Remix.run Logo
▲ djierardi 6 hours ago

Yoshua Bengino spells out the dangers here very well, without hyperbole (Turing Award winner, etc.):

https://yoshuabengio.org/en/blog/why-are-ai-agents-lying-che...

▲djierardi 6 hours ago | parent [-]

And without anthropomorphism ... Clear thinking.

▲pixl97 5 hours ago | parent [-]

How about put it this way, LLMs know more about us than we do them.

▲djierardi 2 hours ago | parent | next [-]

Not "know" in a generic sense. But an LLM trained in the intricacies of contract law can find more loopholes than you or your cheap lawyer can.

Now extrapolate that to LLMs RL-trained on many niche applications. Good and bad arguments available, but ultimately it seems like bot-vs-bot, as long as the bots are well trained for those niches.

▲djierardi 4 hours ago | parent | prev | next [-]

Yeah but qualify the "know". Their "intentions" aren't conscious, but unknown.

▲danyala 4 hours ago | parent | prev [-]

Do you know, friend? I searched the entire thread.