Google’s DeepMind artificial intelligence becomes ‘highly aggressive’ when stressed. Skynet, anyone? – Mirror.co.uk

Posted: February 15, 2017 at 9:20 pm

Google's DeepMind is one of the most famous examples of artificial intelligence.

Last year it famously defeated the world's best Go player at the tricky Chinese board game. It's also being used at Moorfields Eye Hospital to recognise eye diseases from scans.

But new research shows that DeepMind reacts to social situations in a similar way to a human. Notably, it started to act in an "aggressive manner" when put under pressure.

Google's computer scientists ran 40 million different turns of Gathering a fruit-gathering video game that asked two different DeepMind participants to compete against each other to collect the most apples.

When there were enough apples to share, the two computer combatants were fine - efficiently collecting the virtual fruit. But as soon as the resources became scarce, the two agents became aggressive and tried to knock each other out of the game and steal the apples.

The video below shows the process - with the DeepMind "gamers" represented in red and blue while the apples are green. The laser beams are yellow - and while the combatants don't get any reward for a hit, it does knock the opponent out of the game for a set period of time.

Video Unavailable

Click to play Tap to play

Play now

"We characterize how learned behavior in each domain changes as a function of environmental factors including resource abundance," the team wrote in a paper explaining their results.

"Our experiments show how conflict can emerge from competition over shared resources and shed light on how the sequential nature of real world social dilemmas affects cooperation.

"We noted that the policies learned in environments with low abundance or high conflict-cost were highly aggressive while the policies learned with high abundance or low conflict cost were less aggressive. That is, the Gathering game predicts that conflict may emerge from competition for scarce resources, but is less likely to emerge when resources are plentiful."

The results are interesting in that they show computers are able to adapt to situations and modify their behaviour accordingly.

Many experts have warned of the dangers of true artificial intelligence in machines. Elon Musk singled out DeepMind in particular as one to keep an eye on.

"The pace of progress in artificial intelligence (I'm not referring to narrow AI) is incredibly fast. Unless you have direct exposure to groups like DeepMind, you have no idea how fast it is growing at a pace close to exponential," he wrote in 2014.

"I am not alone in thinking we should be worried."

"The leading AI companies have taken great steps to ensure safety. They recognize the danger, but believe that they can shape and control the digital superintelligences and prevent bad ones from escaping into the Internet. That remains to be seen..."

So while Google's super-smart computers may be content to beat each other up in a race to collect virtual apples, the prospects for the future could be worrying. Especially if your name's Sarah Connor.

Video Unavailable

Click to play Tap to play

Play now

Video will play in

See original here:

Google's DeepMind artificial intelligence becomes 'highly aggressive' when stressed. Skynet, anyone? - Mirror.co.uk

Related Posts