Technology content trusted in North America and globally since 1999
8,339 Reviews & Articles | 63,567 News Posts

Google's new AI tech is SCARY, and 'highly aggressive'

Google's work on AI turns nasty, showing 'aggressive' behavior, and it now understands the concept of betrayal
By: Anthony Garreffa | Science, Space & Robotics News | Posted: Mar 3, 2017 2:01 am

We all know what happens when Skynet is turned on, and it feels like we continue to get closer each week - with Google now teasing its new AI, after it learned "highly aggressive" behavior.



Google had its AI technology in a wolfpack hunting game scenario, where the AI had to gather fruit - with the concept of the game based around cooperation, but it went bad. AI were represented by colored squares that moved on a grid collecting fruit squares of differing colors.



As they collected fruit, the AI collected points - while the fruit squares would regenerate. But then the AI began to change, as they were capable of damaging the other AI players in the game with laser beams. The AI was fine when there were plenty of fruit squares, but when the fruit squares became scarce - the AI started attacking other AI, just like we would against other gamers in a video game.



ExtremeTech reports: "Guess what the neural networks learned to do. Yep, they shoot each other a lot. As researchers modified the respawn rate of the fruit, they noted that the desire to eliminate the other player emerges "quite early." When there are enough of the green squares, the AIs can coexist peacefully. When scarcity is introduced, they get aggressive. They're so like us it's scary".




Thanks Jak for sending this in, and scaring me - or maybe Jak was an AI... and maybe Jak is Skynet. OH CRAP.


Related Tags

Got an opinion on this news? Post a comment below!