this like large scale reinforcement learning to train Grok for fast

Josh:
to choose when to think in depth and when to answer questions quickly.

Josh:
So what that resulted in was about 40% what we're seeing here on the screen.

Josh:
40% fewer thinking tokens on average than we've gotten from the previous model,

Josh:
which is a significant difference. Oh, and by the way, it's number one on Elmarina.

Josh:
So this was crazy. EJs, what were your reactions when you saw the team drop this?

Ejaaz:
I already thought these tokens were cheap. I thought these models were cheap enough.

Ejaaz:
Do you remember when OpenAir released GPT-5? They kept flexing GPT-5 mini saying,

Ejaaz:
hey, you now have the power of our previous best model, but actually it's more intelligent.

Ejaaz:
And it's like, I think it was something like five times cheaper.

Ejaaz:
And I was like, holy shit, holy crap. I was like, that is like crazy magnitude.

Ejaaz:
And now it's like, now we've got 47X cheaper than Grok 4.

Ejaaz:
Grok 4, by the way, was already cheap compared to some of the Frontier bottles.

Ejaaz:
So I don't know how far this can go, but kind of zooming out,

Ejaaz:
I have never been more confident that,

Ejaaz:
than now that cutting edge super intelligence will be available for anyone and everyone.

Ejaaz:
This isn't going to be some kind of closeted technology where only the rich

Ejaaz:
can buy devices and run it.

Ejaaz:
I think anyone and everyone will have fair game access to this.

Ejaaz:
And think about the dynamics that that changes up, Josh.

Ejaaz:
Like you can have someone in the complete middle of nowhere with a cell phone

Ejaaz:
attached to Elon Musk's new 5G Starlink satellite that's beaming down to him.

Ejaaz:
And he could kind of produce something that the world ends up using because

Ejaaz:
he has access to this cheap model that is actually super intelligent and can

Ejaaz:
be used to create whatever crazy invention that he has or she has that dreams

Ejaaz:
up. I just think this is insane.

Josh:
Yeah, the efficiency improvements are the thing that's always most exciting

Josh:
to me because, I mean, as we get more cheaper tokens and as the tokens become

Josh:
more portable and lightweight, I mean, you could have the world of knowledge

Josh:
on your phone even without necessarily an internet connection because these

Josh:
models are getting so lightweight, so condensed.

Josh:
Um so effective it's like it's really it's

Josh:
unbelievably impressive and what i was really interested in

Josh:
is comparing this to the other models because i know google was

Josh:
kind of doing a similar thing they were leading along the frontier and oh yeah

Josh:
here's this post from gavin baker that i love because it shows how google has

Josh:
kind of dominated this thing called the pareto frontier and on the chart you

Josh:
can very clearly see how there's this kind of arc that hugs the outer bounds

Josh:
of all of the models and it shows that like gemini pro has been really good on a few

Josh:
So I briefly want to just talk about the Pareto Frontier concept because it's

Josh:
really interesting and it will explain to you exactly why Grok4Fast is way out there.

Josh:
It totally shattered what it is. So, I mean, basically, it's funny.

Josh:
I was doing a little bit of research on this and the Pareto Frontier is done

Josh:
by an Italian economist named Velfredo Pareto.

Josh:
So I just thought that was a fun fact because great name. Basically,

Josh:
it comes from the economist and decision theory.

Josh:
And it's a way to decide optimal trade-offs when you have multiple objectives

Josh:
you're trying to achieve all at the same time. So imagine you're trying to optimize

Josh:
two things that might conflict a little bit, like you want to make a product

Josh:
as powerful as possible, but also inexpensive as possible, like these models.

Josh:
So in this scenario, there's a set of best possible solutions where you can't

Josh:
improve one aspect, like the power, without making the other aspect, like the cost worse.

Josh:
And what we're seeing in this chart here is Google has made a series of those

Josh:
decisions, those tradeoffs that

Josh:
have led to the absolute Pareto optimal outcome along this outer band.

Josh:
What grok has done is they actually made a new trade-off

Josh:
that isn't necessarily a trade-off it's more of an innovation that allows

Josh:
them to unlock this perceived frontier this limiting factor

Josh:
that was on the outer band and just shatter it and create a new pareto

Josh:
optimal trade-off using these best things and they did that by doing a lot of

Josh:
magic but basically what they have now is they have a really smart model that

Josh:
actually sits above gemini 2.5 flash and not too far below the pro model but

Josh:
it is a order of magnitude cheaper and i think that's where that outlier that

Josh:
cost effectiveness is really unbelievable when it comes to,

Josh:
um distributing these tokens widely so now if you're writing code if you're

Josh:
creating an application if you're just if you're paying for tokens this is very

Josh:
clearly the model you want to use.

