Yann LeCun's insightful comments on Neil's post: "A very pointed and sensible critique by the prolific Neil Lawrence of Nick Bostrom's book Superintelligence.I think Neil's piece expresses a lot of the uneasiness (or frustration) many of us, down in the trenches of AI research, have been feeling about Bostrom's book. There are limits to intelligence. There are limits to the predictive abilities of even tremendously superior intelligence (due to partial observability, chaotic behavior, or sheer randomness)."

I especially agree with the limits due to limited information and chaotic behavior. All our work in AI shows diminishing returns. The problem gets harder and harder as predictive accuracy increases, and additional information becomes more and more difficult to acquire and less and less effective. Yes, big data helps, especially over the last decade or so. But big data has diminishing returns too, and there's only so much good data available.
Shared publiclyView activity