The 2 things about the game of Go that primarily peaked my interest when I first discovered the game were:
- It is believed to be the oldest game in human history that is still played today - by some estimates 4,000 years.
No computer Go program had ever before in history defeated a human Go player above the skill level of an experienced & talented amateur level. vs for decades computer programs had been able to defeat master chess players. Computers we pretty good at individual phases or the game, but could not grasp whole board play.
Mathematicians have calculated that to completely calculate the next eight moves ahead, on an full sized 19 x 19 Goban, would require computing 512 quintillion (5.12×1020) possible combinations. But different methods are used to calculate possible moves, another method comes up with about 250^150, or 10^360 possible moves. Options beyond thought, requiring intuition and imagination to be a great player.
While Go has simpler rules than chess, Go’s simplicity permits much higher flexibility, more options, higher complexity. [ in a similar way I see tenkara as a system with simple basic principles, that allows maximum options of styles and methods of tenkara fishing]
I haven’t spent much time on Go for several years. Close to two decades. Looking about my house trying to find the many Go books or equipment I purchased during my peak activity. I discovered my last purchases were about 2002. I also discovered on-line websites, where one can play computer Go, no longer recognized my userid (Crossroads).
Probably the thing that drove, edged out, my fading interest was by the early 2000s my son had become old enough to begin wanting to explore the wider world beyond home, and that took priority. Though I did teach him how to play Go, and the games we played were competitive and fun, because both of us were equally matched poor players.
Anyway, I just discovered item #2 changed dramatically about 4 years ago.
The people at DeepMind created AlphaGo. The first version used the input from human game records going back hundreds of years. And for the first time a computer defeated a human Go Master. Then several of them. A later version was only programmed with the rules of Go, and some information about basic principles of how the game is played. Then was allowed to improve it’s level of play by playing games against itself.
It used tabular rasa learning - blank slate learning. Allowing the program to be highly creative in how it played games of Go. AlphaGo was quickly able to defeat every human Go master it played. Complete shutouts in some cases, and also revealing some styles of play no human had yet discovered. You can watch several of the matches on youtube, but you would have to be a dedicated Go addict. The games are of 5 ~ 6 hour duration.
My two takeaways from these events are:
AI is going to have a huge effect on human society, the types of employment available in the near future. Will AI soon have better intuition than humans?
And secondly, there are some advantages to learning how to do things on your own. Humans can also be more creative learning how to do things on their own.
[Developing unique styles, examples can be seen in the skill of self taught guitar players or the different fishing methods of master tenkara anglers whose styles direct different preferences in rod, line, kebari characteristics developed by these master tenkara anglers. ]
AlphaGo Zero: Discovering new knowledge
From the video description:
"Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go. AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play. In doing so, it quickly surpassed human level of play and defeated the previously published champion-defeating version of AlphaGo by 100 games to 0.
If similar techniques can be applied to other structured problems, such as protein folding, reducing energy consumption or searching for revolutionary new materials, the resulting breakthroughs have the potential to positively impact society. "
[ or maybe negatively impacting society]
AlphaGo Zero: Starting from scratch
How the Computer Beat the Go Master
[there is a most interesting and informative 7 minute video on the below linked SA article. I could not find a way to embed the video. Worth watching even if you don’t read the article]
DeepMind was a London based AI pioneering company, acquired by Google in 2014.
On one hand AI and it’s impact to change society is something to be concerned about.
But, otoh, since Google started using AI for it’s digital translation of languages, Japanese to English. It now, in my view, does not function as well as it did four years ago, and worked much better before AI. AI’s potential for better or worse may depend on where it is used.
The point is - today a lot of attention is focused on the thought processes that goes on - on the 19 x 19 playing field of this ancient game played by people on these beautifully made, and (in these examples) expensive Gobans (Go boards) that may affect our future in unpredictable ways. The key for machine or man seems to be developing - imagination.