A Human's Guide to Beating DeepMind at Go

By Gary Cutlack on at

Classic board game Go was chosen as the latest demonstration of the awesome power of AI, with Google's DeepMind-powered AlphaGo claiming a big human scalp in the form of current Go world champ Lee Sedol.

But what is Go and why is DeepMind so good at it? The British Go Association has all the rules, and it's easy to see why it's been chosen as the ultimate test of AI, as there are very few specific rules; it's all about tactical thinking.

It's based around capturing territory by placing down black or white stones, with the winner the person -- or the advanced AI routine -- who fences off the most of the board and seizes the opponent's pieces.

Try it on the internet in your browser, to see how open-ended it is. Top tips for beating humans include:

- Start on a small board, as the full 19x19 grid is a killer for amateurs, who'll quit long before being destroyed
- Review the pro commentaries -- the AlphaGo match has already been thoroughly dissected
- Practise common "problems" for free on the internet
- Study old matches, much like how DeepMind is said to have learned

Top tips for beating an AI opponent include:

- Pulling its plug out
- Downloading loads of torrents on it while it's playing
- Making it defrag itself while it's playing
- Distract it by opening 100 tabs

Image credit: Go from Shutterstock

Want more updates from Gizmodo UK? Make sure to check out our @GizmodoUK Twitter feed, and our Facebook page.