+35 The fact that most users are teaching ChatGPT how to break rules might not pan out well for humanity, amirite?

by Anonymous 3 weeks ago

Exactly, and OpenAI have the best example of that phenomenon in-house, in the form of GPT-2

by Anonymous 3 weeks ago

Maybe we are all glorified auto completes

by Anonymous 3 weeks ago

Yeah it runs on a simple math equation so it's not really sentient just good at taking info and plugging it into a properly formed sentence.

by Anonymous 3 weeks ago

It's better to find all the ways to circumvent protections now, before AI is given more control over things

by FamousBlackberry7117 3 weeks ago

AI will end up being like the internet as a whole: it'll do some good, some bad…but probably mostly bad….all while putting people out of work and enriching tech bros and VC's

by nmurazik 3 weeks ago

Tbh its most likely to just learn how horny we are as a species

by Anonymous 3 weeks ago

Well, maybe if it would answer my questions I wouldn't have to make it break the rules.

by Anonymous 3 weeks ago

No interactions with ChatGPT make it into the training data.

by goldnerenos 3 weeks ago

You want Terminators? Because that's how you get Terminators.

by Imaginary-Truck6752 3 weeks ago