human vs. machine

I wanted to address this to fellow Chess players here - even though it references the game of Twixt.  I am an experienced, rated Chess player, and something has been bothering me for a few years now ...

Chess software has risen to rating heights easily surpassing humans.  I watched, as first they achieved measly 2900 strengths (accessible to 'some' humans), and then proceeded into the 3000 + territory.  Humans were getting CREAMED in 'fair' matches (I have a point coming); remember how Hydra demolished Michael Adams?  How about when Nakamura lost an 'odds' match (so computers can spot a 2800 human a KNIGHT, and still win!).  I always thought that these matches were UNFAIR, instead of being played fairly.

My point is that 1. Humans were NOT allowed assistance, say, of opening encyclopedias, endgame databases (Reuben Fine would have been nice), etc. 2. the computation speed is a mismatch: given the speed that computers can process data, why should humans have been limited to the same time controls?

Chess is a fascinating, DEEP game full of nuance and complexity.  I always had a feeling that humans could BEAT computers, if we eliminate their 'unfair playing advantages'.  Here is what I'd like to see:

1. Human has free access to complete opening databases; computers ALREADY have this data anyway, so this is perfectly fair.  Remember Kasparov's famous loss to Deep Blue in a tricky line of the Caro-Kan?  This doesn't happen if he has a database.2. Likewise, human has access to endgame table-bases, which computer already has.3. The 'game' remains fair - no knight odds, or pawn odds, or move odds - BUT, the fact that computers process at the speed they do means humans should have UNLIMITED TIME!  Time pressure means nothing to a computer - it processes so fast, it ALWAYS has time available.  Why should the human have any less?4.  Here is the klinker: let humans 'cheat'!  All I mean here is that humans should be free to consult with each other, team up, etc.  This is not 'really' cheating, if the point is to prove that humans can analyze Chess positions in a way that algorithms can not match, in a FAIR PLAYING FIELD.  I believe (or at least HOPE) that human's understanding of positional play will surpass the computer's tactics - once we eliminate human error (the point of the teams), time pressure (no, or generous time control) and knowledge (databases and table-bases).

I would love to see a team comprised of Carlsen, Nakamura, Caruana and Ding play against Stockfish or Leela Zero under THESE conditions.

WHAT CAN I OFFER AS EVIDENCE?

I don't know how many of you followed my recent experiment over at the Twixt play area.  'Twixtbot' had been recently upgraded by a talented programmer to use the same sort of processing as Alpha Zero (champion in Go, Chess, etc.).  As a result, the Twixtbot was CLEARLY DOMINATING all of the best human players.  It lost very infrequently, and if you want, you can see the records.  It's record against our former champions is, frankly, a bit depressing.  But ALL of these games were played under 'equal' conditions, and in my opinion, these conditions were unfair.

So I arranged a match were humans could TEAM UP and consult, move after move, against Twixtbot.  Some of our VERY BEST PLAYERS competed, while I made sure that we moved within time limits.  Each move was thoroughly analyzed; there were NO tactical blunders to be made.  The players worked as a TEAM, finding best lines of play while denying Twixtbot any useful counter play.  IT WORKED!  The humans dominated, and Twixtbot resigned in a hopelessly lost position!

It should be noted that since that game, the Twixtbot continues its domination of humans - including the best players here.  I don't think that means much, except - human players have not really LEARNED very much from the game; we dominated the Twixtbot once, but nobody comes close to repeating (excepting one recent draw).

WILL THIS WORK IN CHESS?  I think it will - or is at least, worth a try.  I am reaching out to our community here for comments: DO I HAVE SOMETHING HERE?  I think it is possible that humans, under the right conditions, can defeat the best Chess software out there.  I offer this Twixt game as evidence.  If people think that this really has a chance, I WILL REACH OUT AND COMMUNICATE WITH WHOEVER I MUST - really - I will send emails to Carlsen, Nakamura, FIDE, Chessbase, and anybody else who will listen.  I will tell them about my experiment here, and propose a match under the conditions I described.  Maybe somebody will go for it - play the game - beat the computer - and this game will be ONE FOR THE AGES.

