At the Forefronts of AI (Thoughts on AlphaGo)

Livestream link (English) here if you’re interested in watching the rest of the matches. More information about AlphaGo here. They also have a paper published in Nature for those that are academically driven.

———–

I foolishly stayed past 3am last night, without studying a word for my midterm this morning,  unable to stop watching Lee Sedol, a 9-dan professional Go player, test his wits against the mighty technological advancement that is AlphaGo – a Google-owned computer program (powered by hundreds of CPUs) that is able to match even professional level Go players in the difficult, but logical board game.

It was game two and (spoiler alert) he had lost the previous game the night before (or day before, in Korea time). The victory by AlphaGo came as a surprise to all the professional commentators – none of them had predicted that the program would be such a powerful opponent judging from its performance last October against Fan Hui. While it would take decades for a professional player to reach this level, AlphaGo was able to play against itself millions of times within the past months and establish a truly powerful playstyle.

Prior to watching the first livestream, I only had a preliminary understanding of Go – I’ve only played it once or twice and only casually in the free time during math class in high school. So it’s obvious that I’m no Go professional, but even though I barely understood half of the moves that AG and Lee Sedol were making (and the moves I did understand were due to the well-thought out explanations and predictions from Michael Redmond during the stream), I could appreciate the heavy implications this battle brings into both the artificial intelligence scientific and Go communities. And those implications, for humanity and for AI, was what had me watching all night long.

lee-sedol-alphago

I was cheering for Lee Sedol the whole way through because I didn’t want to believe that humanity could be bested by machine in such a complicated game that often pushes human intelligence to the limits (as seen in high pressure professional matches). It comes as no surprise that my heart sank when the commentators began estimating or counting the score in the mid-game and he was once again behind and losing to AlphaGo. There was so much pressure on Lee to perform that I couldn’t help but feel for him. On one hand, AlphaGo winning is a step forward in AI, a step forward in technology, and in that sense, a step forward in humanity. Even so, it felt like the opposite – oppressive in the sense that even the greatest human achievements are nothing in the face of a machine. Almost like the AI was undermining all the years that these players spend perfecting their  game -making even the professionals at the top of the ladder seem flawed – and that’s uncomfortable to think about.

Even more unsettling to me was the way that the AI approaches the game. If the system believes she (note: AI are usually considered female) has already won, she no longer seeks to play the optimal moves and instead plays weaker moves to see how the opponent responds. Even in such a high pressure and difficult match, she’s playing with her opponent. From the view of the engineers behind AG, it’s amazing and it shows how far the AI has come. From the view of Go players, it just feels bad.

The whole situation is filling me with mixed feelings. On one hand, I’m in awe of the power that this system possesses. On the other hand, I can’t let go of that feeling that in the end, humans will be replaceable. It makes me think: what can’t artificial intelligence accomplish?

Furthermore, as a gamer, an AI that can beat the best of the best at a game is frightening. We spend so much time fine-tuning our skills, playing against other players, and working towards being better game after game. It’s depressing to think that the fruits of our efforts after years of gameplay is worth less than months of training a computer. It’s even scarier to think of players that can use these computers to train instead of playing of other people. If something like AlphaGo could become commercialized (though doubtful that Google would release it), what would be the point of Go players playing each other anymore for practice? The whole social aspect of the game would go down the drain…Just thinking about that consequence makes me shiver.

It’s these uncomfortable, but exciting thoughts that  draw me to continue watching the livestreams. It’s incredible how far artificial intelligence has come and even if the implications make us uncomfortable, I believe it’s worth the pursuit. And although it’s highly doubtful that Lee Sedol will be able to win the series at this point (the winner after 5 matches gets a million dollars), I’m really hoping that he’ll be able to take one game off AG. I think it’d give us a glimmer of hope that there is something incredible about the human brain that even a machine cannot read. Fingers crossed. Go Lee Sedol! 🙂

Even if he isn’t able to win, I suppose we’ll just have to accept it as also a good thing for humanity…right?

Advertisements

6 thoughts on “At the Forefronts of AI (Thoughts on AlphaGo)

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s