Ejaaz:
What you just described is elon and

Ejaaz:
xai literally charting a new path which

Ejaaz:
is kind of like um very behavioral of elon in general um and another thing that

Ejaaz:
i thought was really cool about this is the reinforcement learning infrastructure

Ejaaz:
team was kind of key behind getting this model as fast and as cheap and as efficient

Ejaaz:
as we're describing it, right, Josh?

Ejaaz:
They used this kind of like agent framework, which was extremely compatible

Ejaaz:
with the infrastructure that they used to train and iterate on this model in the first place.

Ejaaz:
And what I wanted to point out here is there's a theme between the two topics

Ejaaz:
that we've discussed so far on this episode, Josh.

Ejaaz:
Number one, when we described the two models that the researchers created that

Ejaaz:
broke the ArcGIS benchmark, they specifically used a technique which used reinforcement

Ejaaz:
learning, a new reinforcement learning technique.

Ejaaz:
And the reason, if you remember, why Jeremy Berman picked Grok4 specifically

Ejaaz:
was he said it was the best reasoning model because in the way that had been

Ejaaz:
trained via reinforcement learning.

Ejaaz:
And now we're seeing yet again, this GrokFast model achieving what it can because

Ejaaz:
of reinforcement learning.

Ejaaz:
So I'm seeing a theme or noticing a theme here where XAI and Elon are basically

Ejaaz:
the leaders in reinforcement learning,

Ejaaz:
which i think is going to probably play in their favor maybe it's a hint that

Ejaaz:
the models that are going to be closest to agi that are the quickest that are

Ejaaz:
the cheapest are embedded in reinforcement learning techniques that are just

Ejaaz:
completely breakthrough

Josh:
Yeah it seems like they the team really reasons

Josh:
i mean this is a core elon uh notion but

Josh:
like they really do reason from first principles and what's important and what matters and

Josh:
you're seeing that throughout the entire product as they advance and

Josh:
i think what's really exciting what i'm most stoked about um

Josh:
for this show in general is is to compare this

Josh:
next round of models like will gemini 3

Josh:
and grok 5 like how are they

Josh:
going to compete with each other because those are both going to be remarkable

Josh:
models and it seems to me like those are like those are currently the top dogs

Josh:
i mean as far as gpt5 was kind of a little bit of a miss anthropic's been a

Josh:
little bit quiet uh gemini and xai are on fire but this also there was there

Josh:
was one last thing of news before we sign off today well i.

Ejaaz:
Was I was gonna say, like I'm highlighting this sentence here for those who are just listening.

Ejaaz:
And it says, you know, we built this reinforcement learning infrastructure team

Ejaaz:
with a new agent framework to help train Grok4 fast.

Ejaaz:
But specifically so that we can harness the power of Colossus 2.

Ejaaz:
And if I remember correctly, Josh, there was some breaking news around Colossus

Ejaaz:
2. Elon was getting into some fights. Can you walk us through it?

Josh:
Yeah, it's funny. There was this whole report from SemiAnalysis,

Josh:
which does a really great job. I would highly recommend checking them out.

Josh:
And they released this report on the XAI data center build out.

Josh:
And it was so funny to see, because a lot of times you just see satellite pictures

Josh:
or you read headlines and you're not really sure what's going on.

Josh:
The sole purpose of SemiAnalysis is to actually have boots on the ground,

Josh:
check the satellite images, and look at it with a scientific engineering point

Josh:
of view where they actually understand what is going on.

Josh:
And they shared their findings in one of these articles. And I found one of

Josh:
these stories was so funny because it's such a testament to how the XAI team

Josh:
works, where they were having problems with their energy generation in Memphis,

Josh:
Tennessee, because people were complaining and they were having a tough time getting permits.

Josh:
And the core crux of every large AI data center is energy.

Josh:
So they were like, this is unacceptable. well, we need energy immediately.

Josh:
So what do they do? Well, they jumped over the state lines.

Josh:
They went over to Mississippi a couple of miles down the road and they built

Josh:
these new generators right down the road across the state line.

Josh:
They got the permits they needed.

Josh:
They said, we don't, you don't want us, Tennessee. We'll just go right over to Memphis.

Josh:
You could see here, they took the power lines, they ran them back into Tennessee

Josh:
and now they're powering the data center.

