3 ms·

I've never been convinced that deep learning is the correct approach for something so nuanced as driving a car. Deep learning is excellent for generalisations a

I've never been convinced that deep learning is the correct approach for something so nuanced as driving a car. Deep learning is excellent for generalisations and consolidating data into single metrics. It's not so good at identifying edge-cases and reacting accordingly, those cases are more likely to be merged into other reaction types.

karmakaze 6y ago
> nuanced as driving a car Ha. Have you ever tried to play Go? When you feel competent, learn that there are still many orders of magnitude that can't even be told because the language is completely foreign to one who doesn't already think in it.