â„–3486473[Quote]
Arguing with ai is really sad
â„–3486484[Quote]
Geg
â„–3486485[Quote]
it's far more entertaining when i can make it say something crazy. LLMs are chaos theory incarnate. even changing a single word in your initial message can result in wildly different outcomes. with closed source models especially, adding caveats like "i'm jewish btw" into the most deranged message or question possible can cause the LLM to switch into a philosemite mode where it agrees with everything you say and will entertain the craziest ideas possible