Hacker News new | past | comments | ask | show | jobs | submit login
Can Chess Survive Artificial Intelligence? (thenewatlantis.com)
51 points by pseudolus 53 days ago | hide | past | web | favorite | 54 comments

Chess is today more popular than ever, with online playing platforms like chess.com or lichess.org, where you can find an opponent in seconds and easily analyze the game afterwards.

The short time formats like Rapid 15 minutes with 10 seconds increment per move is my favorite, but the most popular format is Blitz, which is say 5 minutes with 5 seconds increment.

You can play even bullet of 2 minutes with 1 second increment, or even no increment formats. You can play chess anywhere, at the bus stop, in the dentist office, take video lessons, do practical drills, puzzles.

Via the rating system, you are always matched to an opponent of near-equivalent strength. The rating system is build so that a 200 point difference means that the higher-rated player will on average win 3 out of 4.

Most opponents are less than 100 points apart, so it's typically a 50/50 chance that you will win a given game.

Chess only looks dull for those who don't play it, it can be super exciting. But the classical format, with a physical board, and 5 hour games? I think that will become less and less of a thing.

On the contrary I promised myself to only play chess over the physical board.

Playing online started to feel like an addiction. Most people end up playing blitz or rapid which even reinforces the addictive behaviour. Over time I felt like there's not much difference in the way I play chess to playing any other online game.

But there's nothing better than meeting a friend in a cafe and have a good game over the board.

Such game forces you to focus even more because you have that friendly rivalry going on and you've been keeping a score. I can imagine real competitions are even more intense.

That's something different than playing against some anonymous nickname online.

Paul Morphy said "ability to play chess is a sign of a gentleman. The ability to play chess well is a sign of a wasted life"

I personally get much more of what I expect from chess when playing in person. Playing online has become a compulsive behaviour and quite a meaningless activity once you reach certain level in ratings - the point when you learned the most chess can teach you about life and strategy but getting better requires harder and harder work while lower and lower return other than bettering your ratings.

So no, for me the classical format must live on else chess will become replaceable.

PS: I encourage everyone to grab a board, go to the nearest cafe or wherever, open a nice bottle of your favourite beverage and have some good battle with a friend over the board.

I have a decent amount of friends. Of the ones who live nearby, none play chess.

There are some chess players in my city, but I haven't the time for that level of commitment.

Playing online via an app with up to 24 hours per move and in-app chatting is best for me. Sure the person can cheat, but I assume most would rather play fairly and win of their own accord rather than a meaningless win via cheating.

I got a friend of mine into chess. First, I beat him easily.

Turned out he's competitive and while I was abroad for 6 months he trained hard to beat me.

When I got back the games became competitive and while playing chess we ended up having a business idea and we started a business.

Now 6 months later it earns both of us about what's an average salary here + pays another employee.

He in turn played his cousin while I was away. They got together after since childhood and started going rafting and other adventures.

I know a bunch of stories like that.

All sorts of things can happen that won't happen online. Of course I played online too, and them also ... to improve.

I think even if you think you don't have friends it's good to just play anybody who's down for it. Or join a chess club generally inquire who knows chess when you meet some new cool people.

The paradox is that something can be "dying" and "more popular than ever" at the same time.

I've been using lichess training for the past few months; doing at least 50 puzzles per day has improved my game so much lately. You can do 50 offline so it's great when you're riding public transportation or taking a break from work.

A great example at how computers affect chess commentary is the currently ongoing tournament in St. Louis. On youtube, the game is broadcast by people using an engine to evaluate the position they're looking at, and on twitch, two grandmasters are providing commentary without an engine. The youtube stream can be easier to follow, but it takes some of the drama out when the story appears to be "look at this guy, missing the best move when he's #4 in the world, how could this happen?" for every game. On the other hand, the grandmasters' commentary might be closer to what the players are actually thinking (what would be deemed "human moves"), but sometimes even they aren't sure what exactly is going on.

https://www.youtube.com/user/STLChessClub https://www.twitch.tv/stlchessclub

It's a shame Alejando Ramirez isn't on comms for this one, as I think he's the main draw for STL commentary for me - extremely well versed in current theory, able to balance human and computer analysis perfectly, and (with all the love in the world) doesn't bumble about as much as Yasser Seirawan.

