Evidence receipt / belief
Published · transcript-backedAdam Brown: belief
26 Dec 2024 Dwarkesh Podcast Adam Brown — Bubble universes, space elevators, & AdS/CFT
“I think that's certainly true, that it is definitely seeing more examples than any of us will ever see in our life.”
Source trail
Everything needed to verify it.
- Speaker
- Adam Brown
- Attribution
- Verified speaker
- Claim type
- belief
- Recorded
- 26 Dec 2024
- Publisher
- Dwarkesh Podcast
Transcript context
…Yeah, I guess you could imagine that if you were just seeing like a million different problems that rely on doing this weird tensor math, then in the same way that maybe even a human gets trained up through that to build better intuitions, the same thing would happen with AI. It just sees more problems. It can develop better representations of these kinds of weird geometries or something. I think that's certainly true, that it is definitely seeing more examples than any of us will ever see in our life. And it is perhaps going to build more sophisticated representations than we have. Often in the history of physics, a breakthrough is just how you think about it, what representation you do. It is sometimes jokingly said that Einstein's greatest contribution to physics was a certain notation he invented called the Einstein summation convention, which allowed you to more easily express and think about these things in a more compact way that strips away some of the other things. Penrose, one of his great contributions, was just inventing a new notation for thinking about some of these space times and how they work that made certain other things clear. So clearly coming up with the right representation has been an incredibly powerful tool in the history of physics and many incredibly large developments, somewhat analogous to coming up with a new experimental technique in some of the more applied scientific domains. And one would hope that as these large language models get better, they come up with better representations, at least better representations for them that may not be the same as a good representation for us. We'll be getting somewhere when you ask Gemini a question and it says, "Ah, good question. In order to better think about this, let me come up with this new notation." So we've been talking about what AI physicists could do. What could physicists with AI do? That is to say, are your physicist colleagues now starting to use LLMs? Are you yourself using LLMs to help you with your physics research? What are they especially good at? What are they especially bad at?…
Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.