Nat Lambert: using models for slides and plans makes AGI feel far away
giffmana · x · 2026-10-19
Nat Lambert notes that pushing models beyond coding/ML — making slides, coherent plans, or longish content design — makes him feel "AGI is going to take a while." giffmana replies that it boils down to the models' lack of empathy with the recipient's state of mind. A short but telling exchange on current models' weaknesses in open-ended knowledge work.
Related event: Nathan Lambert: Models Still Far From AGI Outside Coding(2 posts)→
More from AGI Musings
- Multitasking with agents? Slow down and focus on your P0, says Anthropic cofounder — trq212 · 2026-10-20
- MIRI's Nate Soares: I'm 100% OK With Ignoring Everything AI Leaders Say — notkilleveryoneist · 2026-10-20
- Author calls for freezing frontier AI to handle harms already here, in new book Obsolete — GarrisonLovely · 2026-10-20
- Same superintelligence, wildly different outcomes: judgment is earned by daily use — McDonaghMatthew · 2026-10-20
- Andrew Chen mocks AI doomers buying $5M+ SF homes while predicting doom — andrewchen · 2026-10-20
- Cosmos Institute: Skills from verifiable domains like math and code transfer to writing — lawhsw · 2026-10-20