How Would You Use AI to Cheat?

[Kirby]

In the process of losing 1000 games, your skill level improve from 20k to 1d. Even though you are still losing badly to Leela Zero at the skill level of 1d, your way of thinking will be quite different from your way of thinking at 20k level.

I definitely think your way of thinking will be different. It's just not clear to me that it'd be more enjoyable. Isn't there even some saying: ignorance is bliss? :-p

Personally, I'm impacted significantly by game result. I feel pretty bad if I lose, regardless of "how well I played". The fact is, even if I played pro level for almost all of the moves, if I lost, my opponent outdid me. So my enjoyment of the game is pretty strongly correlated with win/loss. Yeah, my life is happier if I win - that's fun. But if I lose, it sucks.

That's why I'm trying to see what it is about go that is enjoyable if you take away the winning/losing aspect. After knowing the basic rules, any strategy that you add on top is for the purpose if better increasing your chances of winning, is it not?

Note: I do find enjoyment from tsumego problems from time to time. And that's because they have an objective (solving the problem). Solving the problem helps me achieve the goal I was trying to achieve. But if I set out to play a game, my objective is to try to win the game. If that doesn't happen, I don't get much enjoyment. In the same way, I don't feel good about a tsumego problem until I've solved it.

[EdLee]

Personally, I'm impacted significantly by game result.
...
But if I set out to play a game, my objective is to try to win the game. If that doesn't happen, I don't get much enjoyment.

Very basic human nature; not for all people, but for many, yes.
Very understandable. Cliche: play random-bot; guaranteed ~100% win and happiness, no ?

[handofaduck]

Like a lot of people, my first thought was to use the bot to avoid blunders. Hopefully at most one per game. But I have had another thought.

Instead of me picking one of the bot's top three choices, let the bot pick one of my top three choices. So instead of me playing like the bot, let the bot play like me. To put it differently, let me play my best game. Assuming the bot is good at evaluating my plays, OC.

It's already possible in Lizzie to come close to this. I'll hide this because some people may not want the temptation.

Just turn off (set to false) show-next-moves, show-leelaz-variation, show-subboard, and show-best-moves in Lizzie's config.txt (used to be Lizzie.properties.) You'll be left with a UI that only shows the board, the win rate, and the game tree graph. So you can try moves and you won't get Leela Zero's recommendations, but you'll know if they make your position much worse than its best move. If you don't like the result, you can back up and try something else. It's a score estimator on steroids. Of course, this can be abused and if you click on every empty intersection it sort of defeats the purpose. I tried this once against a bot. It scared the hell out of me. I will not do it again. Even this knowledge is too powerful. That's why I stopped using score estimators years ago. I only say this because I know many payers are addicted to score estimators that are on various servers.

[John Fairbairn]

Personally, I'm impacted significantly by game result. I feel pretty bad if I lose, regardless of "how well I played". The fact is, even if I played pro level for almost all of the moves, if I lost, my opponent outdid me. So my enjoyment of the game is pretty strongly correlated with win/loss. Yeah, my life is happier if I win - that's fun. But if I lose, it sucks.

Everybody is different, of course, but I wonder if some of the apparent differences are down to cultural conditioning. We now live in a world where professional sport is a constant and dominating presence. Some of the intensity the pros show on the field and in interviews must rub off on amateurs. When I was young, if a player (pro or amateur) scored a great goal or whatever, his acknowledgement was the merest nod of the head as he walked back to his position. Nowadays, shirts are ripped off, bums are waggled and fingers jabbed down opponents' throats, and I expect the crowing continues on social media. Tiny kids follow these examples.

Conceivably it's worse in America, with bigger prizes for pros and things like sports scholarships for amateurs. The concept of "Coach Carter" bears no resemblance to a PE teacher in British schools.

I feel a little sorry for amateurs brought up in that atmosphere. It might be fun to celebrate at the time but it takes a lot away from deeper enjoyment of the game. I've been shocked at how little baseball stars know of the history of their game.

At any rate, my own enjoyment of go covers a much wider spectrum than the result. Observing the history and personalities is part of it, as that feeds into a wider appreciation of the rest of life. Another part is seeing how other people think - again this is more about people than the game and so is richer. Yet another part is notching up milestones in my own understanding of the game. The game I most remember in my life was one where I first played a particular tesuji that I'd only seen in books before. I obviously "knew" it otherwise I wouldn't have played it, but what I got from that game was the richness of that tesuji, that I had overlooked before, in its impact on the rest of the game. Apart from that, whether I won or lost any particular game is mostly a blur.

