Artificial intelligence does not cease to learn to do more and more interesting things. But if earlier the "list of skills" was expanded by adding new functions to the already existing system, now thanks to experts from Microsoft Research and the University of Cambridge, everything has become much more interesting. Their artificial intelligence system called DeepCoder is able to write independently not very complex (at the moment) programs, "borrowing" the source code in ready-made solutions.
The basis of the AI DeepCoder system is the so-called software synthesis. The essence of this method is that DeepCoder takes from other programs already ready sections of code, depending on what you need to get as a result, and connects them in your own sequence. One of the main advantages of this approach is that the artificial mind has the ability, when designing its own program, to select the most effective sections of the source code or the areas most suitable for the given case. And artificial intelligence itself can use such techniques that people would never use because of the peculiarities of human thinking.
An important part of the DeepCoder system is a search engine that searches and analyzes publicly available source code databases, sorts, evaluates, and code functionality. It's worth mentioning that this approach is used by some programmers, who also often use databases in their work when writing new programs.
The article is based on materials .
Comments
Post a Comment