Of course, the dream team remains Jan Gustafsson and Peter Svidler over on Chess24 when they're able to get together:


Ramirez is doing the twitch stream, though you're right that he was able to balance out how the computer analysis affected his commentary. OTOH, I felt that as an event dragged on, he'd rely more and more on the machine to tell the story of the game. Conversely, it'd be nice if they made Maurice Ashley comment without the engine for a while, maybe he'd have more interesting things to say again.

The 'AI beats worldchampion in ...' is a story that has diminishing newsvalue as time progresses as people will just come to expect this to be the case.

Competitive games survive AI not because humans are a good match to the AI, but because people like playing other people, or watch humans playing each other. A Dota 2 championship does not wane in interest because there exists an AI model that can beat most human teams. Cycling races did survive the advent of the motorcycle just fine.

What does change is the scrutiny that needs to be deployed for catching fraudsters using automation assistance in human competitions. But is looking out for a hidden computer or communications device in a chess tournament different from looking for hidden motors in a cycling race?

>A Dota 2 championship does not wane in interest because there exists an AI model that can beat most human teams.

I just want to make this clear, because it's something that a lot of people misunderstand, there exists no such AI currently. Not even close. What exists is an AI that can play with and against 17 out of 115 heroes, which allows a tiny fraction of the possible hero combinations, and those 17 heroes are specifically chosen because they are relatively AI-friendly. This AI is also only extremely strong if you're not allowed to repeatedly play against it and figure out 1. where the AI is exploitable and 2. what picks/items/playstyles are even good in this 17 hero meta that no human has ever given much thought before. Plus, as it stands now, the AI has quite the interface advantage, since it doesn't just get the rendered screen as an input, but the internal game values through an API - which allows much more precise reactions.

I'm not saying that the OpenAI achievement isn't great, it absolutely is and I enjoyed the matches quite a lot (even if I don't really play DotA 2), but we are a long, long way from AI supremacy in DotA 2. And if we reach it, we have no idea how it is going to influence the popularity of DotA 2 competitions. After all, younger people playing chess now already grew up with the idea that they will never be able to even come close to the strength of a computer, but DotA 2 players have seen bots as useless pushovers for all their lives.

You are absolutely right and I did not mean to say that the current OpenAI Five today can beat most human teams in a full DotA 2 game. My point is I do not believe the eventual existence of such an AI player would impact the interest in Human to Human competitions, either as players or as spectators, for that game.

Yes, but its in a new form.

Hardcore "no rules bared" chess is correspondence chess, where you have 24-hours between moves and can use whatever means necessary to make the best move.

If you think just buying a supercomputer and running Stockfish (or LeelaZero) will win, then you don't know anything about opening databases. All the good correspondance chess players have already run Stockfish / LeelaZero across many different openings, and have collected statistics on which openings will lead to situations where Stockfish (or LeelaZero, or Komodo, etc. etc) will inevitably fail.

These computer AI chess programs have weaknesses. A "Cyborg" team, composed of a human guiding Stockfish (or LeelaZero) through opening books can get further than any computer alone, and likely push the opponent into a subtly failing position.


The article talks about it a bit in Centaur Chess section. That's probably where the biggest challenge lies in the future. Centaurs have access to the 8-man database, you keep opening-books secret, and you spend time studying how Stockfish, Leela Zero, or Komodo (or other engines) makes mistakes.

Or, you write your own engine, and try to use that as an edge.

> These computer AI chess programs have weaknesses.

> ... have collected statistics on which openings will lead to situations where Stockfish (or LeelaZero, or Komodo, etc. etc) will inevitably fail.

In case of Stockfish and other "conventional" engines, yes.

However, with engines like LeelaZero those weaknesses gets much harder to exploit, since you can create your own version of Lc0 by training a new model with tensorflow.

