Latest / In the Loop AI / The Unknown Tech that Enables AGI: Claude Mythos and NVIDIA's Next Generation

Transcript

The Unknown Tech that Enables AGI: Claude Mythos and NVIDIA's Next Generation · In the Loop AI

0:00
Josh: A couple of weeks ago, we covered the Claude Mythos release,
0:03
Josh: the model that found decade old security flaws overnight and scared the hell
0:06
Josh: out of basically anyone who is following the AI story.
0:08
Josh: So much so that the federal government is involved. But the part that we didn't
0:11
Josh: get into is the backend that powered this model.
0:14
Josh: Mythos was built on a chip from March, 2024 that Jensen pulled out of his pocket
0:19
Josh: on stage at GTC, which was the Blackwell chip.
0:22
Josh: It had 208 billion transistors. Everyone treated it like the future had arrived.
0:25
Josh: And yet it took two years of fabrication for us to get the first manifestation
0:29
Josh: of that, which is Claude Mythos.
0:32
Josh: 24 models from Keynote to a working model. It happened with Hopper,
0:35
Josh: it happened again with Blackwell, and it's going to happen again with our future models.
0:39
Josh: But the difference is we have a series of future models that exist today that
0:43
Josh: we can kind of map out to where we're going to be heading based on this trajectory
0:47
Josh: that we've seen with the previous chips.
0:48
Josh: And it's pretty awe-inspiring to see where we are going to go considering there
0:51
Josh: are three generations of chips that have already been announced since Blackwell.
0:55
Josh: We have Vera Rubin, Rubin Ultra, and Feynman.
0:58
Josh: Each one, many multiples more powerful than the last. And when you look at what
1:02
Josh: Blackwell already produced in the very first version, it gets impossible to
1:06
Josh: imagine a world where we don't reach AGI on hardware that's already been designed.
1:10
Josh: Everything that's been announced that is going into production almost certainly
1:14
Josh: is going to produce models indistinguishable from AGI. At least that's what
1:17
Josh: it seems like on surface level?
1:18
Ejaaz: Yeah, so the story here in a single sentence is AGI, like AI models, are already here.
1:26
Ejaaz: We just haven't distributed it because we haven't powered up the GPUs that enable
1:31
Ejaaz: it. So everyone is obsessed with AI models.
1:34
Ejaaz: We talk about our favorite models, how we prompt them, how intelligent they
1:36
Ejaaz: are. But very few people are talking about the fact that
1:39
Ejaaz: The hardware is the thing that powers these things. They train these things.
1:43
Ejaaz: They inference these things.
1:43
Ejaaz: And it's still about 70% of the influence of how intelligent your model is.
1:48
Ejaaz: And the prime example, most recent example of that has been Anthropics Mythos
1:52
Ejaaz: release, right? You just mentioned it. It's discovered a bunch of different cybersecurity flaws.
1:55
Ejaaz: It is this all being powerful thing that the governments around the world, including the U.S.
2:00
Ejaaz: Government, Federal Reserve, they're sharing meetings with the top banks to
2:03
Ejaaz: talk about the craziness of this model we must prepare.
2:06
Ejaaz: There's a lot of doomer news out there in the future.
2:09
Ejaaz: Little do you know that this was powered by a GPU or this was trained by a GPU
2:14
Ejaaz: that was built 20 months ago. So we're talking about almost two years ago.
2:19
Ejaaz: It's called Blackwell. And I want to give you guys an idea of the timeline of what this looked like.
2:24
Ejaaz: So in March 2024, NVIDIA GTC, which is like their developer conference,
2:28
Ejaaz: Jensen Huang comes on stage and he presents this gargantuan scrap of metal.
2:33
Ejaaz: It looks very pretty, by the way. And he goes, this is Blackwell,
2:37
Ejaaz: GB200, GB300, a brand new GPU.
2:40
Ejaaz: We can train frontier models on it. Everyone gets so excited.
2:44
Ejaaz: Their stock price absolutely ascends, right?
2:47
Ejaaz: The thing is, people couldn't get their hands on this until exactly a year later.
