AlphaGo vs. AlphaGo with Michael Redmond 9p: Game 34
Vložit
- čas přidán 25. 07. 2024
- Michael Redmond 9p, hosted by the AGA E-Journal’s Chris Garlock, reviews the 34th game of the amazing AlphaGo vs. AlphaGo selfplay games. The 50-game series was published by Deepmind after AlphaGo's victory over world champion Ke Jie 9p in May 2017.
Check out the sgf file -- with variations/comments -- here: www.usgo.org/news/2020/01/alp...
Produced by Stephen Hu, Allen Moy, Andrew Jackson and Chris Garlock
Thumbnail image of Rock 'em Sock 'em Robots by Lorie Shaull - Own work, CC BY-SA 4.0, commons.wikimedia.org/w/index... - Hry
I love those videos. Keep doing more please!
The hane at E7 reminds me of a particular lesson I got from a Dosaku game, where, while being attacked into the center, he interrupted the flow of his stones to make a cut, and it had the effect of allowing him enough aji to make a counterattack later.
Great game and great analysis. Really looking forward to more of this!
Dave
I find this fascinating yet I have no idea how to play the game lol
You can learn the basic rules on line, but it helps a lot if you're somewhere where other folks play. The rules are deceptively simple. The game itself is more complex and thus more difficult than chess. Even so, amateurs can have a lot of fun playing! Intuition and logic are both important. Go is great mental exercise. It's also an excellent way to develop positive character traits.
I like Chris making observations, it's a good opportunity to learn when Michael approves or corrects him. Chris should be not be afraid of being corrected, if he doesn't get it right, most viewers probably wouldn't either.
+
Thank u Michael, Chris, and Stephen
28:50
In 2020 Leelaz seems to like this variation the one "too bad for white" slightly better (less than 1%) than the one Alphago chose, maybe she judged the moyo is really big and having those 6 stones captured gives white more forcing moves to destroy black's moyo ? Anyway in both variations the game is even (45%-55% which is even game on lizzie with 7,5 komi)
this is so interesting to see how A.I changes over time !
Love the reviews, Michael's explanations are always very clear and helps me asking myself the right questions and getting the right idea more often when i play my games, thanks ! This is priceless !
Michael you're the man
Fantastic game, fantastic explanation: thank you!
This is as good a place to ask as any: when looking at the current Leela Zero version, is this Leela stronger than the AlphaGo (non Zero) version playing here? Of course, a match could tell, but I hope Michael can give his opinion of it, him knowing both AlphaGo and Leela Zero quite well by now.
I like the idea of the live stream though I'll likely never join one. I just like that others with similar questions to mine will be able to get them in. And what a great game! It's unfortunate that the video gets too choppy to follow towards the end. I hope you iron out the technical problems.
If the choice is between streaming and video quality, I choose quality.
+
smashing the like button
What channel?
What is this steamed on?
Twitch.
Lee Schumacher what channel?
www.twitch.tv/usgoweb
The game diverges from AlphaGo Teach's database at move 8. Its only suggestion for white is to block the other way. Here's how it would continue the opening from there: online-go.com/demo/view/438936
"Thanks for the question....ShleemJuice” 😂🤙🤓
omg
In support of Lee Sedol's position, "AI" as it is presented to us is a bit of a fraud. AlphaGo does not play the same game that the best people play, but appears to play a minimax saddlepoint of reducing risks in minimizing losses rather than taking risks in maximizing wins.
Assuming all games end in either a win or loss, isn’t “minimizing losses” equivalent to “maximizing wins”?
@@alanhuang3005 yea and may be he doesn't really know about how humans play when they want to win, they always make compromises too.
Maximizing or minimizing risk is not the same game. The best players are not minimizing their chances of losing until it is clear that they are ahead. Novel plays, innovation, and risk-taking is therefore the province of those who find themselves either winning at the beginning or behind in the later game. Perhaps Michael could advise us as to how many AI-AI games end up being very close. Maximizing wins through risk will likely lead to a certain number of resignations or wide variances in scores.
@@raedwulf0 When a really strong player plays a somewhat weaker player, he tends to play conservatively. Likewise, the weaker player might take more risks to counter that strategy. When equally skilled players play, it depends on the player style. Some players tend to play conservatively and others tend to take risk.
I don't really see how that relates to your statements.
I don't believe AlphaGo has any concepts like "risk". All it does is evaluate the probability of wins if different moves are played and pick the move which gives the best probability of a win. So a move which gives a 53% probability of a win by 1 stone will be chosen in preference to one which gives a 52% probability of a win by 9 stones. In this respect it is very different to a human player.
Dear Chris Garlock, after all this time, you and Michael's chemistry is still awkward. Be a leader of change, get a better co-host.
I liked their chemistry from the beginning
it's a dynamic duo
I don’t think they’re bad, I like it
I like the team too!
He's done a lot to get us these wonderful videos we wouldn't otherwise get. How about constructive criticism before being ungrateful?