The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 17 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 0 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Rao: Trillion-dollar GPU data centers are physically impossible to power
“None of this is physically realizable, right? That's what's really kind of annoying to me. I'm like, okay, you're talking about, you know, a trillion dollars of GPUs. Where the fuck are you gonna get the power for that? I'm sorry, right? Like you're just, you'…”
Rao: xAI's GPU clusters broke rules and are not repeatable
“And the things that have been realized I mean, like the XAI clusters and stuff like that, You know, valiant efforts. No doubt about it. And it's amazing to push through and get it done, but it broke a lot of rules and is doing a lot of stuff very inefficiently…”
Rao: Circular AI mega-deals are justified by coming demand
“I know we talk about it looks bad or whatever, but the bet is that the demand will come, and I think that's actually a good bet. So I see it as like, all right, fine, it looks, it doesn't look great in the short term, and I can see why people are irked by it b…”
Rao: True AI requires fundamental hardware substrate innovation
“My personal take is that we need innovation on the hardware substrate to actually build true intelligence. You're not going to get there with this sort of rudimentary thing that we have.”
Rao: Nervana Was Sold to Intel Too Early
“And so, you know, we sold too early. Sold to Intel too early.”
Rao: Unconventional AI Could Open a Computing Paradigm Lasting 80 Years
“I think we have the opportunity to completely open up a new paradigm that could exist for another 80 years.”
Rao: True hardware innovation requires exploiting non-linear silicon physics
“What's really interesting is if something is linear and predictable, it actually is very easy to simulate. If you can simulate in a digital system, it's less interesting. It turns out, like, silicon has all these beautiful physics that are associated with it. …”
Rao: Unconventional AI has very high chance of hitting 50x efficiency first go
“Are we going to hit a thousand X in the first go? Maybe, maybe not. I don't know. There's a some percentage chance that we will hit it, but I think we're going to hit 50 X with a very high chance.”
Rao: MosaicML grew from $20M to over $700M at Databricks
“When I came there, we had about a twenty million dollar business, and it was, you know, seven or eight hundred million when I left, something like that. I think it's crossed well over a billion.”
Rao: Unconventional AI will probably need $1.5B for its first product
“I think to get to actual first product, it's probably a billion and a half is my guess. And it sounds like a huge number, but I think the opportunity is enormous. So this initial money will get us through much of the exploration and sort of a product, not prod…”
Rao: Company efficiency drops precipitously above 80 to 100 people
“Efficiency drops Really precipitously as you go over, like, 80 to a hundred people. That's what I've seen over and over again.”
Rao: Rat and human cortex microcircuits are virtually indistinguishable
“As a trained neuroscientist, I can look at the, what we call the cytoarchitecture, basically the cell structure of the cortex of a rat And a human, I can't tell the difference. So the human or the evolutionary biology version of Moore's law kind of ended about…”
Rao: Core physics principles of computing haven't changed in decades
“Much of this is built on a foundation of physics that we didn't really go back and change that much. I mean, yes, we built smaller devices, faster devices, but the principles that that it's built on haven't changed in a long time.”
Rao: Unconventional AI will deliver a full-stack system, not just chips
“We have to, we're going to do a whole stack. Like we're not just going to sell a chip. You can't do that because we're changing the whole paradigm.”