Games I lose, especially nowadays, are most often down to simple blunders such as overlooking a snapback. It's not the losing that's annoying but the waste of time caused by missing something that I know very, very well. And if I win, I can usually see that my opponent likewise made the same sort of blunders I make, so I can hardly make a smirk of superiority, can I?

I'm nothing like as interested in AI as many people here, because it lacks the people element, or more specifically (a) it doesn't make the same sorts of blunders as we do, and (b) it contributes nothing to how humans should think. My interest there is more to do with the brain workings of the programmers.

As a specific example of how a loss can give great pleasure, perhaps my favourite is the Japanese pro who blundered and then starting frantically running round the room like a dog chasing its tail. For me, at least, that's much more fun than winning my own games.

[RobertJasiek]

Please explain why you think that AI contributes nothing to how humans should think. AI thinking may be fundamentally different but aren't the results of AI thinking worth considering for human thinking?

[Uberdude]

John/Robert, I also have a (lengthy) response to John's point, so moved it out to a new thread so as not to distract from this "How to AI cheat" one:
viewtopic.php?f=10&t=15805

[dfan]

That's why I'm trying to see what it is about go that is enjoyable if you take away the winning/losing aspect. After knowing the basic rules, any strategy that you add on top is for the purpose if better increasing your chances of winning, is it not?

Here are some things I like about go and chess that don't involve winning:

• Understanding the game better. I see go and chess, like math and physics, as languages that operate in a semantic space other than the usual description of our human experiences. I gain great pleasure from becoming more fluent in these languages.
• Aesthetic appreciation. I can glance at a chess position (go is harder) and it tells me a lot, immediately, in aspects that I have trouble putting into words (see the previous bullet). Just the way that pieces are configured awakens many patterns and recollections and ideas. I love being able to have this rich aesthetic reaction to things like game positions rather than just poems and paintings and nature.
• History. I love that in both games there is a rich history going back centuries, where a large part of the history are the game records themselves, which have been passed down perfectly. I know lots of chess history the way people know go history, and I love that there's a shared knowledge of the important events, from things like "who has been world champion" all the way down to very specific moves from specific games.
• Appreciation of expert play. Just as being better at soccer makes you more able to appreciate professional soccer, I enjoy following along with videos of professional play with commentary (or just strong players streaming), as well as playing over pro games.
• The social aspect. I hang out every week with friends of various skills who all love this game, and it's great to have a common specialized hobby to study and laugh and argue about together. I'm going to the US Congress this year and am looking forward to meeting other people that I "know" from L19 and AYD.

That's not an exhaustive list but it's a start!

[Kirby]

Everybody is different, of course, but I wonder if some of the apparent differences are down to cultural conditioning.

It's possible. I'd imagine that people are influenced by culture wherever they are, from whatever time period.

The game I most remember in my life was one where I first played a particular tesuji that I'd only seen in books before. I obviously "knew" it otherwise I wouldn't have played it, but what I got from that game was the richness of that tesuji, that I had overlooked before, in its impact on the rest of the game.

I have some similar experiences. I remember a game from a long time ago at a tournament. I was 5k at the time, and there was a risky sequence that I thought my work - but it was at the limit of my skill to read it out. I played it out and it worked. I felt a lot of good feeling not just from the result, but because of the sequence I played.

But I have to wonder - had I spent the same amount of time reading out that sequence, played it out, and *still* lost the game... would I have had such a pleasant feeling about the same process? I don't think so. At least my feeling would be diminished. It would be something like, "yeah, that was a cool sequence, but I still lost".

I guess I can get pleasure from miniature sequences or moves I play out in the game, but winning the game gives me more confidence that I'm not fooling myself with their novelty. If I still lose, it's like my skills didn't seem as good as I had thought.

I'm nothing like as interested in AI as many people here, because it lacks the people element, or more specifically (a) it doesn't make the same sorts of blunders as we do, and (b) it contributes nothing to how humans should think. My interest there is more to do with the brain workings of the programmers.

Agree that it lacks the people element. And further, since a lot of bots use some form of deep and/or reinforcement learning, the learning process is a bit of a black box to the programmers. So if I play an AI, I don't get the sense that I'm learning anything about the programmer's brain - I just kind of think, "huh - the algorithm figured this out after a lot of training".

[Kirby]

That's why I'm trying to see what it is about go that is enjoyable if you take away the winning/losing aspect. After knowing the basic rules, any strategy that you add on top is for the purpose if better increasing your chances of winning, is it not?