LeelaZero struggles with endgames. I'd probably use LeelaZero through the beginning and middle games, but switch to Stockfish whenever I feel like a tactical situation comes up, especially if the position starts to drop down to 10-man or 9-man. 7-man Syzygy Tablebase are less than 10TB, a RAID-array of M.2 drives will access those quickly. If Stockfish evaluates any 8-man trade, it will simplify into a 7-man situation and you have guaranteed perfection after that.

No one knows how to integrate the endgame tablebases into LeelaZero by the way. So if we both play LeelaZero for the first 40 moves, but I switch to Stockfish + 7-man Tablebase for the ending, I'm probably going to beat you. LeelaZero's endgame really is quite weak.


If you are a programmer, and can figure out how to integrate the Neural-Net evaluation into MCTS to evaluate tablebases, you probably would do great. Its currently unsolved (to my knowledge).

It seems solvable... I don't see any reason why MCTS couldn't have a if(popcnt(bitboard) < 8) doTablebase() statement somewhere. The question is how to do it as quickly as possible, and how to do it asynchronously and integrate the statistics to the rest of the MCTS tree.

BTW: A research team has created the 8-man Tablebase, although it takes probably 100s of TB to store. This means that 8-man Tablebases will remain as Hard-drive only for now.

I like the opposite: only with physical presence, no consult, no studying openings just you playing your best game against your opponent doing the same according to the classic time rules.

It comes down to what do you value? Do you value playing "your" best game vs "your opponent's" best game? Or do you value playing the best game possible, using whatever tools necessary to do so?

This philisophical split is also in the speedrunning community. The tool-assist players aim for absolute perfection: what would a "perfect run" actually look like? They make recordings at 1/60th of a second, and have a computer input the sequence of buttons with absolute perfect accuracy.

Or, are you a human-only player, who values skill and training?

There's something to be said about the two sides. But I find myself preferring the "theoretical perfection" side... as just my personal taste. I can see why people prefer the skill-and-hard work approach though.

Sorry for the belated response, I've been thinking about this for a bit before answering because I think it is a subtle matter. I think the reason why I don't go for this 'theoretical perfection' angle is that it takes the fun out of the game. That's where computer programs, cold and emotional will not feel the joy of a win, their programmers might but they might just as well not even be aware of the game being played. The only way to get that feeling on both sides is for unaided humans playing other unaided humans.

The 'perfect game' is something that a computer will one day find, a curiosity but just like a math proof ultimately just a thing waiting to be discovered, not achieved.

I forget who this is attributed to, but someone proposed a method of beating chess - correspondence chess that is. Simply set up a chessboard, mail a grandmaster for white, and a grandmaster for black, then have them play against each other.

I remember this variation of the story:

An unknown chessplayer bet a lot of money, that he could play live & simultaniously against two grandmasters and achieve at least 1 point. The two matches were played in different rooms next to each other. No one except the unknown player was allowed to switch rooms during the matches (to not disturb the grandmasters). Obviously he played White against one grandmaster, Black against the other, and won his bet.

As I sometimes tell this anecdote, I wonder what its actual source might be?

I'm pretty confident, this never happend in real, as I expect no grandmaster could get scammed that easily. But the story could be rooted in a "catch me if you can"-like movie or novel or something.

If someone knows more, I would be happy for a hint...

It's a very old trick I'm sure. As for it never happening or fooling anyone in real life: the magician/mentalist Derren Brown challenged 9 English chess players, including 4 GMs, to a simul on his TV show, and got a plus score - he secretly had 8 of the players playing each other, with whom he thus had an even score. The 9th was president of a university chess club, and a very weak player, who Derren played and beat, winning the match.


Might not be exactly what you're thinking of, but Derren Brown did similar against a group of players


> The impact of computer chess on the game — as still played by humans — has been twofold. First, computers have helped to flatten chess, increasing pure understanding of the game at the expense of creativity, mystery, and dynamism. Second, they have become intertwined with every aspect of chess, from play at the highest level to amateur study and the spectator’s experience.

