Nathan Lambert
speaker
1,814 appearances
3 recordings
2 series
first heard Feb 2025
last heard 1 Feb
Nathan Lambert’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 2 in all, peaking in Feb 2026 with 1.
Appearances
I think to some extent we have capabilities that hit a certain point where any one person could say, oh, okay, if I can leverage those capabilities for X amount of time, this is AGI, right? Call it 27, 28.
But then the cost of actually operating that capability is so, so extreme that no one can actually deploy it at scale and mass to actually completely revolutionize the economy on a snap of a finger. So I don't think it will be like a snap of the fingerprint.
physical constraint rather it'll be a you know oh the capabilities are here but i can't deploy it everywhere right and so one one simple example going back sort of to 2023 was when uh you know being with gpt4 came out and everyone was freaking out about search right Perplexity came out.
If you did the cost on like, hey, implementing GPT-3 into every Google search, it was like, oh, okay, this is just like physically impossible to implement, right? And as we step forward to like going back to the test time compute thing, right? A query for, you know, you ask ChatGPT a question, it costs cents, right? For their most capable model of chat, right? To get a query back.
To solve an Arc AGI problem, though... cost five to 20 bucks, right? And this is- It's only going up from there. This is a thousand, 10,000 X factor difference in cost to respond to a query versus do a task. And the task of Arc AGI, it's not like it's like, it's simple to some extent, but it's also like, What are the tasks that we want? Okay, AGI, quote unquote, what we have today can do Arc AGI.
Three years from now, it can do much more complicated problems, but the cost is going to be measured in thousands and thousands and hundreds of thousands of dollars of GPU time, and there just won't be enough power, GPUs, infrastructure to operate this and therefore shift everything in the world on the snap of the finger. But at that moment, who gets to control and point the AGI at a task?
And so this was in Dario's post that he's like, hey, China can effectively and more quickly than us point their AGI at military tasks, right? And they have been in many ways faster at adopting certain new technologies into their military, right? Especially with regards to drones, right? The U.S. maybe has a longstanding, you know, large air sort of, you know, fighter jet type of thing, bombers.
But when it comes to asymmetric arms such as drones, they've completely leapfrogged the U.S. and the West. And the fear that Dario is sort of pointing out there, I think, is that Yeah, great. We'll have AGI in the commercial sector. The US military won't be able to implement it super fast.
Chinese military could, and they could direct all their resources to implementing it in the military and therefore solving military logistics or solving some other aspect of disinformation for targeted certain set of people so they can flip a country's politics or something like that that is actually catastrophic.
versus, you know, the US just wants to, you know, because it'll be more capitalistically allocated just towards whatever is the highest return on income, which might be like building, you know, factories better or whatever.
And I think going back to my viewpoint is if you believe we're in this sort of stage of economic growth and change that we've been in for the last 20 years, the export controls are absolutely guaranteeing that China will win long term, right? If you do not believe AI is going to make significant changes to society in the next 10 years or five years.
Five-year timelines are sort of what the more executives and such of AI companies and even big tech companies believe. But even 10-year timelines, it's reasonable. But once you get to, hey, these timelines are below that time period, then the only way to create a sizable advantage or disadvantage for America versus China is if you constrain compute. Because
Talent is not really something that's constraining, right? China arguably has more talent, right? More STEM graduates, more programmers. The US can draw upon the world's people, which it does. There's tons of foreigners in the AI industry.
Yeah. I mean, many of them are Chinese people who are moving to America, right? And that's great. That's There's that talent is one aspect, but I don't think that's one that is a measurable advantage for the US or not.
It truly is just whether or not compute right now, even on the compute side, when we look at chips versus data centers, right, China has the unprecedented ability to build ridiculous sums of power.
clockwork right they're always building more and more power they've got steel mills that that like individually are the size of the entire u.s industry right and they've got aluminum mills that consume gigawatts and gigawatts of power right and when we talk about what's the biggest data center right opening i made this huge thing about stargate their announcement there that's not that's like once it's fully built out in a few years it'll be two gigawatts right of power
And this is still smaller than the largest industrial facilities in China. China, if they wanted to build the largest data center in the world, if they had access to the chips, could. So it's just a question of when, not if.
Chips are a little bit more specialized. I'm specifically referring to the data centers. Chips, fabs take huge amounts of power. Don't get me wrong. That's not necessarily the gating factor there. The gating factor on how fast people can build the largest clusters today in the US is power.
Now, it could be power generation, power transmission, substations, and all these sorts of transformers and all these things. building the data center. These are all constraints on the US industry's ability to build larger and larger training systems, as well as deploying more and more inference compute.
Less than 10 years or five years to above, right? China will win because of these restrictions long term unless AI does something in the short term, which I believe AI will do, you know, make massive changes to society in the medium short term, right? And so that's the big unlocker there. And even today, right, if Xi Jinping decided to get, you know, quote unquote, scale-pilled, right, i.e.
Showing 1341–1360 of 1,814 · page 68 of 91
← Previous
Next →