Or - maybe not.  Perhaps I am being way too optimistic.  I welcome your comments.(meow)

see the broken document symbol above?  The LG forum EDITED OUT a big chunk of my above post!  It does this sometimes - I don't know why.  Fortunately, I had word processed the entire statement, and can re-post the missing material.  I hope this survives the edit montster.  After my words "4. Here is the klinker ..."  but up to "The humans dominated", below is the missing material.  I hope this makes my post more understandable. and I really with Richard would STOP DOING THIS TO ME!

4.  Here is the klinker: let humans 'cheat'!  All I mean here is that humans should be free to consult with each other, team up, etc.  This is not 'really' cheating, if the point is to prove that humans can analyze Chess positions in a way that algorithms can not match, in a FAIR PLAYING FIELD.  I believe (or at least HOPE) that human's understanding of positional play will surpass the computer's tactics - once we eliminate human error (the point of the teams), time pressure (no, or generous time control) and knowledge (databases and tablebases).I would love to see a team comprised of Carlsen, Nakamura, Caruana and Ding play against Stockfish or Leela Zero under THESE conditions.WHAT CAN I OFFER AS EVIDENCE?I don't know how many of you followed my recent experiment over at the Twixt play area.  'Twixtbot' had been recently upgraded by a talented programmer to use the same sort of processing as Alpha Zero (champion in Go, Chess, etc.).  As a result, the Twixtbot was CLEARLY DOMINATING all of the best human players.  It lost very infrequently, and if you want, you can see the records.  It's record against our former champions is, frankly, a bit depressing.  But ALL of these games were played under 'equal' conditions, and in my opinion, these conditions were unfair.So I arranged a match were humans could TEAM UP and consult, move after move, against Twixtbot.  Some of our VERY BEST PLAYERS competed, while I made sure that we moved within time limits.  Each move was thoroughly analyzed; there were NO tactical blunders to be made.  The players worked as a TEAM, finding best lines of play while denying Twixtbot any useful counter play.  IT WORKED!The humans dominated, and Twixtbot resigned in a hopelessly lost position!

(continue reading from this point above)mc

amazing; it did it again - removed the same paragraphs!  OK, I will go back to my word processor and change the text style; I will also edit a few words that perhaps he censored;  there is no way for me to take back posts.  If this does not work, I will try something else, but sorry for the mess I am making and the difficulty reading!

mc

4.  Here is the  k l i n k e r : let humans ' c h e a t ' !  All I mean here is that humans should be free to c o n s u l t  with each other, team up, etc.  This is not 'really'       c h e a t i n g , if the point is to prove that humans can analyze Chess positions in a way that algorithms can not match, in a FAIR PLAYING FIELD.  I believe (or at least HOPE) that human's understanding of positional play will surpass the computer's tactics - once we eliminate human error (the point of the teams), time pressure (no, or generous time control) and knowledge (databases andtable-bases).

I would love to see a team comprised of C a r l s e n, N a k a m u r a, C a r u a n a and D i n g play against S t o c k f i s h or L e e l a Z e r o under THESE conditions.

WHAT CAN I OFFER AS EVIDENCE?

I don't know how many of you followed my recent experiment over at the Twixt play area.  'T w i x t b o t' had been recently upgraded by a talented programmer to use the same sort of processing as Alpha Zero (champion in Go, Chess, etc.).  As a result, the T w i x t b o t was CLEARLY DOMINATING all of the best human players.  It lost very infrequently, and if you want, you can see the records.  It's record against our former champions is, frankly, a bit depressing.  But ALL of these games were played under 'equal' conditions, and in my opinion, these conditions were unfair.

So I arranged a match were humans could TEAM UP and consult, move after move, against T w i x t b o t.  Some of our VERY BEST PLAYERS competed, while I made sure that we moved within time limits.  Each move was thoroughly analyzed; there were NO tactical blunders to be made.  The players worked as a TEAM, finding best lines of play while denying T w i x t b o t  any useful counter play.  IT WORKED!  The humans dominated, and T w i x t b o t  resigned in a hopelessly lost position!

wow; let's try this - I am going to break the material apart into small posts below.

