False economy What’s in it for me? Understand what makes the

the 
development of deep neural networks.
Since the turn of the twenty-first century, AI has 
made remarkable progress. The key to this 
progress has been the development of deep 
neural networks or DNNs.
The precise mechanics of how these work are 
rather complicated, but the basic idea behind 
them is fairly simple. Somewhat similar to the 
human brain, a DNN consists of thousands of 
simulated neurons linked together and arranged 
into hundreds of complex layers. By sending and 
receiving signals to and from each other, these 
layers of neurons are able to do something 
called deep learning. That means they can teach 
themselves how to do things with little or no 
human supervision; they don’t have to be taught 
by their human creators, like computer programs 
of yore.
By harnessing the deep-learning power of 
DNNs, AI was able to defeat one of its longest-
standing adversaries: the ancient Chinese board 
game of Go. Played with white and black stones 
on an open grid, this strategy game is even 
more complex than chess, despite its simple 
appearance. For instance, whereas chess has 
only 20 possible opening moves, Go has 361. 
And by just the second round of play, the 
possibilities balloon all the way up to 128,960!
Because of the game’s complexity, an AI Go 
program needs to be able to engage in very 
creative, responsive and on-the-fly strategic 
thinking to win a game against a skilled human 
opponent. For decades, such a victory was one 
of the primary benchmarks against which the 
power of AI was measured. And from the 1970s 
all the way through the early 2000s, AI failed to 
meet the mark, losing even to novices and 
children. The game was just too complex for it.
But then along came a start-up called DeepMind 
that specialized in deep learning and was 
acquired by Google in 2014. In that same year, 
the team from DeepMind deployed a DNN-
powered program called AlphaGo against a 
professional Go player, Fan Hui. It beat him five 
games to zero. It then went on to play in 
tournaments, where it trounced every human 
opponent it encountered – including the reigning 
world champion!
The benchmark of winning at Go had finally 
been reached, and the incredible deep-learning 
powers of DNNs had been dramatically 
demonstrated. But as you’ll find out in the next 
blink, AlphaGo was just a taste of things to 
come.