AlphaGo to Zero: Revisiting AlphaGo vs Lee Sedol, Game 4, with Michael Redmond 9p

[@lukabozic5]

I was one of those introduced to Go through the AlphaGo event so I can relate what Garlock is saying at 13:06

[@Lylairl]

Bought right away, thank you guys for continuing to produce some of the best and easily accessible go content.

[@Clara-anise]

Its neat to me that Lee Sedols turnaround move where he wanted to play something as confusing as possible because there were at least some variations where it worked out for him is very similar to how AI like alphago play when they're losing

[@1Eagler]

Thank you Both for commenting these games, y were wonderful.

[@ZhangyXD]

i was a 0.0025 dan, after this video i think i became a 0.25 dan

[@jacklyfgardon9254]

Good job here guys as always. Thanks you.

[@USGOWeb]

Our pleasure!

[@luisbanegassaybe6685]

Just purchased the ebook, thanks for your commitment guys

[@mitthrawnuruodo8276]

Where did you buy it?

[@USGOWeb]

Hope you enjoy it!

[@decidrophob]

Wholly enjoyed the book!!!! Thank you very much always for the dedicated work.

[@liberodentro]

Thank you Chris and Michael 9 dan for the fantastic review videos you are making! I was wondering if it could be possible to cast a series of reviews on the games between Lee Sedol vs Handol AI played during the “Lee Sedol retirement series” . There Lee Sedol played with a 2-stones handicap and lost, except for one game where he pulled out another of his extraordinary miracles, like move 78 in Alpha Go game 4 or the famous wedge in the “Broken Ladder” game

[@pianoforte611]

1:06:30 Wish they linked to that game in which Michael Redmond lost a bunch of stones and still won.

[@kiennham8638]

might help to put the link to the book in the description of this video. It took me a few clicks to actually find the book since it is still so new

[@USGOWeb]

Great suggestion; done. Thanks!

[@messitup]

We all know David Silver played P8 to inspire hope for humanity. It was a 5-0.

[@simonassatkus2298]

Why AlphaGo is not used to determine fair komi? From 50 self played games it was clear that 7.5 is too much.

[@Clara-anise]

I could be wrong but I believe it gives a slight advantage to black when the komi is 6.5, meaning that the really ideal komi is probably something like 7, but that would lead to draws. Since top pros aren't quite as strong as alphago, their komi don't need to be quite as high, so that's why the standard in pro matches is 6.5 these days

[@simonassatkus2298]

@@Clara-anise Is there any public data on how good alphago as black is dealing with 6.5 komi?

[@Clara-anise]

@@simonassatkus2298 I'm not sure whether or not AlphaGo can use arbitrary komi or if it's stuck at 7.5, but I know that KataGo, which can handle arbitrary komi, gives a slight preference to black at 6.5 (I think about 55%)

[@simonassatkus2298]

@@Clara-anise Thanks for the input, never tried KataGo before. I wonder how game analysis would differ there :)

[@MelindaGreen]

@@Clara-anise It would be easy enough to find out, using simonas satkus' suggestion. Let it play itself for 1000 games with 6.5 komi and see if the win rate gets closer to 50% or not. My guess is that it will, and might even justify returning to 5.5.

[@ConsciousBreaks]

1:51:53 LoL Michael

[@lstealth]

My question is can alpha go eventually figure out the perfect start as black, to no matter what white does they will loose?

[@JasminLeblanc]

Yes. But we are leagues away from that. Tic tac toe is a solved game. Chess is not solved but we are getting close with the way ai and sheer computing power is getting better and better. BUt go? thats still many many leagues away. Expect much scarier stuff with ais to become common in your daily life before Go is solved.

[@lstealth]

@@JasminLeblanc such as what, what can advanced a.i be used for?

[@blubbfisch98]

@@lstealth automatic healthcare, pre-defined career paths and choice of romantic partner, complete substitution for human interaction... There are many more and most of them sound incredibly scary now but will be common in 200 years

[@Pabloparsil]

I dont think so. It could develop one strategy where nobody knows how to play against It, but you can't prove that it's unbeatable

[@lstealth]

Is there anyway the alpha go programmers , made alpha go down a few notches from hard-core to medium maybe to give 1 win to this poor man who might have off himself loosing 0 to 5?

[@abhigo7788]

Yo

[@joshuaaa3337]

do you think that Alpha Go lost game 4 in purpose to not hurt the feeling of Lee Sedol? I am talking about the people who control Alpha Go. I think that if they did not do that Lee Sedol could committee suicide!

[@jacksonfitzsimmons4253]

No way. They had already beaten him 3-0 and the game was solidly in AlphaGo's favor before 78. Also you have to keep in mind that this version of AlphaGo was obviously very strong, but it was still far from perfect and it's totally feasible that a pro like Lee Sedol could find a way to win

[@zhentian1977]

Have you watched the documenter?

[@gnsgml11]

Go watched the documentary where they show their reaction instead of asking stupid questions

[@lstealth]

I also believe he would commit suicide , you might be on to something, however I also believe the guy who played this game his whole life at top level could squeeze atleast 1 win out.

[@Averymindset]

Although i believe that the alphago team didnt purposely let Lee win. I do believe that if Lee, who is from strict asian country, playing at the top of the world in his whole life, losed 5-0 in front of every eyes of go player in ther world, he would have commit suicide.

[@amadormakabenta4260]

The halting camel legally soothe because raft bailly tick but a garrulous nest. xenophobic, true cardboard

[@Omniumae]

Exactly what I thought