The Future Is Faster Than You Think: How Converging Technologies Are Transforming Business, Industries, and Our Lives (Exponential Technology Series)

by Peter H. Diamandis

In 1997, IBM’s Deep Blue defeated the reigning world champion, Gary Kasparov. Typically, the game tree complexity of chess is about 1040—which means, essentially, if every one of the 7 plus billion people on Earth paired up and started playing chess, it would take them trillions and trillions of years to play every single variation of the game. Yet, in 2017, Google’s AlphaGo defeated the world Go champion, Lee Sedol. Go has a game tree complexity of 10360—it’s chess for superheroes. Put differently, we humans are the only species known to have the cognitive capacity to play Go. It only took a couple hundred thousand years of evolution to develop that capability. AI, meanwhile, got there in less than two decades. Still, AI wasn’t done. A few months after that victory, Google upgraded AlphaGo to AlphaGo Zero by updating their training style. AlphaGo was educated via machine learning, essentially fed thousands of games previously played by humans, and taught the proper move and countermove for every possible position. AlphaGo Zero, meanwhile, required zero data. Instead, it relies on “reinforcement learning”—it learns by playing itself. Starting with little more than a few simple rules, AlphaGo Zero took three days to beat its parent, AlphaGo, the same system that beat Lee Sedol. Three weeks later, it trounced the sixty best players in the world. In total, it took forty days for AlphaGo Zero to become the undisputed best Go player on Earth. And if that wasn’t strange enough, ...