i was wondering if people play GO in taiwan
Yes. Competitions are televised.
oh iāve never turned on my tv since coming hereā¦
do you know where i can learn?
or buy a Go board?
or whats it called in chinese?
or if chess is played here in clubs? like informallyā¦iām not good enough to compete.
weichi
Iām gonna use this old thread to talk abut Googleās Deepmind AlphaGo. AlphaGo is a Go playing computer that uses two deep learning neural networks, one that does policy decisions, and the other evaluate which move is more likely to yield a winning outcome.
AlphaGo has beaten the European Go champion Fan Hui 5-nil, with no handicap. Itās the first that any computer program has been able to defeat a professional human Go player without handicap.
Artificial Neural Network simulates how biological neural networks processes information, and thus can learn by experience. Itās what drives google and facebookās image/facial recognitions.
AlphaGo will play the legendary South Korean player Lee Se-dol. Very excited to see what happens then.
en.wikipedia.org/wiki/AlphaGo_versus_Lee_Se-dol
AlphaGo won the first Match. Lee Sedol began the game rather unconventionally. Many professional go players like to use unorthodox moves to confuse computer programs. Lee might have tried to do the same. For that he got in a bit of trouble early on, as AlphaGo didnāt really fall for it. However as the game progressed Lee gained a significant lead, while AlphaGo made a couple amateurish mistakes.
As a result, AlphaGo took more time to think throughout the game, using up almost all of its 2 hours. However, the tide was turned near the end of the match, when AlphaGo went on an offensive streak, and Lee responded too conservatively. AlphaGo secured the area on the top left with the 116th move, while distracting Lee by taking a little corner on the top right. In the end Lee had more on the board, but since he chose to play first, he ended up losing after counting the 7.5 komi to AlphaGo. Lee was eventually forced to resign.
Game two tomorrow, live stream link: youtube.com/watch?v=l-GsfyVCBu0
If Lee begins the game conventionally, he is still favored to win Game 2. Never the less, this win is still an incredible achievement in the field of neural-network machine learning.
Itās impressive to see how far we got on neural networks!
The issue with neural network is that we are mimicking the organization of biological brainās, when does it stop being just machinesā¦
Not looking good for Li Sedol againā¦ Li has used up all his regular time and is limited to 1 minute per hand. Every time Li tries to take out AlphaGo in a localized battle, AlphaGo manages to survive. The complex fight at the center, which was supposed to be Li strong suit and for a while looked that way, all just went down hill for him in late game. Even outside localized battles AlphaGo is still out-thinking Li.
Itās hard not to have mixed feelings watching one of the best human players in the most difficult human game ever constantly scratching his head and shifting in his seat like heās constantly playing catch up.
By the way, AlphaGo seems to be trying to write the Hanji character for Death āę»ā on the board. Hope thatās not a message for all of us.
Li resignedā¦ W2 for our future overlord.
As I received in my inbox today: DeepMindās co-founder explains, this breakthrough is just the beginning. The ultimate goal is to āsolve intelligence.ā (No word on how to solve for a lack of it.)
And here is a short list of game computers still canāt win. (Sadly, I donāt see ākill all the humansā on the list. 
)
[quote=āRicarteā]As I received in my inbox today: DeepMindās co-founder explains, this breakthrough is just the beginning. The ultimate goal is to āsolve intelligence.ā (No word on how to solve for a lack of it.)
And here is a short list of game computers still canāt win. (Sadly, I donāt see ākill all the humansā on the list. )[/quote]
The issue is just copying biology isnāt āsolving intelligence.ā We can totally simulate the human brain today with hundreds of servers armed with the most advanced GPUs, all we need is enough money and resources (the electric bill will be enormous). We then have to spend years āteachingā that neural network, and āitā will literally be intelligent. Does that mean humans solved what it means to have intelligence?
In the future people will most likely spend time finding out what is the best way to āteachā future neural networks, so we can get the best results in the shortest amount of time. Real understanding of why mimicking synapses results in intelligence probably would remain a mystery to us for generations to come.
For example, this is a GoPiGo:
dexterindustries.com/shop/go ā¦ ter-kit-2/
which is just a raspberry pi mounted on a RC car with sensors.
This GoPiGo project:
dexterindustries.com/GoPiGo/ ā¦ orm-brain/
recreates the brain of a tiny worm called Caenorhabditis elegans in the Raspberry Piās memory. A C. elegans only has 302 neurons. The entire simulation is only about 5000 lines in python. Without actually writing the code to tell the bot how to react when itās about to hit something, this GoPiGo with a simulated worm brain can avoid obstacles on its own and search for what people programmed it to think is food.
Thatās great and all, but we have absolutely no idea how it works, we just know if we copy the wormās neuron connections, it somehow instinctively knows how to move aroundā¦
Good point.
But what should be the main goal? To develop a system that can replicate a humanās brain or to understand how does it work? Even though we are far from reaching the former, it is still easier than the latterā¦
I didnāt know about the GoPiGo project, but looks very interesting! Iāll check more on that.
Chinese top Go player Ke Jie, who famously defeated Li Sedol back in January, made a statement on March 9th saying āeven if AlphaGo defeats Li Sedol, it still stands no chance taking a win from me.ā
Well, Master Ke, of course AlphaGo canāt defeat you, because when the game is held in China, AlphaGo wonāt be able to connect back to Googleās servers due to the Great Firewall of China.
Well, after another devastating defeat in Game 3, Lee Sedol finally defeated AlphaGo after it made a unrecoverable mistake in move 79. Lee was in trouble right off the bat, but when AlphaGo enclosed on Lee from all sides, Lee made an amazingly unexpected move on the 78th move. AlphaGo wasnāt able to respond to Leeās move and Lee reserved the tide and forced AlphaGo to resign by move 180.
Demis Hassabis, the head of AlphaGo, tweeted that AlphaGo realized the 79th hand was a mistake by move 87, but by the it was too late to find a way to come back from it.
First 90 hands.
I get the feeling that AlphaGo was trying to write the Chinese character ē¬ this timeā¦ It looked more like it before 7 of his pieces were picked away by Lee. Interestingly, one of those pieces was the 87th hand, when AlphaGo realized he made an mistakeā¦
Quite interesting. Gonna watch these videos and try to learn how to play it.
For got to mention, Chinaās top Go player Ke Jie, who dissed Li Sedol for losing, played his first game against AlphaGo today. He lost by 0.25 point.
Someone predicted "Li Sedol wonāt be remembered as the first to lose a Go game to AlphaGo, he will be remembered as the last person to win a Go game against AlphaGo. So far it seems like the case.
Next game is Thursday.
This time they even have team games, pairing AlphaGo with another human player to play against two other human players.
If anyone is interested, I noticed this place the other day. Maybe not the only one.
ROC Go Association
äøčÆę°ååę£åę
No. 69, Leli Road, Daāan District, 106
02 2735 7519
That was frigging amazingā¦
For the most part it Ke even seemed like he had an small advantage. Then AlphaGo drew a house on the board and Ke forfeited.
The most freaky thing is AlphaGo started the game the same way Ke did in their first match. That is just eerie. Spiting your opponent by using an opening that he just lost with, and then win anyway, is just such a human thing to doā¦