2:52
Ejaaz: So to give you guys an idea of the timeline, he announces it in March 2024.
2:56
Ejaaz: Then by the middle of the year, they discover there's like a bit of a design
3:00
Ejaaz: flaw and they amend that.
3:02
Ejaaz: And then by the end of 2024, early 2025, they start shipping these units of
3:07
Ejaaz: Blackwell GPUs out to the top frontier AI labs.
3:10
Ejaaz: But there's an important nuance here, which is it's just the GPU sitting in a data center.
3:15
Ejaaz: They aren't actually powered up. It's not until 6 to 12 months after that fact
3:20
Ejaaz: that these GPUs were finally powered up,
3:23
Ejaaz: used to train models, which is why we now start to see these new AGI-like models
3:27
Ejaaz: like OpenAI SPUD and Claude Mythos come to fruition.
3:31
Ejaaz: So the point is, there is a long gap between the frontier GPUs being announced
3:36
Ejaaz: and rolled out to them actually being powered to train the models.
3:40
Ejaaz: We talked about Elon Musk and XAR a lot on this show before.
3:42
Ejaaz: They actually have the largest arsenal of these Blackwell GPUs.
3:47
Ejaaz: They bought about a million of them.
3:49
Ejaaz: The crazy part about this now is they're not like one, two, but three new NVIDIA
3:53
Ejaaz: GPU models that have been announced in the recent NVIDIA GTC.
3:57
Ejaaz: So there is a major lag between Frontier hardware and the new AI models that are being released.
4:03
Ejaaz: And people don't understand this. And we want to tell you the story.
4:05
Josh: You just remember GPT-4, how long ago that was and how that felt like the huge,
4:11
Josh: most pivotal model that OpenAI ever released.
4:13
Josh: I mean, that was the big one right after ChatGPT came out. That was trained
4:17
Josh: using the Hopper chips. You know, the most recent model.
4:20
Ejaaz: Hopper's a word I haven't heard in a while, Josh.
4:22
Josh: Yeah, well, you know, GPT 5.4, the most recent model that we're using every
4:26
Josh: single day on ChatGPT. That was also trained on Hopper chips.
4:30
Josh: The same chips are training models from GPT-4 to GPT-5.4.
4:35
Josh: And it's a testament to how the efficiency gains of software can actually increase
4:40
Josh: the throughput of hardware.
4:42
Josh: And I think I want to use that as an example because what we just got recently
4:47
Josh: with Mythos through Anthropic, that seems to be the first real implementation process.
4:52
Josh: Of a true Blackwell model. And rumors are that SPUD, the new open AI model,
4:56
Josh: is going to kind of be the same in terms of power that is coming as it relates
5:00
Josh: to the first Blackwell model.
5:01
Josh: And even if we don't actually iterate on the hardware, the amount of progress
5:06
Josh: we're going to get from Blackwell models alone seems like it is going to be
5:10
Josh: difficult to imagine it doesn't become some sort of an AGI, right?
5:13
Josh: It's like when you think about the difference of intelligence between GPT-4
5:16
Josh: and GPT-5.4 and how far we've come, that applied to Blackwell at this new scale,
5:22
Josh: seems crazy but that's not even the crazy part because
5:25
Josh: we have an entire roadmap of these three generations of
5:28
Josh: chips that are coming that we can very clearly map to
5:31
Josh: the gains that we're going to see and i think that's when things
5:34
Josh: get like particularly disturbing because on the chart that we're looking on
5:37
Josh: screen now we have blackwell that's where we are right now blackwell is a significant
5:41
Josh: improvement over the previous model but then we have vera rubin which jumps
5:46
Josh: from 20 petaflops to 50 petaflops that's a two and a half to five times multiple
5:50
Josh: on the compute then we We have Ruben Ultra,
5:53
Josh: which is scheduled for the second half of 2027.
5:56
Josh: That is a 14 times multiple.
5:59
Josh: And then we have Feynman in 2028, which is an estimated 30 to 50 times multiple.