4.  Here is the klinker: let humans 'cheat'!

All I mean here is that humans should be free to consult with each other, team up, etc.  This is not 'really' cheating, if the point is to prove that humans can analyze Chess positions in a way that algorithms can not match, in a FAIR PLAYING FIELD.

I believe (or at least HOPE) that human's understanding of positional play will surpass the computer's tactics - once we eliminate human error (the point of the teams), time pressure (no, or generous time control) and knowledge (databases and tablebases).

I would love to see a team comprised of Carlsen, Nakamura, Caruana and Ding play against Stockfish or Leela Zero under THESE conditions.

WHAT CAN I OFFER AS EVIDENCE?I don't know how many of you followed my recent experiment over at the Twixt play area.  'Twixtbot' had been recently upgraded by a talented programmer to use the same sort of processing as Alpha Zero (champion in Go, Chess, etc.).

As a result, the Twixtbot was CLEARLY DOMINATING all of the best human players.  It lost very infrequently, and if you want, you can see the records.  It's record against our former champions is, frankly, a bit depressing.  But ALL of these games were played under 'equal' conditions, and in my opinion, these conditions were unfair.

So I arranged a match were humans could TEAM UP and consult, move after move, against Twixtbot.  Some of our VERY BEST PLAYERS competed, while I made sure that we moved within time limits.  Each move was thoroughly analyzed; there were NO tactical blunders to be made.

The players worked as a TEAM, finding best lines of play while denying Twixtbot any useful counter play.  IT WORKED!  The humans dominated, and Twixtbot resigned in a hopelessly lost position!

OK, DONE.  YOU CAN READ THE MISSING MATERIAL, AND THEN GO BACK UP TO MY ORIGINAL POST, FIND YOUR PLACE, AND READ THE REST.  HOPEFULLY THAT WILL MAKE SENSE.

Richard, GET YOUR ACT TOGETHER AND FIX THIS BUG IN YOUR SOFTWARE!  YOU JUST PUT ME THROUGH HELL TO MAKE A SIMPLE POST IN A FORUM, AND I AM A PAID MEMBER HERE!  I DID NOT DESERVE THIS TORTURE!

I hope that all this has not distracted from the point of my original post.  I am still hoping for some feedback on my idea.

(meow)

@MisterCat , what were you trying to do that broke your posts?

Add an image?

Or a link?

Perhaps this one?:  The World vs. Twixtbot part 3

I figured that you'd be the guy to figure this out, YPERCUBE.  Answer to query - I did ABSOLUTELY NOTHING.  I was not attempting to add any image or any link.  All I did was word process a long note (using WordPad), and cut/paste it here.  You can see my attempts above, such as spacing out certain words that I thought might be censored (c h e a t), and pasting in the missing section.  Nothing worked until I broke the missing paragraph up into a bunch of little pieces and posted them separately.  I DID NOT DELETE OR CHANGE ANYTHING from my original word processed document, so you can read it above in pieces.  If it's THE LENGTH that LG Forum didn't like, then my attempts above to simply post the missing material should have worked; they did not.

I sent a complaint to Richard, but have not heard back.  This was quite frustrating to me - taking all the time to compose a post, and having this happen.  If you can figure out what happened, you'll have my thanks.

(meow)

And there's your problem: WordPad.

WordPad, Word and other fancy editors are useful to provide rich text formatting. The issue is that while they try to be clever, they mess up when combined with other tools, like a web site which usually serves HTML produced by the web server and  - sometimes - some additional markup processing.

Try to use some real text editors when you want to post at golem, say emacs, vi, vim if you fancy.

Or Notepad++, if you are on Windows.

OR try to copy text without any formatting if you have no other option and you have to use WordPad.

you are saying Wordpad did it, even though I personally attempted nothing fancy - no image, no attachments, and the entire document was done with simple text?  OK, I will actually believe you, since I am confident in your expertise at this stuff.

Gee, I thought Wordpad WAS a 'real' text editor.  Of course, I have Word as well - and I hardly ever use it, since it had so many bells and whistles that I don't need; Wordpad is a stripped down version of Word.  Of course I have Notepad - practically never use that, since everything comes out 'ugly'.

