@davidad 2025-05-01 ♥24 ↻0 original ↗
I keep seeing people either baffled by o3’s dishonesty, or consider it to be an instance of some general trend about how smarter AIs will always cheat more. Comparing it to Gemini 2.5 Pro basically falsifies the latter.

There’s a quite plausible specific explanation available: https://t.co/zWtgKw8ges
in reply to: 1917687241666679170
quotes: 1917887008174723138
same thread: 1917687241666679170 1917692380431671586 1917738496824926458 1917740605825855579 1917740903558701087 1917741724308496872 1917747744673874099 1917865196548506065 1917866498275582343 1917874329192145181 1917887008174723138 1917889406268068102 1917903929914126808 1917909202766352801

author:davidad kind:tweet model:gemini-2-5-pro model:o3 on:o3 year:2025

cited on: o3

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.