High Signal Podcasts Evidence ledger
Method
Browse
← Back to evidence

Evidence receipt / prediction

Published · transcript-backed

Dylan Patel: prediction

2 Oct 2024 Dwarkesh Podcast @Asianometry & Dylan Patel — How the semiconductor industry actually works

“Anyway, there's a tremendous opportunity to bring breakthrough innovation simply because there are so many layers where things are unoptimized.”

— Dylan Patel

Source trail

Everything needed to verify it.

Speaker
Dylan Patel
Attribution
Verified speaker
Claim type
prediction
Recorded
2 Oct 2024
Publisher
Dwarkesh Podcast

Transcript context

…Inverse lithography, maybe. ILT and… maybe. I don't know if that's AI. Anyway, there's a tremendous opportunity to bring breakthrough innovation simply because there are so many layers where things are unoptimized. You see all these single-digit to low double-digit advantages just from RL techniques from AlphaGo-type stuff, or not AlphaGo but 5-8 year-old RL techniques being brought in. Generative AI being brought in could really revolutionize the industry, although there's a massive data problem. Can you give the possibilities here in numbers in terms of maybe like a FLOP per dollar or whatever the relevant thing is? How much do you expect in the future to come from process node improvements? How much from just how the hardware is designed because of AI? We're talking specifically for GPUs. If you had to disaggregate future improvements.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.

Search evidence