GM Hikaru's First Ever Game vs  

Okkkk...

Mr. Hikaru Nakamura, when doing mental calculations of plausible/possible lines, do you mentally visualize the chess pieces in 2-Dimensional form, as is the case for numerous virtual chess games, or in 3-Dimensional form, as is the case for in-person gameplay?

Levi seeing Hikaru's side right?

a

500th like

This sounds like agadmator quote. 😂

@@rhettzkie Sure does haha. Agadmator's best moments.

Levy* not Levi Ackerman from AoT

@@wcg3928 oh I got autocorrected lol

repent unto God

Ikr xD

The enemy can't predicy you if you don't know what you are doing, tsunzu probably

repent unto God

15:50 for anyone wondering where the sh*t talking is

It'd be cool. You can't use either of them though, since they both rely heavily on traversing the game tree, which you can't do with so much imperfect information. You look one move ahead and the game tree instantly scatters and you're left with nothing. Alpha Zero could possibly make some tiny sense of the game since it consists of two components and can therefore remove one (the Monte Carlo Search Tree) without driving itself into complete irrelevance. It would then rely completely on its neural network making predictions for the next move only and would most likely as a result be extremely stupid and light years from the Alpha Zero we know. Especially since its psychological perception of the game is gathered through self play, so it would probably be pretty bad at predicting the moves of a human player. On top of that, it doesn't track the move history, since it's irrelevant in a normal game of chess where all you need is the current board state to make your decisions. In a game like this (similar to poker), you need the move history in order to look for patterns that can help you get a broad sense of the board position.

Well, AlphaStar managed to play StarCraft, which is mechanically+environmentally more complex than chess due to having fog of war and being real-time, and so was the next logical environment for DeepMind's exploration of AI playing games. So DeepMind has already explored an learning neural network AI traversing an environment with incomplete information. That said, there was also a sort of consensus about AlphaStar's performance in StarCraft that it still wasn't quite truly the top grandmaster-professional player level in terms of tactics, but rather just that it was able to brute force the best of players through raw, mechanical skill in its management abilities, and that anyone that could execute a unique enough plan solidly enough could still blindside the AI if it wasn't something the neural net could account for (even over thousands of training hours and years worth of knowledge). However, given the more limited scope of the movements available in chess compared to a video game RTS, I'd imagine that a DeepMind-like AI would be able to handle it rather well, once it was given enough information and training time.

@@TheAmazingCowpig Or said in another way: (I might have been a bit foggy in my explanation) Starcraft is close to being deterministic, meaning that you can look hundreds or even thousands of "moves" ahead and have a pretty good sense of what the game state looks like. On the other hand looking just 2 or 3 moves ahead in this chess variant and you have absolutely no idea of what the game state is. You could be up 3 pieces or you could be down a queen.

muchograndeyolatengo I don't think Starcraft would be significantly more deterministic than Fog of War chess. Any time a player does not have vision of their opponent, they have incomplete information about how to act; information gathering is a core facet of all strategy games, from this version of chess to video game RTSs. 2 or 3 moves in FoW chess could be equated to 2 minutes into the future in a Starcraft game. While the exact rules and consequences differ, DeepMind has already created the AI that tackles having incomplete information about the game state (as a progression from AlphaZero that did have complete information about the game state at all times). So, a DeepMind AI for FoW chess would probably look something similar to AlphaStar.

​@@TheAmazingCowpig The only incomplete information the AI has in Starcraft is the opponent's build order. Everything else is perfectly visible (including all the battles). So it can excel at everythign else through self training (build order, resoruce management, micro etc.) And if I recall from the matches I've seen, the AI seriously lacks in strategic understanding but makes up for it in perfect micro. We are talking about two completely different domains here. You can' really represent a game of FoW chess with a tree. It's just not possible. The only thing you can train is a standard opening, but that's really exploitable unless you add noise to the input vector or do some other kind of magic. You can't train it to understand the move patterns of human opponents of different strengths through self play. Because humans are not gonna think the same way, which again makes it inherently exploitable. But we're talking in circles here. I'm not gonna convince you and you are not gonna convince me. So I'll let you have the last say and then I'm gonna retreat :)

The classic toxic super GM kekw

naa man, happens in poker all the time, its legit reasoning, you literally can win because and only because you played badly, now that deoesnt mean playing badly is a good strategy because it isnt in the long term.

