1

Topic: Neural networks vs programs

https://www.iguides.ru/main/other/neyro … bucheniya/ Belonging Google company DeepMind told about creation  AlphaZero which all for 4 hours could be trained independently in game in chess so well that managed to win the operating world champion - computer program Stockfish. Any neural networks play Starcraft; somebody knows - they play better than initial algorithms from developers of game? Probably, I was mistaken a little, when thought that neural networks are a little general-purpose. It turns out,  as a result hammer in any algorithms, and programmers do not become necessary? Still a question - it is how much difficult to train a neural network? It can be not less difficult, than write the program?

2

Re: Neural networks vs programs

Hello, Khimik, you wrote: K> it is possible, I was mistaken a little, when thought that neural networks are a little general-purpose. It turns out,  as a result hammer in any algorithms, and programmers do not become necessary?  are not effective there where there are good algorithms. We tell, it is senseless to use  for syntactic analysis  the code. Besides, serious  very much . I.e., at least while, in a pocket  not .

3

Re: Neural networks vs programs

Hello, Khimik, you wrote: K> it is possible, I was mistaken a little, when thought that neural networks are a little general-purpose. Like  are not less general-purpose, than .

4

Re: Neural networks vs programs

Hello, Sharov, you wrote: K>> it is possible, I was mistaken a little, when thought that neural networks are a little general-purpose. S> like  are not less general-purpose, than . What is ?

5

Re: Neural networks vs programs

Hello, xma, you wrote: xma> Hello, Sharov, you wrote: K>>> it is possible, I was mistaken a little, when thought that neural networks are a little general-purpose. S>> like  are not less general-purpose, than . xma> that is ? The machine of Turing.