Cool writeup, Rob, thanks! It would be interesting to hear more about how the normal range of inputs might be thought to differ from the range of all possible inputs for some sample cognitive function, and how this promises to deal with intractability. I'm kind of surprised by the idea that you can reliably get such big reductions in computational complexity without sacrificing (too much) performance on mundane tasks -- but maybe that's just my weak theoretical computer science talking.
Cool writeup, Rob, thanks! It would be interesting to hear more about how the normal range of inputs might be thought to differ from the range of all possible inputs for some sample cognitive function, and how this promises to deal with intractability. I'm kind of surprised by the idea that you can reliably get such big reductions in computational complexity without sacrificing (too much) performance on mundane tasks -- but maybe that's just my weak theoretical computer science talking.