comment 0

Can the new bots teach us to play better Go?

Ever since Alpha Go and the other uber-bots arrived on the scene and started trashing top pro Go players, their influence on the Go world has been huge. Strong pros have been adopting numerous plays innovated by the bots, and it’s clear why that’s happening – the bots have been kicking butt.

Recently, players have been downloading copies of a program called Leela and in particular a version or interface known as Lizzie and trying to use it as a teaching tool. One popular Go content provider on YouTube, Nick Sibicky, has just produced several videos using Leela or Lizzie as a teaching tool. I’ve found these to be less than helpful. Here’s the comment I left on one of these videos a couple weeks ago:

It could be that for some players (maybe really strong players?), using a bot is a super-useful tool but here’s another perspective. If I can one day get to your level of play, I’ll be a really happy guy – so I’m happy to learn from your ideas and experience. That is way more interesting to me than looking at constantly shifting winning percentages, and hearing how the bot agrees or disagrees with you. I get that the availability of strong AIs adds a new learning tool and lots of your audience are very interested in exploring it. For me though, I’d rather you talk about games you have spent some time with, which you can use to make some useful teaching points. When you comment in the video that you should look at these games before you pull them randomly, I think you’re right – doing a little pre-work could elevate your videos to a whole new level of excellence. I appreciate your work, but I’m going to sit out the Leela videos.

Today, another content provider – Dwyrin – has published an excellent video called Beware Lizzy. In it, he points out the pitfalls of using the bot to try to improve your game. In a nutshell, the bot will often recommend moves which are fine if you have the reading ability of the bot and can cope with the exceedingly complex variations the recommended moves can lead to. As well, the bot strongly favours invading the 3-3 point – which is OK too – as long as you have the strength to take care of difficult full-board reductions needed if you’re going to give up lots of influence in exchange for cash. If you want to learn to make good basic plays, take care of your groups, and find and attack weaknesses so you can profit, the bots may actually turn out to be less than helpful.

I think there are some things a puny and weak player like myself can learn from a strong bot. For instance, in the heat of a difficult battle with a complex capturing race, it would be useful to see how the bot mounts or fends off a strong attack. Dwyrin has promised to publish a video soon outlining ways in which he thinks Lizzie can help a player improve.

By the way, if any readers here in the Long Branch area would like to learn to play the game of Go (Warning: it takes some dedication to take on this beast), I’d be happy to teach you the game and play teaching games with you.

Have your say...

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s