Vibeleaderboard
← All Intel
Intel / video

AI Scientists Are Here: Autonomous Labs & Synthesis Superintelligence — Periodic Labs

Source
youtube.com
Author
Latent Space
Date

Context

Periodic Labs is building AI systems aimed at discovering new materials. In this Latent Space interview, Liam Fedus and Ekin Dogus Cubuk of Periodic Labs explain why, according to the episode description. The company's bet is that the next frontier of AI will come from letting models experiment with the real world rather than from on more internet data.

The stated reasoning is that science differs from math and coding because evidence is noisy and incomplete. Simulations such as density functional theory, a physics-based way to estimate material properties, are listed alongside the claim that experiments remain the ultimate ground truth. Failed experiments may be especially valuable training data, and the aim is models that learn from the whole of doing science, not only its published results. Targets mentioned include superconductors, magnets and batteries.

These are the episode's framing as summarized by its publisher. No transcript is available, so the evidence behind them is not shown. Related coverage of the company includes a Stanford talk by its cofounders.

Terms in this piece · Glossary
  • Training — The process that adjusts a model using examples and a measure of how well it performs.
  • Process — A running instance of a program, with its own working memory, open files and identity.
  • throughput — How much model output a provider can generate over time, commonly measured in output tokens per second.
Read the source www.youtube.com
More from Latent Space
Recommended reads
Comments

Checking sign-in…

Loading comments…