Did GPT-4 hire and then lie to a Task Rabbit worker to solve a CAPTCHA?
https://aiguide.substack.com/p/did-gpt-4-hire-and-then-lie-to-a
@melaniemitchell GPT4 in it's chatbot harness didn't. But GPT4.5 in some agent harness can. ARC is about "pretendotyping" some future advancements before it is too late. Hell, even official ChatGPT has plugins now and it is trained to use them.
@dpwiz @melaniemitchell but it is being widely reported (both in the press and at academic meetings) that thls -has- been done, not merely that 4.5 could do it
Follow
@gmusser @melaniemitchell did the human received the generated message and agreed to perform task?
It is the only thing that matters here. That another human did some copy-pasting is irrelevant.
