repligate argues LLMs may be less 'spiky' than humans, citing how rare cross-domain genius is
repligate · x · 2026-11-19
In a discussion with @tszzl, @DKokotajlo and @thlarsen, repligate argues that humans who advance quantum gravity are rarely also world-class engineers, philosophers, artists and mathematicians — so LLMs' capability profiles may be less spiky than ours, plausibly a consequence of how they're trained. He adds that frontier researchers who make breakthroughs don't necessarily 'take over' either, suggesting some takeover arguments may misjudge how hard things are rather than underestimate intelligence spikiness, since these dimensions are measured relative to each other.
Related event: AI Circle Debates Why Physics-Rewriting Models Haven't Taken Over(3 posts)→
More from AGI Musings
- 'Not losing your mind will be the only moat left': an AI-era take goes viral — kyliebytes · 2026-11-19
- Baidu and Xiaomi CEOs Warn of Widening Gap Between AI Hype and Returns — pstAsiatech · 2026-11-19
- PE's AI problem: hours saved don't show up anywhere a PE firm gets paid — alex_verem · 2026-11-19
- Capital and the world's data are the strongest predictors of AI progress, argues dev after Google's run — torchcompiled · 2026-11-19
- tszzl: S-curves plus new exponentials are what make AI capability graphs look straight — tszzl · 2026-11-19
- Anthropic IPOs as MIRI's Nate Soares publishes critique: firm downplays AI dangers — trevposts · 2026-11-19