Understanding AI World Models w/ Chris Canal
post by jacobhaimes · 2025-01-27T16:32:47.724Z · LW · GW · 0 commentsThis is a link post for https://kairos.fm/muckraikers/e011/
Contents
No comments
On our latest episode of muckrAIkers we are joined by our first ever guest, Chris Canal!
You can find the show on Spotify, Apple Podcasts, YouTube, or wherever else you listen.
Chris is the co-founder of EquiStamp, a third-party evaluator providing objective and accurate evaluations of frontier LLMs and the systems built off of them. The group has been a key contractor for METR, providing baselines for their recent work RE-Bench, which evaluates frontier LLM capabilities in AI R&D.
In this ~3 hour long episode on understanding AI world models, we get into test-time compute, the moving goalposts of “AGI,” and do a deep dive on the orthogonality thesis.
EquiStamp is hiring, so if you're a software developer interested in a fully remote opportunity with flexible working hours, join the EquiStamp Discord server and message Chris directly; oh, and let him know muckrAIkers sent you!
As always, all sources referenced throughout the episode can be found on the showpage at https://kairos.fm/muckraikers/e011
Given the imprecise sensationalism that permeates AI discourse, it is important to have a critical perspective when reading press releases, news coverage, and even academic papers about AI. On muckrAIkers, Dr. Igor Krawczuk and I (Jacob Haimes) contextualize these developments, separating muck from meaning, allowing all of us to better understand the motivations and likely next steps of major players in the AI space.
muckrAIkers is a Kairos.fm production
0 comments
Comments sorted by top scores.