4 ms·
From one of the comments: >"Professional heavy math people are those who said in the 60s that the perceptron's limitations proved all AI was impossible. And in
by nonbel 9y ago
From one of the comments:
>"Professional heavy math people are those who said in the 60s that the perceptron's limitations proved all AI was impossible. And in the 90s that one hidden layer was all you needed, deep learning was useless."
Can anyone provide the citations for this? I was aware of the latter but not the first one. You can find people still repeating the one layer stuff up to a few years ago just by reading stackexchange.
- mannykannot 9y agoI wonder if the author is referring to the oft-repeated claim that Minsky's and Papert's proof that a perceptron cannot learn the Xor function had a chilling effect on research into neural networks generally, even though Minsky and Papert themselves had shown that multi-layer networks were capable of doing so [1][2]. I realize that even this alleged misunderstanding is not the same as a claim that AI is impossible. The closest attempt of a mathematical proof of the impossibility of AI that I am aware of is the Lucas-Penrose argument from Gödel's first incompleteness theorem [3]. [1] https://en.wikipedia.org/wiki/Perceptrons_(book) https://en.wikipedia.org/wiki/Perceptrons_(book) [2] Minsky M. L. and Papert S. A. 1969. Perceptrons. Cambridge, MA: MIT Press. [3] http://www.iep.utm.edu/lp-argue/ http://www.iep.utm.edu/lp-argue/
- nonbel 9y agoThanks, the first ref looks like it may be the one.
- marcosdumay 9y agoBesides, heavy math people were also those who proved that recurrent multi-level neural networks were Turing complete and pushed the field up from the ground. But sorry, I don't have citations. It's stuff I've read a long time ago, in random books picked at library shelves by looking at the covers :)