Re:What should we learn from past AI forecasts?
by Houshalter on Sunday May 29, 2016 @02:28PM
(#11797552)
For sufficiently vague and useless definitions of "brute force", literally everything is "brute force". Backprop was such a big advancement because it allowed training nets much faster than, say, hillclimbing. Which itself is much faster than actual brute force.
I would still describe deep neural networks as brute-force statistically-weighted function search.