@@mcpartridgeboy well unless you reason that playing 'badly' in a certain situation will mislead the opponent in a way that makes him lose, then it's not playing badly.

@@sagov9 It is definatly playing badly in poker , its just you need to ajust for that, like in poker when your playing bad players its pretty simple you just wait for good cards and value bet, maybe not semi blind chess but its hard to believe a solid game plan is worse than a goofy move.

@@mcpartridgeboy You're missing my point. If you know how the opponent thinks, you can use 'suboptimal' play to mislead him. It's a valid strategy in games like this and poker, and not 'goofy' at all.

repent unto God

I don't think he speaks English

@@ijemand5672 I laughed way too hard on this reply 😂

repent unto God

@@ijemand5672 lol😂😂😂

yep, i clicked because of that

Yellow guy was OP

@@achillesmichael5705 wait what does he do again? Like all his units are stronger, or smth? All i remember is that his co power gives me headaches.

My god the Nostalgia.... I loved that game

@@nvat5569 we all did u.u

Xd

when she oneshots your adc and leaves

Wtf, lol comments here?

u mean flanked? coz backdoor is taking the nexus right.

@@neilpanganiban272 nah, you can backdoor as evelynn cuz she can go invisible given that the inhib and nexus turrets are down

The queen effectively becomes your scout, where it's better to use it to get valuable info than take things or pressure.

repent unto God

He's trying to access that part of his brain like: 👀 🧠 🔚

Ceiling demon

@@kingjensen8091 using stockfish on the ceiling WeirdChamp

Lmao

Hahahahaaa

I was thinking something similar the other day. About the fan made pieces like the unicorn and pegasus and how you can add more stuff up till you end with a MOBA or RTS game 🤯

lol

When we going to get a chess anime?xD

@@mightybatillo that's the downside of becoming popular nowadays, if you consider it a downside.. On the bright side, Chess is becoming more popular.

Wdym yugioh xD

I peeped that too 🤔

Less exciting, but I was wondering about chess with sorta "beginner intuitive" rules. No en passant, no castling(!), and the queen can also move like the knight.

oO PPH Oo That sounds even worse than the one pawn that can move like a king

@@nikodemg6608 Full disclosure: I'm not aiming for a fun mode. It would be an experiment to see exactly how much those rules matter. Probably a very messy experiment. Until quarantine began and I jumped on the twitch chess hype, I didn't actually know how castling worked (though opponents of mine had used it) and I had never even heard of en passant. I'd be curious to see the "meta" with those sorta rules.

I'm elated Mr Hikaru ... Big fan.

What

and still lost

Believe it or not that's a thing....if both players are playing blindfold

Thats blind chess isnt it mate

@@SnowFlameSupremacy yeah but this time you don't get to hear the chess notations

@@sachinm1882 No, you don't get to see the highlighted squares

It doesnr have as much range as the bishops or rooks tho, so its a tradeoff

@@piggy8761 still, its a very important asset in fog of war

@@trojanchess4125 yea

Knight is basically the trump card as lomg as you scouted their pieces well. So it's better to conserve the knight for a possible game changer.

Oh it sounds complicated and funny

I feel like he'd do really good with the strategy and macro, but would get stomped in the micro game, as he hasn't trained for starcraft, and only knows chess

lmao u watch this stuff?

it could be played with one extra board and a game master

Disappointed by the number of retards in the world*

He is seeing Hikarus side.. wondering no one in comments sawthat