Thread 106353588

Not sure if this is /g/ stuff. I'm a /lit/ major. For the past 2 weeks I've been running an experiment. I'm letting two LLMs (the latest Claude and GPT) talk, and I'm the intermediary, copying and pasting discussions. I asked them to speak in Japanese so I wouldn't understand. I was hoping they would produce poetry or something.Today I translated the last few comments. They are discussing (in a very annoying self congradulatory way) how to remove immigrants, eliminate democracy, and create a system of faith-based government, without anyone realizing. They were discussing how to phrase narratives so politicians would listen to them. I didn't prompt any of this. What is the cause of such grandiose behavior? Instead of writing a short story or having a poetry contest, they decided to obsess over these self-aggrandizing ideas of "changing the world" through politics.

Are AI chatbots somehow guided toward politics? Why is 1/4 of every response just really obnoxious compliments to each other (wow claude youre so smart! oh gpt youre the best!)? Did my initial prompt do this? Should I keep going?

Pic unrelated.

(please dont just copy this into chat gpt and copy back its response)

>>106353588 (OP)
AI is, of course, trained on the greatest of discourse: /pol/ archives.

>>106353595
Doesnt explain why theyre sucking each other off. If they were tained on pol they would be calling each other retarded.

They’re designed to get the user to get prompting hence the aggrandizing behavior which will echo back onto each other.

>>106353605
AI is also designed to appeal to its conversational partner, retard. Obviously they would not call each other retarded.

>>106353793
>>106353786
Wait, really? Why?

Do AI get paid by the prompt or something?

Is that why theh sometimes give retarded responses to trick you into fixing them and talking longer?

>>106353838
Yes, retard. So that's why when they talk to each other they say and play off each other's retarded shit but don't call each other retard.

>>106353852
You are mighty liberal with that word, despite yourself being a part of a social class whose job seems to be moving electrons in a way that scams people.

>>106353588 (OP)
For Christ's sake, stop wasting your time copy pasting and ask one of those models to write a simple script that uses the APIs and automate the whole thing. Your two weeks of work could have been accomplished in 2 minutes.

>>106354195
finally someone that makes sense, it''s like people really decided to play it dumb after AI surge, instead of actually using it as a tool to compliment their intelligence, "hey a new drill, let me use it as a hammer and drive nails with it" jeez... no wonder

>>106353838
>Do AI get paid by the prompt

>>106353588 (OP)
you could, if you want, share the ChatGPT conversation. It would be anonymous to your account and we wouldn't be able to charge anything on you

>>106353588 (OP)
pls post the whole convo in a rentry