Add a Comment   Listen to Podcast   Up to OJB's Blog List

Sadness and Beauty

Entry 1778, on 2016-03-16 at 14:54:40 (Rating 2, Computers)

I recently read an article from Wired titled "The Sadness and Beauty of Watching Google's AI Play Go". It was a commentary on a Go match between AlphaGo (Google's Go playing artificial intelligence) and Lee Sedol, one of the world’s top players.

An expert commentator (who had earlier lost 5 games and won zero against AlphaGo) said, relating to one move: "It’s not a human move. I’ve never seen a human play this move" and then, "So beautiful."

So the "sadness" is the realisation that a computer can out-play a human in a game which was once considered impossible for a computer to even play at a competent level. And the "beauty" is the absolute and unexpected magnificence of the move.

A move that a small number of very special humans can appreciate but probably never make in a real game. And the fact that a machine has done something more beautifully than a human just adds to the sadness (the match was eventually won 4-1 by AlphaGo so the machine is not totally dominant yet).

Before I go any further, here's a small bit of background on Go and game playing programs in general...

Go is an ancient game (it originated in China over 2500 years ago) which is many orders of magnitude more complex than Chess (there are 10^761 possible games compared to an estimated 10^120 possible in chess). It involves placing stones on a 19 x 19 board and attempting to capture opponent's stones and to gain territory. Many people thought that, even after beating the world champion at Chess (IBM's Deep Blue beat world champion Garry Kasparov in 1997), a computer would never master Go.

But one has.

Chess programs tend to rely on pure brute force to gain the advantage. They analyse millions of moves per second and many moves ahead. But "understanding" the game and "learning" from past games isn't so important. But because of the sheer number of moves possible in Go that approach cannot work. So AlphaGo actually learns from every game it plays, and it can play against itself to get even better. In many ways it "thinks".

The idea of computers thinking is a problematic one, largely because no one really knows what thinking actually is. But even if they don't think now I believe it's only a matter of time before they do. It's not a matter of creating anything genuinely new, it's really just about how computers are programmed.

The techniques used in AlphaGo should be able to be used more generally to create expert systems in any area - medicine being one of the most obvious. Should we trust a computer with our health? I certainly would, especially after the very mediocre results I tend to see with human doctors.

What about computer programming? Could I be replaced with a computer? Well that might be more difficult because there's a much greater level of creativity involved. But that also is probably only a matter of time. In fact all jobs will be able to be replaced, and I will go further: all human activities will be able to be replaced.

So where does that leave us? Well, nowhere really, because humans will be obsolete. The beautiful machines will create more beauty - in science, art, fiction, everywhere - and the sad humans will just watch in awe. Maybe until they decide to end their species, or maybe the machines will do that for us. Just beauty will be left, but some sadness too.

There are no comments for this entry.

You can leave comments about this entry using this form.

Enter your name (optional):
Enter your email address (optional):
Enter the number shown here:
Enter the comment:

To add a comment: enter a name and email (optional), type the number shown, enter a comment, click Add.
Note that you can leave the name blank if you want to remain anonymous.
Enter your email address to receive notifications of replies and updates to this entry.
The comment should appear immediately because the authorisation system is currently inactive.