For machine learning, put in a word
We are experiencing a boom in machine learning methods, and it may seem that there is no problem that a hundred-million-layer network would not solve if it had infinite time and an unlimited data set. Machine learning shows excellent results when working with images, speech. Neural networks can "remember" the answers to examples drawn on paper and predict where a tennis ball will fly before it touches a racket.
They say that deep learning, genetic algorithms and other new-fangled disciplines can solve problems of any complexity. And apparently this is true. Unless of course you have a lot of time and all the data of this world.
One of the most fundamental methods of machine learning is the genetic algorithm. As you undoubtedly know, the great genetic algorithm of Mother Nature worked two billion years before multicellular life appeared. Eight hundred million years later, the genetic algorithm built a species whose representative is typing this text. Do we have the patience for three billion years of computational experiments and can we afford to go that route?
Cybernetic root of evil
— . ", " , , , .
" , , ", — .
. , . , , . " ?". .
, . , , , , , . , , . , .
?
. ?
. . , . ? - ?
? , , . , . , . , . , , . . , , — , .
— . , . , — , . . . . . . , . , — .
. , . . ! , ? , - , . , .
, — . , . ? - . . , , , , . — . , . . — . . , . ? . .
, . . ? — . , . , , . . , , , , . , , , . , . , , , , . , , .
— , — , .
. , . — . , , . , , , . , , .
, ?
, . , . , . . ? , . ! , , , , . , , . .
, , ?
, , . , . ! , .
, , , , . . , , , . . — : , , . . . . . , , , , , , , , . .
— . , . . . , . , .
. . . , , , . .
[1] . , , . . .
. . . . , . , .
? , . . , . . , , , . . . — .
. , , . . .
. . , , , . . , . . , . , , . , , — .
, . , , . , . .
, . , . , . . . . , , . , , .
When constructing artificial intelligence, it should be remembered that artificial neural networks do not solve the problems that most of them solve natural neural networks. It should be remembered that learning and, ultimately, intelligence are optional, expensive, dangerous, high-level functions that are alien to mess with pixels and individual actuators. So it was hundreds of millions of years ago and to this day nothing has changed, for such is the path of cybernetics.
[1] Vyacheslav Dubynin: "Brain and movement." https://www.youtube.com/watch?v=XNMGhUG3YDA