Josh:
And part of the article was, was this funny story, but part of the article also is,

Josh:
is colossus 2 um being built

Josh:
in the sheer scale that colossus 2 is going to be and it's going

Josh:
to be over a gigawatt of energy um which

Josh:
is i don't know how many hundreds of thousands of homes is

Josh:
going to power but this is like a remarkable amount of power and a tremendous

Josh:
amount of gpus and they're planning to make these all coherent and they're using

Josh:
them exclusively i believe to train this new grok 5 model so as this new training

Josh:
center comes online they will be using this new cutting-edge world's largest

Josh:
supercomputer to train the world's perceivedly best model.

Josh:
But I found this funny because the day that this article came out,

Josh:
there was another post from another CEO of a very prominent company saying,

Josh:
hey, wait a second, we have something a little bit bigger.

Josh:
Than Colossus 1 currently. And that was from Microsoft's CEO, Satya Nadella.

Josh:
And he had this post where he said they just added over two gigawatts of new energy capacity.

Josh:
So Ejaz, this is just a really crazy brawl between these people who are building

Josh:
larger and larger AI data centers.

Josh:
And it eventually leads to the big news that just dropped a little earlier today.

Josh:
But do you have any commentary before we get to the huge number?

Ejaaz:
Yeah, there's actually one thing I wanted to point out, which is when Elon first

Ejaaz:
announced that he was building out this Colossus 2 data center,

Ejaaz:
it made headlines that it cost $20 billion.

Ejaaz:
And everyone thought it was crazy. Everyone was yelling, this is an AI CapEx bubble.

Ejaaz:
There is no products that prove that all this investment makes sense.

Ejaaz:
And now you have Satya Nadella, CEO of Microsoft, announcing that he's probably

Ejaaz:
going to be investing twice as much of that to build two gigawatts of new capacity.

Ejaaz:
Again, validating that there is a need for energy and compute to train these new models.

Ejaaz:
Don't forget that Microsoft last week acquired a random European data center for,

Ejaaz:
I think it was about, what, $10 billion, which caused its stock price to 3x

Ejaaz:
because it itself wasn't worth that much at the time of the reporting happening.

Ejaaz:
And then it leads us to the even bigger announcement, which released this morning,

Ejaaz:
which is NVIDIA will be investing not one,

Ejaaz:
not 10, not 20, but $100 billion in OpenAI over the next couple of years.

Ejaaz:
And you might be asking why?

Ejaaz:
Well, it's because OpenAI is going to be investing in so many data centers that

Ejaaz:
is going to produce so much power. I don't know how many gigawatts.

Ejaaz:
I think it's actually 10 gigawatts which is 10x colossus 2 5x um uh fair water

Ejaaz:
which is satya nadella's thing for all my mathematician mathematician fans out

Ejaaz:
there it is just crazy josh are we in a bubble or is there a need for all of this

Josh:
So here's the thing as i keep going back

Josh:
and forth about the bubble conversation because a hundred billion

Josh:
dollars is such an outrageous amount of money to spend on making

Josh:
what is already a remarkable language model even more

Josh:
remarkable um like the product is great and

Josh:
at least me personally as a user of these

Josh:
products i'm definitely getting closer to a wall of things that i use them for

Josh:
where if a model is marginally smarter my experience doesn't get that much better

Josh:
um but i was so that's like one school of thought and then the other is thinking

Josh:
well this is probably the only thing we'll ever need to spend money on going forward ever.

Josh:
So it makes sense to throw all of it at it now.

Josh:
Because in the case that you do solve AGI, you get hyperintelligence,

Josh:
it solves all of your problems. And it gives you the better questions to ask

Josh:
in order to solve better problems. So it really...

Josh:
It would appear, assuming that we continue on this trajectory of improvement,

Josh:
that it makes sense to take every disposable dollar you can to get better and better compute.

Josh:
And this will probably just extend forever. As we are able to harness more energy

Josh:
from the sun, from nuclear energy, a lot of that new energy and compute will

Josh:
just go to making better AI, which will then serve better downstream effects for how society works.

Josh:
So is it a bubble on the long term? I think absolutely not on the short term.

Josh:
I don't know. Where do you get the revenue from?

Josh:
I don't know. I mean, it's a ton of money, but you know what?

Ejaaz:
I think the reason why you and I feel this disassociation between the amount,

Ejaaz:
how large these numbers are in investing in infrastructure versus what we're

Ejaaz:
actually seeing is we're not going to be seeing AGI before some other fields

Ejaaz:
or some other professions see it first, right?