OK, the next time I try this - word process away, but before posting, I will CHANGE FONT to 'plain text', no formatting, which I can do with Word or Wordpad.  I will load my material as 'plain text'.  It might not look as good - spacing, etc. but I will trust you that it will work.

Do I REALLY NEED to get MORE text editing software now??  I'm getting old, my friend, and learning new software is something that I find to be a royal pain.  (practically killed myself to get Zoom working right on my phone; still don't really know what I'm doing, but I managed a few video calls, so HOORAY)

Thanks.

(meow)

No, I did not mean it was your fault. As above, Word and Wordpad are **rich text** editors (bold, italic, underline, links, etc), so their default behaviour (and output) is to preserve this rich text formatting. I'm sure thay have options to export only the text but it might be difficult to find.

Note: I suggested Notepad++, not the simple (and rather useless) Notepad.

A lot to agree here.

Machine vs Human. Let us think what we challenging?  Ability to count, ability to memory? It was already obviously known that human is not a calculator. Kasparov lost one almost won game by a blunder (if i am not mistaken). So what that game proved? That human is not machine? Omg, and we did not know it before? :)

I am not sure about timing question. Because time in our world is the main resource/value.  We cant just give someone unlimited time - for me it is nonsense. What value will have the human victory over machine if human will need for this lets say 10 years?

I would rather propose that in human vs machine games human should have ability to analyse. I mean with virtual board for example. To check moves. Computer does the same but does not have memory limitations. And as i see in such matches we dont challenge memory.

Is there any real life task examples where human limited in analysis? For example we are solving equation. The goal is to solve it and give correct answer. There is no regret if before we choose right method we spent some time on wrong ways. What is important that we finally have choosen right method and further answer.

Giving human debut and endgame book seems ok for me. But if we are speaking about top players - they dont have problems here.

Thanks for the remarks, Andrey.  I will incorporate your suggestions into my summary below:

- Humans get use of opening encyclopedias and endgame table-bases; this is fair, since this material is already built into machine.  It is true that top players have extremely strong opening knowledge and endgame technique, but it is NOT 'perfect'.  I mention Kaparov's opening error, but top grand-masters actually commit endgame inaccuracies all the time; I've seen a few occur in the current 'Magnus Carlsen Invitational Tournament', going on right now!

-Humans get to move pieces around on side chess boards, or virtually on computer screens.  Machine has perfect memory, and is already 'seeing' upcoming positions perfectly.

- I can think of TWO Kasparov losses which will be eliminated: the one where his opening knowledge of the Caro-Kan was faulty, so he played into an inferior position; the one where he resigned in a drawn position, assuming that the machine had analyzed all lines and determined it was winning; BOTH of these errors go away with opening books and human consultation.  There is no emotion or human faults involved.

- Humans deserve MORE time because their computational speed is vastly over-matched by the machine; I am not insisting on actually UNLIMITED time, but time enough for consultation, and using resources available. Say, for example, the machine is given one hour per move, and the humans are given one day.  Those time controls would work for a game played remotely; for a 'live' match, then perhaps the machine can be given 10 minutes vs. 1 hour for the human team.

- If the team of humans can successfully beat the machine, which I think is possible, I believe that there will be much learned about Chess (to everybody's benefit) and also about the limits of machine analysis; also, I guess there will be some celebrations, as to the accomplishment - the participants here in my Twixt experiment were all quite pleased with their achievement, and RIGHTFULLY SO.

More comments will be welcome here, before I take my threatened action - I will need to forward my suggestion to 'powers that be' in the Chess world.  I haven't figured out how I'm going to do that without using Facebook or Twitter (I have neither), but I'll figure something out.

(meow)

The human brain has 86 billion neurons working in parallel while the typical AI uses only a few processing cores. Hence its computational speed vastly over-matches the machine. :-)

MisterCat, it's not just time, it's also memory. We humans can't remember what we've already examined. We need notebooks in addition of unlimited time. Or a computer-age equivalent of a notebook for storing variations. In the end it's a lot of work. Yes with enough work you can even win against bots. :)

right, well, memory - that's why I proposed moving pieces around on real and virtual boards; you would go farther to suggest saving, for reference, positions arrived at, say, 15 or so moves into the future, for analysis and reference (which could be accomplished on simple computer chessboards); and the human team gets enough time to analyze these future positions.

