ICML Keynote on Automated Science

math_rachel · x · 2026-07-15

This post shares the slides and full transcript of an invited ICML 2026 keynote titled "What will be left for us to work on?", highly recommending a full read.

A core takeaway is that viewing "automated science" merely as problem-solving is a fundamental misunderstanding. The goal of science isn't just to get tasks done; it's to form human understanding. The author emphasizes that human understanding is not friction to be eliminated by automation—it is an indispensable part of science itself. Without it, we lose all the derivative value it brings.

The keynote also distinguishes between "recursive self-improvement, human-like AI, superintelligence, and economically transformative AI," arguing that these concepts do not inherently imply one another.

Related event: random_walker Releases ICML 2026 Keynote: What Will Be Left for Us to Work On?(10 posts)→

Original post →

More from AGI Musings

AGI Musings channel →