Remember earlier this year when Google’s artificial intelligence computer, AlphaGo, defeated “Go” champion Lee Sedong and shocked the world? Well, Google hasn’t stopped working at making their artificial intelligence even more intelligent. In fact, Google’s artificial intelligence computer is now teaching itself and learning beyond what humans even know themselves.
Is This The End Of The World?
Of course, all of that sounds really apocalyptic and scary until you learn that it’s still just practicing it’s Go skills. If you haven’t heard of Go, it’s a popular ancient Chinese game that pits two players against each other. The combination of moves in Go is what makes it so remarkable and such an effective way to test artificial intelligence.
AlphaGo Zero is the new version of the AI that defeated Go champion, Lee Sedong, last year. It has gone ahead and learned, without human input, how to become even better at Go. Basically, through playing older versions of itself, AlphaGo Zero has learned new moves and strategies for playing Go. When put to the test against humans, AlphaGo went a shocking 60-0. It only took 40 days for AlphaGo Zero to beat AlphaMaster – the Go playing artificial intelligence champion.
David Silver, the lead researching working with AlphaGo Zero, had this to say:
AlphaGo Zero not only rediscovered the common patterns and openings that humans tend to play … it ultimately discarded them in preference for its own variants which humans don’t even know about or play at the moment.
Basically, Google’s artificial intelligence computer has completely revolutionized a way to play an ancient Chinese game that has been played by humans for generations. Oh, and it only took it 40 weeks to be able to beat the previous AI master of the game. You aren’t alone if you are imagining this AI becoming sentient, uploading itself to Roombas everywhere, and starting the great robot uprising. However, this AI seems too busy playing Go to concern itself with petty things like world domination.
Google’s Artificial Intelligence Computer Doesn’t Run On Pure Horsepower
Probably the most interesting part of the discovery is that AlphaGo Zero actually had less computing power than the original AlphaGo. We often get so hyped about how many cores a chipset has and how much RAM is packed in that we forget about the other factors in computing. For this experiment, researchers say that algorithms had a much greater impact that computational power. This could be a huge step for AI and computing in general. We can see that we don’t need massive super computers to create an effective AI program. What we actually need is better programming that allows the AI to learn for itself.
Again, that sounds very scary and it does bring up debates we have heard recently. For example, Elon Musk is one of the big names in the tech world that is sounding the alarm bells about AI. How long until AI begins learning for itself and advances beyond what humans can control? Could it shut down networks? Could the AI proliferate itself into internet connected devices and expand its power? The doomsday scenarios are a little overblown at this point but a definite concern if AI continues to grow at the rate it currently is.
For now, Google’s artificial intelligence computer is a cool project with little threat to the world unless you are a professional Go player. I trust that the researchers in charge will make the right decisions for humanity. If not, well then I welcome our robot overlords with open arms.