Deepmind unveils revolutionary AI

One small step towards a general artificial intelligence

Many of you may remember when Deepmind, Google's sister company, developed an artificial intelligence algorithm in 2015 capable of learning and playing 57 Atari games at superhuman levels. The only thing programmers told the AI was that it should get the highest score possible so, through a series of countless trials and errors, it learned how to play the games masterfully. Check the video below for a walk down memory lane.

In 2016, another algorithm from Deepmind, called AlphaGo, defeated the world champion Lee Sedol at the game of Go.

Go, a super popular board game played mostly in Asia, offers many layers of complexity on top of chess and many thought an AI would never be able to win over a top human player before 2030.

The achievement was a watershed moment for artificial intelligence and showed that an AI could have an intuition similar to humans. You can learn more details about AlphaGo and the historical significance of a computer beating the top human players by watching this fantastic documentary showing how the Deepmind team was able to pull it off.

The new breakthrough

Fast forward to 2020 and Deepmind pulled another surprise with an algorithm called MuZero, which is capable of learning and playing Go, Chess, Shogi and Atari games with superhuman capabilities WITHOUT knowing the rules of each game.

This is a significant development for artificial intelligence because, like humans, the algorithm is capable of learning the dynamics of an unkown environment by itself and without any help. It is a small step in the direction of creating general purpose algorithms which would replicate human intelligence and, thus, give birth to artificial general intelligence (AGI).

Creating an AGI program, which technically would be more intelligent than a human, could be a point of no return for the human race. In a few days, an AGI could harvest the computing power of the whole Internet, reprogram itself and become millions of times more intelligent than all the humans who ever lived.

Why should you care about this?

We are still decades away from a possible AGI outcome. Right now, no one knows what would happen when computers become more intelligent than we are. But we understand it probably will be humanity's last invention.

With an AGI program, humans could either expect dystopic outcomes such as the one depicted in the movie Terminator or solve humanity's greatest problems and challenges such as diseases, climate change, quantum physics, aging or poverty.

For students interested in computer science, ethics, law or even politics, the field of Artificial General Intelligence is fascinating and will require the brightest minds and the purest hearts in order to help our civilization to develop this technology in a safe and useful way.

Leave a comment