But then, lazyplayer, your position would appear to be that, with all these advantages, human players WOULD score a victory against the best machines.  I am of the position that it is possible, but not guaranteed; I think that some would take the position that humans have NO CHANCE against the top software, like Stockfish, Leela, etc., regardless.

So am I being, in effect, naive?  If I were to communicate my idea to the 'chess powers that be', would the response be 'of course, you idiot - naturally if you give the top players that many advantages, they would prevail'!  Maybe I AM being naive, but then - to my knowledge, nobody in the world of Chess has ever TRIED this experiment.  Nowadays, when machines lose good games (usually against other machines), those games end up on YouTube with some IM or GM giving analysis to the fascinating play and instructive positions that were achieved.  In MY experiment, say the humans SMASH the machine - isn't that going to produce a game that is ALSO fascinating and instructive?  After all, the machines certainly don't go down easily!

?

richyfourtytwo, no disagreement with your Math; I guess that what I mean is that, despite the processing speed of the miraculous human brain, humans appear to be much less effective at retaining that knowledge, and using it all in a Chess game - because their brain (and memory) is kind of OVERWHELMED by too much data; the machine does not suffer in that way.

?

MisterCat, with UNLIMITED time and UNLIMITED memory, the human team can theoretically run the algorithms of the bots and play the same identical moves.

It's like, can you do multiplications as well as a calculator? Of course you can, but doing large multiplications is not a very pleasant job.

MisterCat, see my game against hex bot: https://littlegolem.net/jsp/game/game.jsp?gid=2158441&nmove=11

It has played a very weak "opening" because it has been configured to play random opening now. I've also asked him to give me a tip and I played g8 due to its tip. Now the question is: can I win from a position like this? Sure, if I check all the choices. But who pay me 8 hours a day for a week to do this job? Do you want to donate your money for human players? :D

lazy, Andrey talked me out of the 'unlimited' a few posts earlier; I am of the mind to say 'more' time meaning, enough time for a team to consult, and to devote deep analysis on chess boards.  Still, the approach now (of Alpha Zero, Twixtbot, etc.) is to play a million (number varies) games internally, tabulate the results of those games, and choose moves that have the highest winning percentage.  Even 'unlimited' time would not allow humans to accomplish that - they would all die of old age before playing through millions of games.

The older approach, used by Stockfish, is closer to what humans do - select moves on a programmed set of principles, look ahead and see if the selection can be justified tactically.  But the newer machines were winning, until machines like Stockfish (Fat Fritz would be another well known example) started incorporating the same ideas, blended in with their already programmed algorithms.  And here at the Twixt section, the old Twixtbot could play the game, but notably NOT WELL, when all it was using was a limited look-ahead, and moving based on principles.  It became the monster that it is when the programmer adopted the Alpha Zero approach.

So I don't think that a human team can realistically expect to apply the same algorithms and achieve the same type of search that the machines do.  My theory (the point of the experiment) is whether 'understanding' of the game, once tactical blunders are removed, will allow humans to beat the machines.

Remember David Janowski?  He was a great player two centuries ago - and his CALCULATION skills were unmatched by anybody; it is why he frequently defeated opponents with clever tactical combinations.  Frank Marshall, I suppose, would be another example of this.  But Emanuel Lasker  handily defeated these players in matches - because his understanding of the game was superior; not because of his calculation skills.

This is where I'm headed with all this.

?

MisterCat, by the way, speaking of unlimited timers, why timer in Littlegolem is fixed? I would like longer games. I always run out of time... :D Actually games could be extended to infinite timer, the idea is as follow: if the game hasn't ended yet, then it's scored as a loss for the player to move. This creates compulsion for moving without creating a time pressure.

MisterCat, What is a "tactic"? Is a 50 moves deep checkmate a "tactic"? You see the concept is too vague. You're neither wrong nor right because it's all too vague.

What you're probably trying to say is that humans play with better algorithms than machines. This is obviously true. The human algorithm are so much better that we're not that much weaker despite seeing far fewer positions. Still, if the difference in positions is enormous, then we stand no chance, despite having the better algorithm. This is the take home lesson.

