This post feels like reading 2+2 about 5-8 years ago. This was a big argument back then, probably about 50% of people on each side (2+2's favorite, tom dwan, was also on your side if memory serves). Most people seem to have realized it's wrong by now. Heads up poker has an optimal strategy which cannot be exploited. And while that strategy could not take maximal advantage of its opponents every error in tendency, it would have an edge against any player in the world right now.

As an example: should you play j6o vs a button raise in NLHE? Make the wrong decision, but play optimally from that point, and you've still already lost. Bet sizing too small on the river with a polarized range? Another loss. Too big? Loss. Bet with wrong type of blockers? Loss. Fail to bluff with blockers or with a weak hand? Loss. Bet turn with a mid-strength hand because you don't want your opponent to see a free river? Loss. Sub-optimal preflop raise size? 3-bet size? Don't check-raise bluff with the optimal type of hand[s], to the right size? Loss. All these errors add up, even if they are small.

See here for info about a strong heads up Limit poker bot:

http://poker.srv.ualberta.ca/about
http://poker-blog.srv.ualberta.ca/20...sequences.html

Regarding the topic on hand, it's starting to look pretty rough for Lee Sedol in this match. It's still very difficult to know how AlphaGo's strength compares to Lee Sedol's unless you're a top pro yourself (and even then it's very difficult), but it looks like the bot is most likely stronger than Lee Sedol, and possibly substantially stronger.

While it's exciting to see progress being made, I would say that overall this is probably a bad thing for poker, online poker in particular. It certainly would not be a major surprise to see very strong poker bots appearing in the next couple of years, but I hope we have longer before they begin to infect the games to a substantial degree.

solved without any sarcasm. even nosebleed degens go broke. humans no longer needed for our pattern recognition ability. start hedging life plans.

if alphago wins this match 5-0, while continuing to never seem headed for a loss, then who's going to argue that it's not better than the best human?

before this match, the pros seemed to universally believe that it had no chance to win. they were taking bets on whether alphago could win 1 game.

I was shocked 20 years ago when kasparov was beat. i know i laid odds against big blue. fool me once... bet alpha and feeling good.

This is amazing. A couple of years ago, I thought computers were like 5 dan amateurs and 15 years ago like 15 kyu.

It's also been playing moves that commentators think are bad, but end up being good. There's reason to believe it might be FAR better than the best human.

There was an awesome moment in the second match, for those who haven't been watching, where AlphaGo makes a move and the commentator literally did a double take, and seemed super confused. Lee Sedol actually got up and left the room for about 5 minutes, while he was on the clock. He was shook.

How much time left at the end of game do these players typically have? Is the clock that much of an issue?

Lee Sedol ran out of time at the end of this match, so it did become an issue. When their clock runs out, they have 1 minute for each move after that, as well as three 'overtimes' (they can choose three to go over that one minute mark three times).

Ah thanks. How long is the overtime?

60 seconds each

Deep Mind is freaky. discovering new patterns unseen by pros and knowing how it'll play out. it has it's own intelligence which is both different and better than human players.

there are also machines that can now beat 8-bit video games, without being told the rules to the games, but by being able to track pixels.

with more power, they could apply DM to anything. why not let it learn human language? attach nouns to images, which google has done. attach verbs to videos. apply a gaming mindset in approaching anything it wants.

i don't see how it doesn't make humans obsolete. it's limited by energy and needs to attach it to some limbs, but there are all-purpose robots like Baxter coming online. having humans explore space or spending time fixing human bodies, seem like an unnecessary and inefficient complication for something superior to us.

believing that a superior intelligence is going to spend its existence serving the needs of humans is self-centered on our part. DM makes me believe the singularity is in the forseeable future.

any human born today is going to have to outperform machines superior to DM. lol

Probably already been asked already, but what implications does this have on poker? How far away are we from a perfect solution to 6max NLHE?

In the second game both players went into overtime called byo-yomi

It is standard for both players to go into byo-yomi.

These programs should improve the level of human go, as players will have a really strong opponent to play against at a handicap.

AlphaGo will help refine and challenge Go Principles. One game involved moves by AlphaGo along the fifth line which generally is discouraged, the principle being Don't push on the fifth line

google may have already built it. if not, then it's likely they're working on it in secret. the same engineers behind deep mind were talking about building AI to beat human games, and the two examples they used were Go and poker. then they reveal DM last year, and all of a sudden it's wiping the floor against the best humans. this was supposed to be a decade or more away.

go is to chess, as chess is to checkers. believing poker is unassailable is burying your head in the sand.

maybe its something about google but i would strongly root for a google bot in a HU poker match, but was quite pleased when claudico got crushed.

I have no doubt that they will be able to conquer HU poker (and 6max to an extent) its just a matter of how long till they turn their attention to it.

Really strong play by AlphaGo. Will be interesting to hear how much stronger Deepmind think it is, I would guesstimate +300 elo.

The third game was a massacre. Wow.

It bodes poorly for humans when they don't know if they're ahead or not, while the computer clearly knew it was winning far before the end when it started playing moves on a empty part of the board, ignoring the 'action' and confusing the analyst. The next stage is to be being played without realising it. Hello Skynet, ty Google

Congrat, very win, where my coin?

In game 3 Lee Sedol played a Chinese Fuseki opening that has much theory. He attempted to engage AlphaGo in Ko fights to no avail.
As noted above, AlphaGo ignored some of these to play quiet moves. Simply astounding.

I don't think you can get this to play against on your laptop. I would like to play against it with a handicap. Not sure what the strongest go program you can buy is.

Looking like 4-0. The beast is an hour ahead on the clock and the human has his head in his hands. It took 10 yrs after Deep Blue for commercial programs to spank top chess grandmasters. Expect it to be a few years before the same happens with Go. However, the DeepMind code seems a lot more sophisticated and proprietary than chess programs so unless they publish the code, it might never make it out of the lab.

This thing is also running on an absurd amount of hardware as was Deep Blue at the time.