Open AI co-founder on AGI



Here's a talk where Open AI co-founder Ilya Sutskever talks about the issue of AI progress and AGI. (Before the timestamp he was talking about Open AI's specific projects such as Dota 2 and their hand robot).

tl;dw: (italics indicate my interpolations):

  • AGI in the near-medium term can't be ruled out based on current rates of progress.
  • AI progress is more constrained by hardware than by conceptual advances (partly because it's hard to make conceptual advances if you don't have the hardware to test them with).
  • Companies and researchers are willing to make increasingly large investments into compute hardware (at rate greater than Moore's Law).
  • Open AI is also working on what we would consider Friendliness (but seemingly not at the same level of rigour as MIRI).