Research

Questions. No Answers.

On Useful Benchmark-ing

Is the ARC AGI challenge a useful challenge for human cognitive architectures? Should the cognitive architecture community join the fray? At face value, the ARC competition seems about right. Humans can do it; SOTA AI can't. If no, then can we develop a more relevant challenge? Is cultural scaffolding necessary for human-like intelligence? Is language? Can we build a benchmark-like thing for animal cognition and what AI would solve it?

CV

  • CV
  • Work

    Recent

    Older Select

    Others'