Alphabet reports that its DeepMind Technologies AI unit has successfully learned how to play 57 Atari video games. And the computer system plays better than any human.
Atari, creator of Pong, one of the first successful video games of the 1970s, went on to popularize many of the great early classic video games into the 1990s. Video games are commonly used with AI projects because they challenge algorithms to navigate increasingly complex paths and options, all while encountering changing scenarios, threats and rewards.
Dubbed AGENT57, Alphabet’s AI system probed 57 leading Atari games covering a huge range of difficulty levels and varying strategies of success.
“Games are an excellent testing ground for building adaptive algorithms,” the researchers said in a report on the DeepMind blog page. “They provide a rich suite of tasks which players must develop sophisticated behavioral strategies to master, but they also provide an easy progress metric —game score—to optimize against.
Read More at Tech Xplore
Read the rest at Tech Xplore