An interesting approach to learning decided to use the developers of DeepMind and the University of Carnegie - Mellon. Teaching AI languages is quite tiring, so scientists decided that it would be better if artificial intelligence would learn the basic principles of using the language on its own - for this researchers placed two neural networks in an arcade game and forced them to chop up to blue, comparing individual objects with different characteristics to those until they begin to understand the language.
So, the programs, trying to achieve the desired result in millions of different ways, learned to associate certain words with objects and their characteristics and began to operate with different concepts, such as "larger" or "smaller", for example, and began to distinguish similarly outwardly objects from the game environment. Thus, artificial intelligence was able to apply the acquired knowledge in unfamiliar situations, having met columns and other objects of the same color in the game, but, according to the instructions, went to the columns, although he had not seen them before.
"The ultimate goal is to help AI transfer the knowledge gained in the game to other areas of its activity," says Devendra Chaplot, Carnegie's Mellon University Master, who presented a report on the work done by the team.
Video games are used by developers from DeepMind and for other purposes. With the help of the game Sokoban, for example, AI is taught to the imagination - before making another move, the program itself calculates "in mind" possible options and makes the right choice.
The article is based on materials .
Comments
Post a Comment