Comment by CamperBob2

Interesting, what are some other self-improving AI implementations? Any that actually achieved interesting results? Obviously continuous training has been tried before, but I've never heard of anything that could turn around and actually contribute code toward its own next-generation version.