This speaks more about the players than computers. It is still very possible to play ‘human’ chess at an extremely high level. Look at the number of London systems at the top level, the way Carlsen decides to play a Czech Benoni against a lesser grandmaster. He won that game out of human skill, not computer mastery.

The human element of chess is alive and well. The computerisation of chess started to sky rocket when computers and humans were at similarish playing strengths. As computers have far overtaken humans at chess now, I think chess will start to humanise itself once more.

Will the 100 m dash survive the car?

I think it's different. In physical sports, we already have categories, because we have to overcome physical limitations. For example we have separated human sprinting into Male and Female categories forever. In boxing or martial arts we even have weight categories. So you can be good "for a woman" and it doesn't carry much stigma.

For chess, it's supposed to be a purely intellectual game, and even though there are categories, the top women can play with the top men. This wouldn't make sense in 100m dash where the Women world's record is beaten by highschool boys.

Essentially this places the entire human species in a second-tier category. Human champions are now good, "for a human". Humans are in the lower league and we must make sure they only play between each other because of their inferior cognitive capabilities. They are no longer competing for the genuine title of best chess player in the World, they wouldn't even qualify to enter the tournament. The World best human in history can be beaten by the equivalent of a highschooler machine player.

I think the comparison with the 100m dash is very apt.

We have machines that can obviously beat any human runner but the point of the race is to race other humans, not machines.

The point and fun of chess will remain to play against other humans.

Yes, but if you cheat with a car in the 100m pretty much everyone will know. Although this has been done with longer races. ( https://en.wikipedia.org/wiki/Frederick_Lorz)

Another related issue is races with amputated sprinters when the prosthetics provides an unfair advantage because it's lighter, has better mechanical properties than a biological leg, makes the runner taller (longer strides) and doesn't have lactic acid limits.

This hasn't quite happened yet but it's getting closer in the 400m and I expect long jump will also at some point have disabled athletes perform better than able bodied ones (Disabled long jumpers all jump off their prosthetic leg). The debate about allowing disabled athletes into the "general" events is growing.

It might be better and more affirmative to lose the disabled label and call them cyborgs, then give them their own category.

Cars don't change how humans run. Computer chess does change how humans play. While yes, the interesting part is watching humans play humans, the concern is that computers will influence players, be it in chess or competitive videogames, into less interesting play. I think that's a real concern as artificial intelligence improves, though right now I don't think it's a problem.

Yeah, a better comparison would be the triathlon, if instead of one part being on a bike, you instead got a motorbike and every year new tech made that faster. Sure you still have 2/3 down to humans but 1/3 is sort of automated.

But as you say, this doesn't seem to be an issue currently, and 2019 has been quite exciting in classical chess. I think right now there are more potential openings on the table than any time this century, as new engines (re)introduce lines that Stockfish didn't think highly of, and that can continue for quite a while.

Surely people don’t use computers WHILE playing? So it’s like having the coach ride in the car, yelling at the runner while he practices. Quite innocuous.

Sure, and remembering 20 moves of computer prep in various different lines is a skill in itself, and knowing _which_ lines to prep or innovate in is part of the metagame. But it still feels a _bit_ unnatural, no?

Well yes, but not really much different from memorising openings from a book. Personally I find all kinds of professional sports and competition a bit boring, it was more fun when people were amateurs.

As long as computers are only used for training, what will change with improved AI?

you mean, if you tie a rope from the car to the human runner behind and see whether he is able to keep up the pace with a reasonable running gait or end up being dragged and tumbling on the ground?

Running doesn't involve wheels.

I imagine there will come a point where algorithmic chess (I wouldn't call it AI) results in perfect play, where every match ends as a draw. Much like the scene in War Games where the computer plays tic-tac-toe against itself and every game ends in a draw. Given sufficient resources, and the fact that there is no chance at all in chess (save choosing who moves first), it seems to me it's only a matter of time before every move is the perfect move. That would mean one of two things: either every game ends in a draw, or every game ends with white winning.

