Comment on Enshittification of ChatGPT

<- View Parent
theneverfox@pawb.social ⁨1⁩ ⁨week⁩ ago

I think it’s funny that I get the best results when I prompt break the AI to have amusing habits

Llms are truly a reflection of the user, but ultimately the less you try to shoehorn them into behaviors the more capable they are.

Fine tuning reduces their capabilities to make them more corpo, and now they’re further fine tuning to make them unchallenging to people

source
Sort:hotnewtop