6:06
Josh: On the current chip stack that we have today, assuming that we get no software progress at all.
6:12
Josh: And what we saw with the Hopper chips is that we got a tremendous amount of
6:16
Josh: progress just from software.
6:18
Josh: So when you combine this 30 to 50 times multiple with a maybe another 100 times
6:23
Josh: multiple on software, if we make another breakthrough, we're looking at some
6:25
Josh: pretty insane improvements here that like are really hard to wrap your head around.
6:30
Ejaaz: I want to point out that these improvements, these multiples that you just mentioned
6:33
Ejaaz: are just on the speed and power of these hardware modules, right?
6:39
Ejaaz: So it's going to work 3x harder or 14x harder, but it's also going to cost you
6:44
Ejaaz: a lot less to be able to train the same type of intelligence or model.
6:47
Ejaaz: So the intelligence per density, which is a unit that we completely made up,
6:52
Ejaaz: and we don't know if it exists, but it somehow rhymes in my head at least,
6:56
Ejaaz: is improving and it's going to be cheaper with each successive model.
7:00
Ejaaz: But if you want to get a bit of context as to like what that looks like in terms
7:04
Ejaaz: of like the models that you use today and what it's going to look like tomorrow,
7:07
Ejaaz: we have this other table here, which kind of like maps it out.
7:11
Ejaaz: So with Blackwell today, you get about a two to three X more intelligent, crazier model, right?
7:16
Ejaaz: That's what Claude Mythos is supposedly meant to be. It's like a larger size.
7:19
Ejaaz: It's trained on these Blackwells.
7:21
Ejaaz: You're going to see a bunch of models similar come out from OpenAI and XAI over
7:24
Ejaaz: the next couple of months.
7:25
Josh: Just to pause you there, these are already models deemed too dangerous to release for the public.
7:30
Ejaaz: Yes. Just some emergency meetings literally being called by the federal chair, top banks.
7:36
Ejaaz: Actually, I read something yesterday that the NSA is using or conferring or
7:42
Ejaaz: re-engaged with Anthropic, as well as the Pentagon and the U.S.
7:46
Ejaaz: Defense Department, after banning and blacklisting Anthropic because it's so powerful.
7:52
Josh: And that's where we are today.
7:53
Ejaaz: That's today. So that's right here. 2026, two to three X, right?
7:57
Ejaaz: Yeah, crazy. Now, you might notice that by next year, we have a larger multiple
8:01
Ejaaz: on the original multiple.
8:03
Ejaaz: By next year, we're going to have a 10 to 15x improvement purely through Vero
8:08
Ejaaz: Rubin GPUs. Now, I must emphasize...
8:10
Ejaaz: This does not include post-training. This doesn't include all the fine,
8:14
Ejaaz: fancy techniques that AI labs themselves will implement to make a smart model.
8:18
Ejaaz: This is just the hardware.
8:21
Ejaaz: It's like buying the hardware and training a model today versus next year,
8:25
Ejaaz: you're gonna get a 10 to 15x more intelligent model, but it gets even scarier.
8:28
Ejaaz: 2028, 30 to 50x.
8:31
Ejaaz: 2029, 100 to 200x. Now, I haven't seen these multiples in any other industry
8:37
Ejaaz: for any kind of performance or hardware improvement.
8:39
Ejaaz: So I can't wrap my head around this because it looks like just a few small numbers
8:45
Ejaaz: that are getting larger, but these are multiples of its predecessor,
8:49
Ejaaz: which means that we're probably going to get AGI,
8:53
Ejaaz: honestly, by the start of next year.
8:55
Ejaaz: And they're trained on hardware that currently exists and is rolling out.
8:59
Ejaaz: I don't know. I'm just kind of scared reading all of this, to be honest,
9:01
Ejaaz: because what happens if we have universal access to this?
9:04
Ejaaz: There's going to be a load of malicious actors which can use these models for
9:07
Ejaaz: various different things. But also, I don't know what these models are going
9:10
Ejaaz: to be capable of. They're going to be so much smarter than humans themselves.