Here are some things I like about go and chess that don't involve winning:

• Understanding the game better. I see go and chess, like math and physics, as languages that operate in a semantic space other than the usual description of our human experiences. I gain great pleasure from becoming more fluent in these languages.
• Aesthetic appreciation. I can glance at a chess position (go is harder) and it tells me a lot, immediately, in aspects that I have trouble putting into words (see the previous bullet). Just the way that pieces are configured awakens many patterns and recollections and ideas. I love being able to have this rich aesthetic reaction to things like game positions rather than just poems and paintings and nature.
• History. I love that in both games there is a rich history going back centuries, where a large part of the history are the game records themselves, which have been passed down perfectly. I know lots of chess history the way people know go history, and I love that there's a shared knowledge of the important events, from things like "who has been world champion" all the way down to very specific moves from specific games.
• Appreciation of expert play. Just as being better at soccer makes you more able to appreciate professional soccer, I enjoy following along with videos of professional play with commentary (or just strong players streaming), as well as playing over pro games.
• The social aspect. I hang out every week with friends of various skills who all love this game, and it's great to have a common specialized hobby to study and laugh and argue about together. I'm going to the US Congress this year and am looking forward to meeting other people that I "know" from L19 and AYD.

That's not an exhaustive list but it's a start!

I'm definitely with you on the social aspect - that's one of the reasons I frequent L19. Regarding the first point about understanding the game better, to me, understanding the game better is tied to understanding how to win better. Outside of that, what is there?

[Tryss]

For me, it's not to win, but to be intellectually challenged and "solve" the interesting problem in front of me : "how to win this game?". If all I wanted was to win, I would just play against a 20k without handicap, but then the "problem" is trivial.

When I win by a large margin or very easily, I'm not as pleased compared to when I win a very close game (actually, I prefer losing a very close game than winning a game without challenge). A win is only important if it's meaningful

You talked about tsumego earlier, it's kinda the same : solving a 30k tsumego don't give me any feeling. Solving a complicated dan level tsumego? That's really satisfying !

[dfan]

I'm definitely with you on the social aspect - that's one of the reasons I frequent L19. Regarding the first point about understanding the game better, to me, understanding the game better is tied to understanding how to win better. Outside of that, what is there?

As I said, it is similar to my love of math and physics. The fields of math and physics aim to optimize how much we know about the world and the logical structure of mathematics. The fields of go and chess aim to optimize how best to play those games. The study of all these fields gives me a lot of satisfaction of pleasure, which is true independent of the thing being optimized.

I'm not asking you to feel this way, of course! I'm just explaining my own feelings. I think there are a wide variety of ways in which people enjoy games, which is part of what makes them so great. For some people, 100% of their interest is in the competitive aspect; for some, it's close to zero and the rules of the game just provide a rich context for exploration and appreciation. I'm somewhere in the middle.

[Elom]

Time for some sort of 'AlphaScout', a neural network designed to identify dishonest play.

[moha]

Time for some sort of 'AlphaScout', a neural network designed to identify dishonest play.

This probably doesn't help in the long run since a net can also be taught to cheat undetectably, imitating typical human error distributions. Actually you may not even need a specific net for that.

[Kirby]

For me, it's not to win, but to be intellectually challenged and "solve" the interesting problem in front of me : "how to win this game?".

I agree, and this is kind of my point. The enjoyment has a tie to winning the game - your objective, is to solve the problem that you've stated: "how to win this game?".

Accordingly, from that perspective, if I *lose* the game, I have failed at my objective. Comparing to tsumego problems, it's as if I got the tsumego problem incorrect.

Getting the problem incorrect is not fun to me. I feel satisfaction only if I've solved the problem correctly. Similarly, it is hard for me to feel much satisfaction in playing a game when I have not ultimately "solved" the problem of "how to win this game?"

[Kirby]

As I said, it is similar to my love of math and physics. The fields of math and physics aim to optimize how much we know about the world and the logical structure of mathematics. The fields of go and chess aim to optimize how best to play those games. The study of all these fields gives me a lot of satisfaction of pleasure, which is true independent of the thing being optimized.

I guess you are saying that you get satisfaction from the study of go, rather than the result of actually playing a game of go.

I like study, too, but I seem to enjoy the feeling of feedback. If I am doing go problems, I like the positive feedback of seeing, "yes, I solved the problem correctly." In terms of playing a game, the only feedback that I really recognize is the result of the game.

If the feedback isn't positive, either with a go problem or with playing a game, I feel that I've failed in some sense, and it's not a positive feeling.

I'm not asking you to feel this way, of course! I'm just explaining my own feelings.

I'd like to feel the same way, and I appreciate that you're sharing your feelings.