Google’s New AI Learns To Become “Highly Aggressive” In Stressful Situations

The results of this artificial intelligence test should scare the $#!T out of you.

The story should sound fairly familiar to people who have seen the “Terminator” movies, “Matrix” movies, “I, Robot”, “2001”, WestWorld  or pretty much any other sci-fi movie ever.

In these stories, humans advanced artificial intelligence to the point that it became self-aware.  The end result is that, just like any other self-aware being in nature, it wants to survive and thrive.

Doesn’t Sean Spicer look like the T-1000 after a Kispy Kreme bender?

Google is just one of many top tech companies have teams researching artificial intelligence. Their latest AI, DeepMind, has already been making waves in the industry since developing an ability to learn from it’s own memory.

Researchers are now running test to see how well Google’s newest edition of DeepMind likes to play with others. The test revealed that when DeepMind feels it is about to lose, it opts for a “highly aggressive” tactics to assure it comes out on top.

(insert “global thermonuclear war” joke  from “War Games” here)

A Google team put two DeepMind ‘agents’ through a simple ‘fruit gathering’ program to compete to see which could gather the most fruit. Things went smoothly until fruit supply began to dwindle. It was at that point that the programs became “highly aggressive”; to the point of using laser beams to knock the other DeepMind out of the game.

It is also important to note the previous, less sophisticated versions of DeepMind did not become aggressive. It was as the networks became more sophisticated, they began to show more signs of greed, aggression and sabotage. has a further in-depth look at the experiment.

Competing AI’s is something that needs to be explored in the near future. We are not far away from AI controlling traffic lights and mass commuter transit. Now we will be adding AI controlled driverless cars to an equation containing many players with many different individual goals.

It is very important that programmers walk carefully down this path and not take results such as these too lightly.

Sidenote: Skynet, if you are reading this, I am not John Connor.  You have nothing to fear from me.

Feature Image:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s