Dave Blundin

speaker

31,926 appearances 100 recordings 1 series first heard Apr 2025 last heard 19 Sep

Dave Blundin’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 74 in all, peaking in Jul 2026 with 8.

Appearances

newest first · ▶ plays the moment

Moonshots with Peter Diamandis · Robinhood's Vlad Tenev on Tokenizing Everything, OpenAI's 6 Misalignment Reports, Figure's Robot Makes Beds | EP #292 · 19 Sep 2026 podcast

And I think you can do all that automatically now with an AI agent just thinking through the math. voice-verified

And they said, okay, well, now that it's rippled through and we've simplified the math tremendously, we can just run a quick test and see if this approximation is as good as the softmax version was, which is a lot harder to compute. voice-verified

And so, you know, a lot of people would think math is arcane, math is irrelevant, math is over here, it's some other thing, you know, these you know, Fermat's last theorem and whatever, it's all over in this wing. voice-verified

But in reality, it directly ties to the optimization of the AI within its own performance and then the self improvement, recursive self improvement loop. voice-verified

So I mean it it's exactly the same process that you were just describing where you just you just have a simple mental model of a mathematical adjustment and then you ripple through all the way to the NVIDIA voice-verified

GPU performance at the transistor level. voice-verified

Well voice-verified

Yeah. voice-verified

And then I tweak the input and I get a massively different output. voice-verified

You can formalize the the the radical difference from expectation. voice-verified

You can you can measure that and and formalize and bound that too. voice-verified

Um so you know, like everybody knows somebody that is ninety nine percent of the time perfectly rational and then but when they're off, they're really off. voice-verified

I mean like they're dangerous. voice-verified

And and that attribute exists in neural nets too. voice-verified

You know, if you if you don't have the parameters set just right, uh it can be wildly off, which you know in a self-driving car is like holy crap, it it just completely drove off a cliff. voice-verified

Uh and so the deviation is just a cosine difference of the vectors or the activations, but the deviation from expectation is a very measurable thing. voice-verified

So it you know, so you have a decomposition approach, which Vlad was mentioning, you just don't have a kind of a a relative distance. voice-verified

It's from expectation approach, and there's probably ten other approaches we're not thinking of right now that collectively can absolutely be quantified and certified as safe, not safe. voice-verified

You know, when you donate stock, you know, appreciated stock, you don't pay capital gains. voice-verified

You just get the full value of the stock and you donate it. voice-verified

Showing 101–120 of 31,926 · page 6 of 1597