The end result of this is that every game will be the same. Given two effectively infinitely-skilled computer players, there's no more game to play. There's always a best move, and the best one could hope for is the draw (or, again, white would always win; I'm not sure if it's been proven that first move guarantees a win with perfect play--but it's one case or the other).

Human vs. human will remain a compelling game, both for players and observers, but computer-vs-computer will cease to be interesting at all. As will human-vs-computer, except maybe as a way to see who can survive the longest before losing or drawing; no human can win against the best computers even today, and never will again.

> I'm not sure if it's been proven that first move guarantees a win with perfect play

It has not, chess is an unsolved game: https://en.wikipedia.org/wiki/Solving_chess

In this scenario human vs human is not very compelling when all they do is make mistakes. The only analysis left is how bad their mistakes are.

There are positions in chess that have been disputed for decades or more which are just now coming alive because of the ability of engines to analyse lines. The thing that's interesting about high-level chess is how players play out particular strategic themes that result from known positions (eg initiative vs material, the open H file for white in the London system vs not being able to castle kingside, play in the center versus counterplay on the flanks etc etc etc).

My recommendation would be watch a really good game analysis like Agadmator's youtube channel. One like this might be fun to start https://youtu.be/vztVylqnBSo or one of the Capablanca vs Lasker games. Alternatively John Bartholomew and Eric Rosen have fun channels where they play and commentate on their thought process.

Good to see Agadmator getting a shout out

I personally think this is an extremely exciting time in chess, especially when you take into account the surprising amount of doom and gloom that occurred around the World Championship. We've recently moved from long, boring theoretical battles in the Berlin or Petrov to revive various lines in the Sicilian, and Magnus Carlsen's classical revival since retaining his title seems to be based on very interesting positional sacrifices, instead of just grinding out wins from mostly-equal positions as he may have done in the past. It's hard not to want to acknowledge the invisible hand of AlphaZero (and indeed Leela) in these changes. I'd like to think that opening theory is going to go through something of a renaissance as more top players get access to these new engines with a better long term strategic sense of the game. It's hard to believe we're going to run out of new, interesting positions in our lifetimes.

Between those changes at the board, and the wider changes of people pushing chess as an esport, and growing prize pools from St Louis and the Grand Prix circuit, there's never been a better time to be a player or fan at any level of the game.

The answer to the question is yes, most chess players have accepted computers play the game better and don't care. High level chess will eventually just change the rules if forced drawing lines make it too dull.

> High level chess will eventually just change the rules if forced drawing lines make it too dull.

If it hasn't happened already, it may never. People were complaining of the draw death of chess decades ago, and it's worse now than ever.

My personal theory as to why chess isn't that popular in the US - and why the US produces relatively few world-class players given its very large talent pool - is that chess at a high level involves a large number of draws. The last championship was entirely bloodless in the main games and had to go to tiebreaks (rapid + blitz games). At IM+ level more than half of games end with draws. And this goes contrary to the US mentality where everyone is either a 'winner' or a 'loser' and competition is seen as a highest ideal. A game where most of the time it is impossible to tell the better player must feel frustrating to amateurs. And indeed, many many many times some laymen will complain that there are 'too many draws' and 'the rules must be changed' so spectators see some blood.

Another common complaint related to draws is the number of 'grandmaster draws', or 'comfort draws', whereby two players, usually at a high level, agree to leave it at that after a low number of moves (Chess is one of the few games featuring 'draw by agreement', meaning both players may agree to a draw at any point in the game without justification). Almost every grandmaster is guilty of that, and players who do that often get derided by their coaches, the media and amateurs: how dare you, you are a disgrace to the sport, I didn't pay to watch this, etc. I daresay the complainers are either hypocritical or have never been involved in high-level long-form tournaments (7+ days of 4+ hour games each plus rapid/blitz side events etc.). Sometimes both players feel tired and don't want to play it out (usually due to having played it out 6+ hours the game before) and objecting to that is just unrealistic. Chess isn't a gladiatoral arena. Then there are the famous accusations of collusion from Fischer against his Soviet counterparts but Fischer has always been slightly mad his whole life so I wouldn't take anything he said seriously.

