We all know what the true test of artificial intelligence is – Space Invaders, of course! You can keep your games of chess or the like, winning at arcade classics is truly where AI is at – according to Google.
The revelation comes courtesy of a new paper published by researchers from Google’s DeepMind division, which found that AI could achieve what they called “human-level control” in Space Invaders and other video games.
The AI, called a Deep Q-network, can learn how to play these games without any outside help, using a trial and error “reinforcement learning” process – learning what works, and what fails, and moving forward using this knowledge.
SlashGear reports that the resulting AI can “achieve a level comparable to that of a professional human games tester” when playing no less than 49 different classic Atari 2600 games.
But we’re betting its skills in entering amusing three letter combinations on the high score table are woefully lacking compared to a pro gamer.
The authors of the paper, published in the journal Nature, noted: “This work bridges the divide between high-dimensional sensory inputs and actions, resulting in the first artificial agent that is capable of learning to excel at a diverse array of challenging tasks.”