9:13
Josh: The disturbing thing is that this technology is here. Like this is,
9:17
Josh: it's no longer an engineering problem or a physics problem necessarily.
9:21
Josh: It's just a matter of actually producing the thing and plugging it into an outlet and putting it online.
9:25
Josh: And this is coming. Like there are no novel breakthroughs required to make this a reality.
9:30
Josh: Now, what that looks like on the other side, I don't know, but I think it's
9:33
Josh: safe to assume the velocity of improvement we're going to get is certainly not
9:37
Josh: slowing down. It is turning more closely resemble a vertical line than anything else.
9:42
Josh: And I think it begs the question, like, at what point do we reach AGI and how do we even define that?
9:47
Josh: Because I'm not sure we spoke about that much on the show, but Ejaz,
9:50
Josh: when you say AGI, what do you mean by AGI? What would you be looking for?
9:55
Josh: To declare, okay, we have finally reached AGI.
9:58
Ejaaz: Okay, so this is like my own made-up definition, but it's what will make me go, okay, this is AGI.
10:06
Ejaaz: It would be a single AI model, not many, but a single AI model that advances
10:13
Ejaaz: the frontier of three key major industries autonomously. So I'll pick these
10:18
Ejaaz: industries as examples.
10:20
Ejaaz: Financial industry, so it trades better than the average world.
10:24
Ejaaz: Sorry, then the best hedge fund or investor.
10:27
Ejaaz: It is able to make assessments better than any of the financial analysts,
10:31
Ejaaz: the top experts, et cetera, in that industry.
10:33
Ejaaz: In science, it has discovered a bunch of medical cures for some major diseases
10:38
Ejaaz: such as cancer, Alzheimer's, and stuff like that, that scientists,
10:41
Ejaaz: top scientists at their top level could not figure out. It accelerates their research.
10:46
Ejaaz: And maybe one other industry that I can't think of right now,
10:49
Ejaaz: but it's when these models start doing things that the best of the best humans
10:53
Ejaaz: right now couldn't figure out themselves and couldn't have seen themselves.
10:57
Ejaaz: Do you have a similar definition or?
10:59
Josh: Yeah, I think that sounds right. I think, and again, it's very fuzzy.
11:02
Josh: Everyone kind of has their own custom definition of what they believe AGI is going to be.
11:05
Josh: But for me, it's just AI that's smarter than the smartest human at pretty much
11:10
Josh: any cognitive task that exists.
11:12
Josh: So you can go to this model and it will be better
11:15
Josh: anyone else who you can ask on planet earth about anything and
11:18
Josh: the problem with models today is they're very spiky like you can do this
11:21
Josh: for code probably and it can code better than every
11:24
Josh: human on earth but if you ask it you know a generalized question
11:27
Josh: about something that you really know a lot about there's a
11:30
Josh: lot of times where it's not completely accurate or it will respond
11:33
Josh: as if it has the intelligence of a three-year-old it fails the
11:36
Josh: reasoning tests of a lot of simple things it still feels like
11:39
Josh: it's this very spiky entity once it is fully
11:42
Josh: developed once it is actually better at every cognitive task
11:45
Josh: that includes physical things too that includes like understanding physics
11:48
Josh: of the real world world models that feels like agi and
11:51
Josh: then artificial super intelligence asi feels like
11:54
Josh: it is smarter than all humans combined so it's like if we put all of our brains
11:58
Josh: together no matter how long we tried we can never come up with the things that
12:02
Josh: artificial super intelligence will come up with and i mean will we get there
12:06
Josh: using this chip architecture possibly I'm seeing a 50x multiple,
12:12
Josh: not including the software multiples.
12:13
Josh: And like those compounding on top of each other at the rate that we're moving,
12:17
Josh: seems like the only real constraint is going to be physical.
12:20
Josh: It's going to be actually rolling out these models and powering them on.
12:23
Ejaaz: Well, another crazy thing is, I think a lot of people, including myself,
12:28
Ejaaz: would assume that with every chip upgrade, it's going to be more expensive,
12:33
Ejaaz: and it's going to be bigger.