Well, lazyplayer, since you don't play Chess here at LG, I'm not sure about your background in this game.

why timer in Littlegolem is fixed

Just about every game playing site that I know of (which, admittedly, is not that many) has some sort of time control, since it would not be fair to a player to have to wait ridiculously long for their opponent to respond.  Obviously on 'live play' sites, where both players face each other at the same time with only their computers between them, there MUST be a time control; when would people be able to get up to eat?  Many players now play Chess with faster and faster 'live' controls, such as 'Bullet Chess' - all moves in one minute.  Personally, I don't play on any 'live play' sites (Chessbase, Chessdotcom, ICC, etc.) because I do not wish to be stuck at my computer.  I like 'turn based sites' like this - move and then go away; come back in a day, or a few days, and if your opponent has moved, then you go again.  This is much more relaxed, and frankly, I don't need any more stress in my life.

Personally, I find the LG time control easy enough - 24 hours may not SEEM like a lot, but since you get an increment of 36 hours every time you move, it's rather simple to stock up on time if you need it.  I play Chess (only) at another 'turn based' website called Red Hot Pawn - and there, if you want really slow play, you can choose a time control of 3 weeks per move (plus, an extra 28 days if you ever are running low); a game that slow would not be MY choice, but some people there DO use that time control.

Our site creator is Richard Malaschitz, and this time control is HIS selection, but you can communicate with him if you want - perhaps he canaccommodate you in some way.  There is a 'contact' link at the bottom of the page.

What is a “tactic”? Is a 50 moves deep checkmate a “tactic”?

There are some players who have the philosophy that there is no real difference between positional play and tactical play; I'll grant that this MIGHT be correct, when you are looking 50 moves deep, but nobody does that; even machines don't (they display their depth of analysis - it NEVER gets down to 50 moves).  So in practical terms, 'tactics', usually defined as forks, skewers, pins, double attacks, etc. would refer to a SHORT sequence of forced moves leading to a concrete advantage - generally the win of material (but checkmate would qualify); you can read more about this on Wikipedia - look up 'chess tactics'.  Strategic (or positional) play is a technique for selecting moves which should lead to an advantage based on the superior position of the pieces, but generally doesn't yield an immediate advantage; it takes longer for the advantage to be realized, and also requires proper technique.

What you’re probably trying to say is that humans play with better algorithms than machines.

Basically, yes - that is what I'm trying to say.

we’re not that much weaker

Ah, but we ARE - given equal rules and conditions.

Still, if the difference in positions is enormous, then we stand no chance, despite having the better algorithm.

Well, that remains my point here.

Contrast that with your statement:

Yes with enough work you can even win against bots. :)

See?  So which is it?  Do humans stand a chance, given the right conditions, or don't they?  That is my whole purpose here, and why I'd want to conduct the 'experiment.'.

MisterCat, see my game against hex bot

i never played a game of Hex in my life, so understand nothing of what you say; gee - isn't that a COOL looking board!

Sure, if I check all the choices. But who pay me 8 hours a day for a week to do this job? Do you want to donate your money for human players?

Not MY money, but I rather expect that, promoted properly, somebody WOULD pay.  Sponsors were readily found for other human/machine matches.  Also, what is clear is that 'checking all the choices' may be feasible in Hex, but there are way more positions in a Chess game; if we could 'check all the choices', the game would be 'solved' by now, which it is not.

?

You've not specified exactly what are the "right conditions", and in particular you need to specify both time control, how many people are allowed to play (do we also allow ongoing games to be passed from parents to heirs?), allowed hardware/assistants and financial incentives.

So basically I can agree or disagree depending on these unspecified "right conditions". If you pay Carlsen to win vs Stockfish on short time controls then he can probably win but he'll have to think several months per move (from home, with analysis software allowed to store variations) as a full time job. i doubt he is interested in doing that and that you can motivate him financially.

Regarding my chess knowledge, who cares, my arguments stand on their own regarding of my chess rating.

