Google now has an artificial intelligence it calls “Agent” that has learned how to play retro videogames according to an article on The Guardian.
And not just in basic terms either: Agent has picked up the core concepts of 49 retro games like Space Invaders and Breakout and even developed its own strategies on how best to score big without human intervention.
Watch it get better at Breakout:
Essentially, Agent learns through experimentation and feedback, very similarly to how a baby learns anything. The more it experiments, the more it learns and the better it gets at the given task.
Thanks to this powerful breakthrough, Agent has not only learned how to play Breakout, it’s developed its own strategies on how to score the most points. It didn’t happen overnight; The Guardian says Agent took 600 games of Breakout to work out its most effective strategy, but that’s not really much different to how I learned about videogames as a kid, or how I play these days.
Read that again. An artificial intelligence algorithm has learned enough about a game to come up with a winning strategy, all on its own. Its creators simply set the rules, let it run and watched it evolve from clumsy beginnings into a Breakout master, displaying insights into the game’s mechanics that seem almost human.
DeepMind’s Demis Hassabis says it’s the first step towards creating a general learning system that could one day be used to tackle challenging problems that even humans battle with.
What it won’t do, though, is lead to the rather “Hollywood” idea of the rise of smart machines that will some day become sentient and decide humanity must go. At least, that’s what DeepMind’s researchers believe. They say they’re aware of the potential dangers posed by genuine artificial intelligence, but that we’re “decades away from any sort of technology that we need to worry about”.
Agent is the result of Google’s work with artificial intelligence outfit DeepMind, which you may recall the company acquired back in 2014 for $500 million.
While I like what I am seeing and feel quite excited about the potential applications of the technology, I must admit I’m firmly in the camp that would like to see measures put in place to guarantee AI development that won’t lead to our obliteration.
I’m not alone, either, with prominent scientific luminaries like Elon Musk and Stephen Hawking calling for the same thing.
What do you think? Is AI a genuine threat to humanity, or is it nothing more than paranoia born of Hollywood-inspired imaginations?
[Source – The Guardian]