How I Decided to Skynet Myself

Like most people who follow technology, I was surprised when early in 2016 a team from Google unveiled an AI that could beat one of the best players in the world at Go. The depth and complexity of Go are such that traditional techniques (expert systems and brute force search) are simply impractical for it; the only way to develop a system to win at Go would be to build a large and complex enough Neural Network to handle the game, and then essentially let it play until it learned how to beat us.

Following the outcome of the matches (see here), I was struck by a thought:

Given that Google, with (comparatively) unlimited resources, is able to develop and train a Neural Network powerful enough to defeat the best human players at Go, is it possible for me, with my relatively wimpy laptop, to develop and train a Neural Network that is only smarter than I am?

Thus was born the impetus for the Skynetting Myself project: to develop AI that is smarter than me.

Advertisements