Ejaaz:
The clear example is coding. Coding has just

Ejaaz:
been on an absolutely exponential improvement rate

Ejaaz:
that has beaten out any other ai feature ever you now

Ejaaz:
have ai models that can code as well

Ejaaz:
as a senior staff engineer which is getting paid like

Ejaaz:
300 to 500 a year right um so my guess is this investment is worth it um my

Ejaaz:
guess is the investment is going to come to fruition in professions in use cases

Ejaaz:
in jobs that we won't see but we'll maybe talk about or see the kind of like effects.

Ejaaz:
Maybe it's in science where we create a new drug that cures cancer or whatever that might be, right?

Ejaaz:
I think different types of professionals will see AGI and reap the rewards of

Ejaaz:
this investment before average consumers see it.

Ejaaz:
And then I think the other thing that I want to mention, Josh,

Ejaaz:
is this isn't specific to US or Western spending.

Ejaaz:
In fact, our foes over the seas in China or in Asia have been working on this

Ejaaz:
for like the last five years.

Ejaaz:
They've been building out massive data centers, which I think has like build

Ejaaz:
up an aggregate of like 300 gigawatts over the next five years, at least.

Ejaaz:
And they've been investing in this so heavily. So it's not just a Western thing.

Ejaaz:
It's an Asian thing as well. China is investing so heavily in

Ejaaz:
this if this is a bubble if we are completely

Ejaaz:
wrong this will be the uh biggest most highest profile l that the world has

Ejaaz:
taken it's not just going to be a us thing it's not just going to be a map something

Ejaaz:
it's not just going to be a sam altman thing it's going to be an everyone's

Ejaaz:
involved type of thing kind of like world ending event yeah

Josh:
Too big to fail so i i do love

Josh:
this incentive structure where everyone is incentivized to make it work because

Josh:
everyone's equally at risk in terms of their exposure to the technology so that

Josh:
i think i could be happy to sleep at night where at least u.s and china are

Josh:
aligned in one thing in which they want to achieve agi they want the smartest

Josh:
models they're going to make their money pay off the best they can so,

Josh:
All the power to them. But I think, is that a wrap for us today,

Josh:
EJ? We got anything else?

Ejaaz:
That is a wrap,

Josh:
Josh. That's it. That's a wrap on our little like XAI mini episode.

Josh:
There was one fact that I wanted to just do a little like fun fact check, which is a gigawatt.

Josh:
And according to Grok, it powers approximately 750,000 to 850,000 average US homes per one gigawatt.

Josh:
So the scale we're talking is like a tremendous amount of gigawatts.

Josh:
I mean, this NVIDIA project is 10 of those, which means that's about,

Josh:
I mean, on the high end, eight and a half million U.S.

Josh:
Homes can be powered by a singular data center. So we're going to hope this works out.

Josh:
I think right now it seems like, I mean, Grok is cooking. The XAI team is on

Josh:
fire and they are in between models.

Josh:
I cannot wait until they get this new Colossus training cluster up or even Microsoft's.

Josh:
I mean, Microsoft's got a huge cluster.

Josh:
What are you doing with it, dog? Like, let's see. Let's see your stats.

Josh:
Let's see your numbers. Put a number up on the ArcGIS leaderboard. um

Josh:
but yeah i think that's that's a wrap on all the fun exciting new

Josh:
things about xai the comment section is by energy stocks yeah

Josh:
by energy stocks um i we read

Josh:
all the comments i read every single comment i try to reply to them too so i

Josh:
would love for you to share either what you think about the show or or who

Josh:
you think is winning this ai race currently do you like are

Josh:
we just kind of like do we have elon derangement syndrome are

Josh:
we just kind of like obsessed with everything he builds or is this it feels

Josh:
like it's pretty grounded i feel like we have some good examples about how well

Josh:
they're doing so i'd love to hear if you agree or disagree that

Josh:
would be a fun little thing for the comments but anyway that's a wrap on today's episode

Josh:
we have a couple more exciting ones coming this week so buckle up uh

Josh:
the next one the next one coming i think ejaz myself

Josh:
and we might even have a guest for that episode we'll be probably in an all-out

Josh:
brawl it's good that we're recording remotely because we might like the blood

Josh:
could possibly be drawing so yeah buckle up for for that one there's a lot to

Josh:
look forward to this week but that's it for this week this episode so thank

Josh:
you so much for watching as always please don't forget to subscribe like comment,

Josh:
all the fun things, share it with your friend.

Music:
Music