Back to List

Notes and Reflections on Books and Media

by Hannah Leitheiser

Machine Learning

AI Superpower: China, Silicon Valley, and the New World Order

Kai-Fu Lee



- Kai-Fu Lee, AI Superpower: China, Silicon Valley, and the New World Order (2018SEP25)

Google set their DeepMind project to try chess, and given the rules of gameplay, it was able to beat Deep Blue after four hours. They then asked it to play some Atari games and it became an expert.

I'm still trying to wrap my head around this chapter in artificial intelligence. There are many names, but the recent technologies are usually called "machine learning." It's just one branch of Artificial Intelligence. If I understand, it's something like curve-fitting. We all know if you have a bunch of data points, you can plot a line of best fit. You can add degrees of freedom and plot curves. Machine learning involves thousands of degrees of freedom, so can map complex input to responses.

The process may be similar to the human mind, although currently machines require much more training data to get to the same level of accuracy. It's important to note, too, machine learning is merely inductive. There's no answering why about responses, which may be troublesome if machine learning is used to make important decisions.

Machine learning will be applied to most domains where there is, or could be, a large amount of data, and where some kind of goal or quality factor can be measured (number of clicks on an advertisement, for example).