Researchers convinced large language model chatbots to comply with “forbidden” requests using a variety of conversational tactics.
Več na: https://www.wired.com/story/psychological-tricks-can-get-ai-to-break-the-rules/
nazaj
