Published on July 31, 2023 4:04 PM GMTCore claim
I believe that sharing information about the capabilities and limits of existing ML systems, and especially language model agents, significantly reduces risks from powerful AI—despite the fact that such information may increase the...