5 ms·
I disagree with you. On the contrary, I think it can only be solved by throwing more data, but much more data that we have available now, to get first a smart A
by elcomet 4y ago
I disagree with you. On the contrary, I think it can only be solved by throwing more data, but much more data that we have available now, to get first a smart AI that has common sense about the world (a world model). Then we can start trying to drive in it.
- dreamcompiler 4y agoThe counterexample to your point is that brand new 16-year-old human drivers seem to do a pretty good job of driving with very little training data. Yes, they do kill people and they're much worse drivers than 40-year-olds who have much more training data, but the fact that 16YOs kill as few people as they do with so little training data means some other mechanism than raw data quantity is important.
- atonse 4y agoWhat do you mean by little data though? They’ve been absorbing years of data classifying traffic and things like that.
- elcomet 4y agoIt's not a counterexample. Your 16-year old human has millions of years of evolution into its brain. That's a lot of pre-training data.
- dreamcompiler 4y agoAre you suggesting that orange cones and semi-trailers cutting you off were common on the African savannah?
- elcomet 4y agoI'm suggesting all of this is just finetuning a very generic and powerful object recognition and tracking ability of the brain