Has anyone tried testing GenAI on the halting problem and comparing to human performance? E.g. generate random programs where the halting status is known, and test how effectively GenAI can determine the halting status compared with humans? This seems like the ultimate, and simplest, AGI benchmark. If an AI cannot solve the halting problem as well as humans then it certainly is not an AGI. The halting problem is a level playing field between...
Originally asked by yters on Stack Exchange (ai) (CC BY-SA 4.0).