12:35
Ejaaz: It's going to be clunkier, right? Like the data centers are going to get bigger,
12:38
Ejaaz: it's going to be more expensive.
12:40
Ejaaz: I wish I had a chart to show this, but it's actually the complete inverse.
12:43
Ejaaz: And I'll give you some examples, some numbers to explain that, right?
12:47
Ejaaz: So a reasoning task that costs $1 on Blackwell costs $0.20 on Vero Rubin,
12:54
Ejaaz: which is rolling out as we speak or later this year.
12:57
Ejaaz: And it'll only cost $0.07 on Rubin Ultra, which starts to get released by the start of next year.
13:03
Ejaaz: So the cost is going down pretty massively.
13:06
Ejaaz: Now, by 2028, Jensen announced the Feynman GPU, right?
13:10
Ejaaz: A single rack of that. So we're talking about like just a couple of that.
13:14
Ejaaz: Blocked on top of each other, will process more compute than was required to
13:19
Ejaaz: train GPT-4 that you mentioned earlier, Josh.
13:23
Ejaaz: So the point is, less is more, but somehow more powerful, but also somehow more
13:29
Ejaaz: cheap relative to the intelligence that you're building.
13:32
Ejaaz: And if you assume this intelligence is going to reach this ASI,
13:36
Ejaaz: AGI-like state, it's going to make you money as well.
13:39
Ejaaz: So you end up just having i guess i i'm afraid to say this but the best of old
13:44
Ejaaz: worlds both worlds i don't know what humans are going to be doing but it's great for ai.
13:47
Josh: Basically yeah there's no world in which things don't
13:50
Josh: get better and it feels like right now we're really just constrained by this
13:53
Josh: compute power there's this great meme that i saw online it's
13:56
Josh: it it said uh mythos is too powerful for public release
13:59
Josh: but the reality is is that they're just completely out of compute and
14:02
Josh: anthropic can't actually supply the tokens required to give
14:04
Josh: mythos to the world these optimizations these cost structures
14:07
Josh: yeah there it is we got on screen now great meme
14:10
Josh: great meme but these these cost structures that are
14:13
Josh: going to incur from these new models are going to completely destroy that factor
14:18
Josh: at least for now until whatever that next generation of model is that is so
14:22
Josh: powerful that it's constraining gpus and the interesting thing is that open
14:25
Josh: ai has the same exact thing going on all these models are kind of converging
14:29
Josh: on the same spot but they all seem to be compute constrained.
14:33
Ejaaz: I think what critics will push back on though, Josh, for everything that we've
14:37
Ejaaz: said so far is, okay, cool.
14:39
Ejaaz: You can buy these new hardware things, but why would you do that if you could
14:43
Ejaaz: just wait a few months or six months and buy the next thing?
14:46
Ejaaz: Jensen's just shipping out these products. He's making a load more money.
14:49
Ejaaz: It doesn't make sense. These things are depreciating assets.
14:52
Ejaaz: By the time you've bought the first one and you've ramped that up with power
14:56
Ejaaz: and training your next model, there's already three other new chip architectures.
14:59
Ejaaz: And he would be right, that critic would be right,
15:03
Ejaaz: except that they're massively, massively wrong. And we have proof for that,
15:06
Ejaaz: right? GPUs have now become this anti-depreciation machine.
15:10
Josh: One of the most amazing things about this phenomenon, and it feels like a narrative
15:14
Josh: violation, is the idea that the GPUs that were released three years ago are
15:19
Josh: actually more valuable today than they were at the time they launched,
15:23
Josh: which is a pretty bizarre idea.
15:25
Josh: We have this artifact on screen that shows a chart.
15:27
Josh: And an H100 from NVIDIA cost $30,000 when it launched in 2023.
