Artificial intelligence has become one of the most interesting (and terrifying) realms of computer development. It’s a field that develops as rapidly as computational power does (that is, exponentially), and more recent discoveries have provided tell-tale signs that the relationship between humans and computers is likely to become a lot more complicated and mysterious than that between a coder and a computer of today.
Take, for example, the recent victory of DeepMind’s Go-playing creation, AlphaGo. AlphaGo beat the world’s human master of the 3,000-year-old Chinese game of strategy and intuition, chiefly by executing move that, according to its own calculations, had a 1 in 10,000 chance of being chosen by a professional Go player. Instead of listening to this precedent, AlphaGo looked into its own neural network of self-taught algorithms and saw opportunity where no humans had before. It played the move, threw off the reigning champion, and became the new world Go master.
Some have seen a lot of humanity in AlphaGo’s playing style. They say that AlphaGo’s ability to guess at an outcome from starting conditions it had never seen before can be defined as intuition, and its choice to use such an irregular move can be seen as a product of introspection, and of identity. After all, it’s not like AlphaGo was simply programmed to do a certain move given a certain circumstance. No, instead it was operating off of machine learning, when a computer is fed huge amounts of data and information but given no cut-and-dry way of processing it. Instead, the computer creates its own abstractions to fit what it has seen, in way modeled to mimic the neural networks and processes of a human brain.
Deep learning and neural networks aren’t just used to beat Go champions; Facebook and Google use these kinds of computing systems along with big data to identify faces in pictures, help computers understand distinct speech patterns, and complete a fair amount of other services that have become the norm in our internet age.
AlphaGo’s victory, along with the Chess and Jeopardy victories famously won by computers using artificial intelligence, tend to make people a bit nervous and sad; what humans distinctly own, their intuitive intelligence and accomplishment, is starting to be challenged and surpassed by the very machines they made. You don’t have to be worried about a robot takeover to feel a little fear and disappointment when thinking that there’s another way that humans have become inferior to machines.
That said, it’s important to note that the human Go player that was beaten by AlphaGo did win a few games using some surprise moves. And after playing the machine, he began a winning streak against human players the likes of which he had never encountered. His experience against the machine had trained him to be a better player the way no other humans, faced with his dominance, ever could.
So yes, the machine made a human decision that made it better than any human before it. However, in the process it made the humans around it better too. Perhaps a dual evolution is possible, and humans and computers can use friendly competition to increase their capabilities further than ever dreamed possible.