The Inspiring Story of AlphaGo

The Inspiring Story of AlphaGo

There have been multiple incidents when a human and a machine have faced off in a competition of wits vs. coding; in Jeopardy and in Chess, computers have finally, and famously, bested their human opponents. More recently, Western Go professional Lee Sedol lost a game of Go to a computer using an advanced artificial intelligence algorithm.

alpha go2Some feel a sense of panic at these displays of computational dominance and worry that they’re a sign that computers will rise to power over lesser minds. Others see a bright future indicated by the computer’s victory; they believe that computers and humans will commence to evolve together, and be the cause of each other’s evolution due to a kind of brain-computer link.

Perhaps in no other field of computing is this more true than that of artificial intelligence. AI developers have found inspiration for computing progress in the human brain, and organizations like DeepMind are walking at the razor’s edge of neural and computational understanding.

DeepMind was responsible for the creation of the Go machine that finally became the world champion at the ancient game invented in China 3,000 years ago. DeepMind created what’s called a deep neural network, which is comprised of a network of hardware and software engineered to loosely mimic the web of neurons in the human brain. Neural networks aren’t a singular creation by DeepMind’s development team; tech moguls like Facebook and Google use them to identify faces in photos and to understand human speech.

Neural networks allow engineers to feed a computer information and allow it to create abstractions and understandings on its own based on the loads of information it is fed. This was famously demonstrated when a computer was fed image after image of cats, and eventually learned to recognize cats on its own and differentiate them from similar mammals like dogs, foxes, and the like. Again, this was done not through creating an algorithm that allowed a computer to look for certain attributes that prove that a cat is a cat and examine a picture for those attributes (a code which would be extremely difficult to make work accurately) but instead through simply giving the computer huge amounts of information and allowing it to figure things out for itself.

agIn the case of AlphaGo, the victorious Go machine, its win against the reining human champion was chalked up to a strange and unexpected move on its 37th turn. According to David Silver, the researcher who led the creation of AlphaGo, AlphaGo “knew that this was a move that professionals would not choose, and yet, as it started to search deeper and deeper, it was able to override that initial guide.”

Although there’s no way of really understanding how the computer understands anything when it’s using a neural network, the team behind AlphaGo’s creation believes that in a certain sense, AlphaGo had to start “thinking on its own” to win the game. In other words, it was not making decisions based on a set of rules that its creator had encoded in its digital DNA but instead making decisions based on algorithms that it taught itself:

“It really discovered this for itself, through its won process of introspection and analysis.”

Leave a Reply

Your email address will not be published. Required fields are marked *