15:32
Josh: At its peak because of the scarcity because everyone
15:35
Josh: needs these things it was selling for a four times multiple at 120 000
15:38
Josh: per h100 this is kind
15:41
Josh: of outrageous it was a little exorbitant we don't need to
15:44
Josh: be paying that much money but now that they are old they're not depreciated
15:47
Josh: but there's much better hardware out there they're still holding their price
15:51
Josh: at 30 000 in fact you can see a rebound that happens in late 2025 where the
15:56
Josh: cost of these h100 gpus actually ticks upwards And I think a lot of the people,
16:02
Josh: Michael Burry most famously, who is the guy behind the big short,
16:05
Josh: He created an entire short thesis around the idea that the depreciation schedule
16:09
Josh: of these GPUs wasn't aggressive enough and they were actually going to lose
16:13
Josh: their value and therefore the market was going to deflate because the companies
16:17
Josh: weren't marking these down properly.
16:19
Josh: The reality is, is that not only are they not going down, they're starting to
16:22
Josh: trend back up because the incremental cost for a token is so low with these
16:27
Josh: and everyone's so desperate for compute that they're like, well,
16:29
Josh: might as well spend some extra money,
16:31
Josh: get the H100s and start generating inference tokens with them.
16:34
Josh: It's this pretty amazing phenomenon that's happening.
16:36
Ejaaz: Yeah, so if you're wondering why this is happening, explicitly it's AI demand
16:40
Ejaaz: is growing faster than chip supply can expand.
16:43
Ejaaz: We don't have enough fabs or the manufacturing prowess or the energy grid to
16:47
Ejaaz: support creating and generating more GPUs to satiate the demand that we're seeing
16:52
Ejaaz: in AI across all these different industries, right? It's a very pervasive bit of technology.
16:57
Ejaaz: Now, the data that we're showing you on the screen right now isn't siloed to
17:01
Ejaaz: like a few research papers. This is happening in the market right now,
17:04
Ejaaz: and it's incredibly liquid.
17:07
Ejaaz: So a new phenomenon of companies in AI whose stocks have all skyrocketed are
17:13
Ejaaz: these things called neoclouds, right?
17:15
Ejaaz: So these are like, think of it as like AWS. They supply compute to train your
17:20
Ejaaz: AI models by setting up their own data centers, and they kind of like provide
17:23
Ejaaz: it to you in like a cloud or data center specific structure.
17:26
Ejaaz: Examples would be CoreWeave, for example. The idea here is these data centers or these GPU providers.
17:33
Ejaaz: 70% of the GPUs that they're running are old GPUs that we're showing you on our screen right now.
17:38
Ejaaz: And they're booked out, I'm not exaggerating, 6 to 12 months in advance.
17:44
Ejaaz: In fact, they're done so in contracts and the same providers renew the contracts
17:48
Ejaaz: three months before the contract needs to be renewed just to make sure that
17:51
Ejaaz: they get access to these older GPUs.
17:54
Ejaaz: So the point I'm trying to make, and you mentioned this just now,
17:56
Ejaaz: Josh, is all that matters is can I get AI tokens generated to do the thing that
18:02
Ejaaz: my company needs or answer the prompt that I have?
18:06
Ejaaz: And if the answer is yes, and it's for a reasonable price, I'm down to go for
18:10
Ejaaz: that because the value that you can build and earn on top of that is invaluable,
18:14
Ejaaz: right? They can have a large markup on that.
18:16
Ejaaz: So it makes sense that these assets are kind of like in high demand.
18:20
Ejaaz: And to your earlier point, Michael J. Burry like shorted the entire market saying
18:24
Ejaaz: that these are depreciating assets and he got that completely wrong.
18:26
Ejaaz: And his thesis specifically was based on it can't train frontier models.
18:31
Ejaaz: And he's actually right.
18:32
Ejaaz: The older models can't train frontier models. But what they are being used for
18:36
Ejaaz: is one thing very specifically, inference, which is if someone has a question,
18:40
Ejaaz: how do I get them the answer? How do I process the prompt?
18:43
Ejaaz: That's what the older GPUs are being used for. And they're really damn good at it.
18:47
Ejaaz: And the reason why it's important and essential for AI labs specifically who
18:50
Ejaaz: are training models, who you might think might want the expensive models is
18:54
Ejaaz: they have a ton of inference.
