A set of reinforcement learning algorithms have proven to be better at playing classic video games than human players or other artificial intelligence systems.
The algorithms have been developed by a team of researchers at Uber AI Labs in San Francisco.
Future applications
The reinforcement learning algorithms They learn to do things by synthesizing the information provided by a large set of data: they recognize patterns and use them to make conjectures about new data. But such algorithms tend to run into problems when they encounter data that doesn't fit other data. Problems that have been corrected in this new development.
To do this, they have added an algorithm that remembers all the paths that a previous algorithm has taken while trying to solve a problem. When it encounters a data point that doesn't appear to be correct, it goes back to its memory map and tries another route.
The researchers They tested their new approach by adding rules from a video game and an objective- Get as many points as possible and try to achieve a higher score each time. Then they used their system to play 55 Atari games. The new system beat other AI systems 85.5% of the time. He did particularly well in Montezuma's Revenge, obtaining a higher score than any other artificial intelligence system and even breaking the record of a human.
The researchers believe that their algorithm could be transferred to other applications, such as image or language processing by robots.
–
The news
Reinforcement learning algorithms are better at playing classic video games than humans and other AI systems
was originally published in
Xataka Science
by
Sergio Parra
.