1. 8

  2. 1

    I don’t see why this would be relevant for general intelligence - humans are also extremely mediocre Turing Machine executors.

    Usually the claim for Turing completeness is marshalled against people who state that neural networks fundamentally cannot perform intelligence - which would then require that intelligence is not computable. I don’t think I’ve seen anyone state that NNs are good TM evaluators.

    That said, I’d be interested to hear in which direction events since 2017 have surprised the author.