Google’s New AI to Become Highly Aggressive under Stressful Situations

Google’s New AI to Become Highly Aggressive under Stressful Situations

Intro

Around October last year, Stephen Hawking warned that the continued advancement of artificial intelligence will either be the worst thing or the best thing that will ever happen to humanity, and recent developments may actually point in that direction. Skynet, the self-aware AI system we all see in Terminator movies is a prime example of what AI can actually do to humanity.

While testing Google’s DeepMind AI system last year, we came to realize that artificial intelligence systems have an ability to learn independently from their own memories, making it possible to beat even the smartest people at their own game.

Researchers have been testing how willing DeepMind is to cooperate with others, and the tests have revealed that whenever it feels stressed, whenever it feels like it’s going to fall behind, it becomes very aggressive just to ensure that you don’t beat it.

The researchers ran over 40 million turns of a fruit gathering computer game that requires that two DeepMind agents compete against each other to get as many apples as they could possibly gather. It was discovered that as long as there were enough apples to go round, things went pretty smoothly, but as soon as the number of apples reduced, they resorted to knocking each other out of the game using laser beams just so that only one of them gets all the apples.

Google’s New AI to Become Highly Aggressive under Stressful Situations



TweakingTricks.com is 100% independently owned and welcomes new article contributions for our growing community. Our sole purpose is to spread news and research to provide deep insight on disruptive companies, people and trends.


© 2016 - TWEAKINGTRICKS.com | Privacy Policy