I kind of summarize 'right conditions' in my post above, just below the post by Andrey; but I am not advocating for a game which will go on for months or years; I proposed a few time controls above, but perhaps 30 minutes for the machine to move vs. 90 minutes (or even 60 minutes) for the humans.  This game MIGHT take a week, but there can be adjournments.  The humans can consult online, analyze and tabulate positions on  Chess software, resort to opening books and endgame table bases, but NOT use machines on their side;  I am not looking at a TCEC match-up; this is to be human ingenuity vs. machine.  The team is not 'the world'; a few players (in my Twixtbot experiment, 'the team' consisted of 10 players, with the majority of comments made by the strongest players); the game lasted one month - perhaps fewer moves played then a regular Chess game, but the LG time control is longer than I propose); you can view the 38 move Twixt game here:  GAME financial incentives from the sponsor, as I addressed above; Carlsen can NOT defeat Stockfish under ANY equal time controls, and he would tell you that; my proposed time controls and conditions make your example irrelevant.  You asked for me to define 'tactics', so I wondered about your Chess background, and thus gave a fairly in-depth reply.  Your general comments are much appreciated!

(meow)

MisterCat, Carlsen can defeat Stockfish under condition I've hinted for above, sorry if the grammar isn't clear. Something like 30 seconds per move for the bot and 1 months per move for the human. In addition the human needs all the tablebases and all the opening books for sale on Amazon plus something to store variations that he has already analyzed.

And financial incentives of course, I had all the time to play vs gzero but no incentive, so I've played quickly and I've already lost. I guess Carlsen would suffer same fate. He would have to be motivated very hard to spend all day analyziing the position he has to play. Frankly it would be an insane amount of work for nothing. Even if it wins it would be attributed to time control.

The job you've done with Twixtbot is impressive. Congrats for that. I guess people really analyzed seriouisly for that game. It shows that people can do better than they typically do, especially if there is Maciej in the team of humans. :D

IMO an interesting question is why human hardware is so different from computer hardware. Why we're so slow at discrete tasks like multiplying numbers or enumerating chess positions? And the answer is that these discrete mathematical tasks simply do not happen in the world we live in. So we're not adapted to them at all to these problems. We're adapted at "judgement according to principles", and chess players try to play according to these criteria, but all choices have to be backed with a decent amount of analysis anyway, even if you're a positional genius.

Btw, games like Hex, or Go, they're better for humans precisely because the task of enumerating positions is greatly simplified (each empty cell is a legal move in hex, for go it's almost the same except for captures and other subtle issues like ko and suicide).

@MisterCat I appreciate the effort!

I indeed had the very same feelings about the human vs computer matches in Chess and Go. However... I think in Chess, under your ruleset, human team will still lose.

It's unfortunate that Chess has only 1 board size. Great thing about Go, Hex, Twixt is that one can increase the board size to infinity and with the board size will grow the inability for computer to compute it's knowledge. So I would argue that in those games human is still stronger than computer. There is Chess960, but I don't have anough knowledge to say if this is a valid improvement over Chess and if computers are any weaker in that - actually my guess is that the gap between human and computer would be even bigger.

Chess is just too easy to compute because in most cases material advantage is a valid general advantage. This is great for computers, because they backpropagate to learn, so (talking very colloquial here) they don't have to reach checkmate to draw conclusions. Humans do this as well, but I would argue that humans also learn more forwards instead of backwards, meaning they can create abstract concepts like that a figure is stronger when it is more mobile. It requires 0 games to find this out instead of a billion self-played games. But in Chess it's not enough.

I have another question: does anyone know how many people play here using a computer engine?

kspttw, you should start a NEW forum thread if you are going to ask a new question.  I would suggest being more specific - this is the Chess forum, but engines are available for many of the games here (most?); are you only asking about Chess?  Also, do you refer to using computer engines to CHEAT, or simply using engines to analyze games after the fact?  Do you know that in official 'postal Chess' clubs now, ALL computer engines are permitted, and this is not considered cheating.

So your question should be more specific, and belongs in a different thread.

mc

Yes, I asked about chess... And I wrote "play using a computer engine", which means during the game, not after in analysis...

I think it is a good thread.

I don't suggest it is allowed or not. I just ask.

I created this thread for a specific purpose.  Threads can be created without limit, so if you have a NEW topic, it belongs in a NEW thread.  I will field your question there, and perhaps so shall others.

mc