In theory, can everyday computing tasks be broken down into those that can be solved by a neural network?
An MIT Review recently published this article about a chip from IBM that is more or less an artificial neural network. Why IBM's New Brain Chip May Be Historic | MIT Technology Overview
The article suggests that the chip may have borrowed a page from the future. This could be the beginning of an era of new and advanced computing power. It also talks about programming the chip.
One drawback is that the IBM chip requires a completely new approach to programming. Although the company announced a set of coding tools for the upcoming chip last year (see "IBM Scientists Show Blueprints for Brain Computing"), even the best programmers find training in bruising, Modha says: "It's almost always frustrating." His team is working to create a library of ready-made blocks of code to make the process easier.
Which brings me to the question, can everyday computing tasks be broken down into those that are solved by a neural network (theoretically and / or practically)?
source to share
It depends on the task. There are many tasks for which John von Neumann's computers are good enough. For example, calculate the exact values of functions in a certain range or, for example, apply some filter on the image or save the text in db, read it or save prices for some products, etc. This is not an area where NNs are needed. Of course, it is theoretically possible to train NN to choose where and how to store data. Or teach him accounting. But for cases when it is necessary to analyze a large data set, and the existing methods cannot be considered in NN. Either speech recognition, or the purchase of some paintings that will become masterpieces in the future, can be done with NN.
source to share