AIs often refuse open math problems, citing low odds of success
Jsevillamol · x · 2026-11-22
Jsevillamol reports that when prompting AI to solve open math problems earlier this year, models frequently refused to even attempt them, reasoning that since the problem is open, success was unlikely. He hopes the recent Riemann hypothesis achievement gives the models the confidence they were lacking.
More from Fun
- Ed Zitron becomes the face of AI bubble skepticism, sparking an X spat — inductionheads · 2026-11-22
- 'Call me when Claude brings back the dodo': banter after Riemann hypothesis progress — niloofar_mire · 2026-11-22
- AI-generated 'When a Father Is Born' clip moves viewers, mangled hands betray it — Lucky-Gur6416 · 2026-11-22
- Dev greets Claude with 'hello beloved claude,' jokes he looks crazy — Aryvyo · 2026-11-22
- AI safety drama: clarifying which 'doomers' you mean — and why not Yudkowsky — repligate · 2026-11-22
- Comic skewers AI's 'false sense of achievement': throw ball, cheer, repeat — ThomasSimonini · 2026-11-22