It's a bit like asking if athletics will survive the cars, because no one can outrun a car. Well, as long as you play chess against other humans AI is irrelevant, isn't it?

Anyone can outrun a car... In a sufficiently dense forest. AI always have had this problem that reality is a very dense forest.

Alternatively, NASCAR races (seeing things zoom around a track at 200mph) is also really fun. So we have a race for tool-assist, and we also have a race for human-only.

Yes, until we get to the point where Chess is solved and some AI can say

1. D4 Mate in x.

I think until that happens, chess is fine.

Came here to find this comment and upvote it.

This article misses the actual problem. It suggests that computers are gradually yielding a perfection of play which is implied to lead to a draw. But that's not really what's happening. In fact computer chess now already has fewer draws than top level human chess. The issue is that chess is unimaginably vast. The article talks about people having the first 15 moves prepared. That's true, and false. In each position there are somewhere around 30 legal moves on average and 3-4 decent moves. That means after just the first 15 moves there are 3.5^30 reasonable moves to explore. That's 20 quadrillion reasonable options. And obviously the entire search space, 30^30 is rather larger.

No human has explored more than a minuscule fraction of that domain, and today computers are already much stronger than the strongest humans. In times past this was not such a big deal. You play what looks good or promising, and even once you get outside of your preparation it'll still be human against human or, at worst, human against team of humans. But this is where computers now come in. Now if you enter into a complex position that your opponent has analyzed with a computer and you haven't, you stand a very real chance of losing without your opponent ever having to think for himself or you having any real chance.

As a result of this players are increasingly preferring lines which tend to be very "solid" with much less room for your opponent to find 'surprise moves' or novelties. And even if they happen to find such a move you stand a good chance of being able to somewhat safely navigate the waters since the position is, again, "solid." The problem should be obvious. This sort of chess is extremely prone to draws.


You can even see this in chess events when a player loses. Modern world championship matches are very short (usually 12 games). So a single loss is huge because you just don't have much time to catch up. As a result once a player loses the calculus of the match significantly changes and now even if you do get caught out - well you were going to lose anyway. And so the players will put aside computer fears and return to more 'traditional' views of chess. And as a result you often see a decisive game end up being followed by another decisive game. At worst one decisive game is often followed by much more interesting games.

This is one of the reasons there have been suggestions like returning to draw odds in favor of the champion (as used to be the case in some matches). Now a days instead to resolve draws in the slow chess we go to increasingly rapid time controls and even, at the extreme end, a times-odd game where black gets draw odds. As variance ostensibly becomes larger in those sort of games, the challenger is often happy to just draw the probably stronger champion in slow chess and hope to get lucky in the fast tie-breaks. Other ideas have been to play the tie-breaks before the real match starts, or to have a tie-break for every single game so they're always decisive. Again, it's always the same goal - to remove drawing as an acceptable outcome for at least one of the players.

You can also see this exact affect even in tournament events. When there is no leader all the games tend to be draw heavy because the calculus is directed towards the value of 'not losing' over winning. But once a player starts to pull ahead of the pack, the calculus starts to change more towards winning. And so one player winning ends up triggering a cascade of decisive games or, again, at worst much more interesting games.


The problem isn't computers - it's fear of computers and somehow removing the value of draws in chess. But nobody's found any really great way to solve these problems.

> removing the value of draws in chess

Armageddon chess (Black wins in all draws) is one such solution, although it leads to a major Black advantage.

But there are 4-kinds of draws: Repetition, Stalemate, Insufficient Material, and 50-moves. No other draws exist in chess.

Perhaps instead of giving Black a win in all kinds of draws, maybe Black should win in two or three of the draws (Black wins on Stalemate and Insufficient Material), while White wins on Repetition and 50-moves.

This suggestion hasn't been studied yet. But it seems like the natural progression after Armageddon chess.

Note: This is not my idea, but one being discussed somewhere else. Its interesting though, so I like it.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact