Chaining the evil genie: why "outer" AI safety is probably easyView in threadMitchell Porter3y100One can play this game - adding constraints to make a goal safer - with ChatGPT! https://pastebin.com/eQqtPMSr It shouldn't be hard to automate this process, via a GPT-based agent similar to Auto-GPT. Reply
One can play this game - adding constraints to make a goal safer - with ChatGPT!
https://pastebin.com/eQqtPMSr
It shouldn't be hard to automate this process, via a GPT-based agent similar to Auto-GPT.