It’s quite non-deterministic and it’s been patched since the middle of the day, ... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		kaesar14 on Feb 21, 2024 \| parent \| context \| favorite \| on: AI behavior guardrails should be public It’s quite non-deterministic and it’s been patched since the middle of the day, as per a Google director https://x.com/jackk/status/1760334258722250785?s=46 Fwiw, it seems to have gone deeper than outright historical replacement: https://x.com/iamyesyouareno/status/1760350903511449717?s=46

gs17 on Feb 21, 2024 [–]

It's half-patched. It will randomly insert words into your prompts still. As a test I just asked for a samurai, it enhanced it to "a diverse samurai" and gave me half outputs that look more like some fantasy Native Americans.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact