DeepMind teaches itself to play 57 video games better than humans

Published by:
Posted on: April 7, 2020 9:49 pm EDT

The artificial intelligence master continues to get smarter and more self-sustaining

The video game industry has a great advantage that it gets boosted with new emerging technologies, artificial intelligence (AI) has been part of video games since the first chess video games with a computer-controlled contender. That was a few decades ago, but now AI has been developed and specialized so much that current AI devices are easily surpassing human beings’ abilities. An AI-empowered software called DeepMind has been teaching itself to play at least 57 Atari games and has managed to play them better than any human. This is the last advancement of the AI company Alphabet, and DeepMind is so proficient at the games that it sometimes dubbed Agent 57.

This is not the first time that Atari games are being used to test new AI-software abilities. In fact, the video game industry has been the testing ground for other technologies that have emerged from AI, such as virtual reality (VR), augmented reality (AR) and machine learning (ML), to name a few. Atari’s games provide a varied skill challenge that makes it ideal to achieve reinforce learning. This is not the only video game DeepMind plays well; not so long ago it was defeating the best StarCraft players.

Not only is AI reaching human levels, but it also tested “above the human baseline” in each of the 57 Atari 2600 games. There has been a YouTube playlist released in which Agent 57 can be seen playing all those 57 games. The main goal of AI is not defeating humans, instead, comparing AI-software capacity with humans allows developers to test the advancement of AI to be applied in other areas.