18:55
Ejaaz: They use inference to even train the new models. So it's this new paradigm where
19:00
Ejaaz: all these old GPU architectures are being re-found or repurposed for this really
19:05
Ejaaz: important thing that is inference.
19:07
Ejaaz: So important context to understand if you're investing in some of these companies, for example.
19:11
Josh: Yeah. And why is it so valuable? Well, it's a testament to the software improvements,
19:14
Josh: right? So we have those software efficiency improvements that we didn't have three years ago.
19:17
Josh: So that same hardware generates a lot more value.
19:20
Josh: And if we scroll down to the value multiplier section of this artifact it shows
19:24
Josh: that the cost of a chatbot inference in 2023 was three dollars an hour and now
19:29
Josh: autonomous agents completing these complex tasks is 30 to 300 dollars per hour
19:34
Josh: The value that you can charge for these tokens is significantly higher than it was in the past.
19:40
Josh: And the amount of tokens that you're able to generate efficiently at higher
19:44
Josh: quality is much higher as well.
19:45
Josh: So there's all these converging forces that are just making the market desperate for compute.
19:50
Josh: Nobody has the compute required that they want. And NVIDIA is trying to put
19:54
Josh: it online as fast as they can, but it's not fast enough.
19:56
Josh: And I assume as we go through this, we're going to continue to see varying bottlenecks
20:00
Josh: and the efficiencies will move to where there are bottlenecks,
20:03
Josh: which creates new bottlenecks right now we're seeing some convergence around
20:07
Josh: cpus and cpus seem to be like they're going to be hitting a
20:10
Josh: shortage somewhat soon because we're out of gpus let's move to cpus
20:13
Josh: and it's it's this really interesting dynamic but that is the idea
20:16
Josh: on this nvidia episode or just the chip episode in
20:19
Josh: general that it is hard to imagine a world in which we don't reach
20:22
Josh: agi given the currently announced infrastructure it
20:25
Josh: doesn't require any breakthroughs it's just if nvidia does
20:29
Josh: what they announced on stage through jensen huang through these next three
20:32
Josh: chips it is almost impossible to imagine what the world of intelligence is going
20:36
Josh: to look like and i think it's important to understand is that mythos is trained
20:39
Josh: on a two-year-old chip and no one's really talking about that so it blew my
20:44
Josh: mind hopefully it blew yours as well uh at least found it a little bit fascinating
20:47
Josh: and that is our episode today thank you guys so much for watching we really appreciate it
20:50
Ejaaz: And i know some of you are probably thinking oh there's a bunch of challenges
20:53
Ejaaz: here and josh actually just mentioned one of them which is like you got cpus
20:57
Ejaaz: we don't have enough energy, we don't have enough memory.
21:01
Ejaaz: And that's like another episode that we can get into.
21:04
Ejaaz: So all of those things assumed will be leveled at some point.
21:08
Ejaaz: And we're gonna see all those industries grow versus being constrained.
21:12
Ejaaz: People are throwing trillions of dollars into this industry.
21:15
Ejaaz: So all of those problems should theoretically be fixed.
21:18
Ejaaz: But rest be sure, we will be the first show to cover it and give you those thoughts
21:21
Ejaaz: before it happens, by the way.
21:22
Ejaaz: And Intel is a sneaky one to get into. But we'll talk about that another time.
21:26
Ejaaz: Thank you so much for listening. If you are not subscribed to us, please subscribe.
21:30
Ejaaz: It helps us out massively. We are having banger weeks on YouTube,
21:34
Ejaaz: Spotify, Apple, and wherever you listen to us.
21:36
Ejaaz: Please rate us. Leave us a comment. We love hearing your feedback.
21:40
Ejaaz: There are like thousands of newbies that are listening to the show, welcome.
21:44
Ejaaz: And also give us feedback about stuff that we may not be covering that you want
21:47
Ejaaz: to hear more of. We're always open to feedback.
21:49
Ejaaz: But until then, I guess we'll see you on the next one.