I’m not a well-trained martial artist at all. But I have beaten well-trained martial artists in multiple fights. Apparently, that is not an unheard of phenomenon, either. It seemed to be key that I fight well by some metrics, but as a novice, commit errors that are incomprehensible, uneven and importantly: unpredictable to an expert because they would never do something so silly. I fail to go for obvious openings, and hence end up in unexpected places; but at that point, while I am underestimated because I have been foolish, I suddenly twist out of a grasp with unexpected flexibility, then miss being grabbed again because I have moved randomly and pointlessly, fail to protect against obvious threats, but don’t drop due to an unexpectedly high pain tolerance despite having taken a severe hit, and then take a well-aimed hit with unexpected strength.
This has me wonder whether an AI would have significant difficulties winning against humans who act inconsistently and suboptimally in some ways, without acting like utter idiots randomly all the time—because they don’t take offers the AI was certain they would take, fail to defend against threats the AI was certain they would spot and that were actually traps, stubbornly stick with a strategy even after it has proven defective but hence cannot be budged from it even when the AI really needs them to, etc.
Yet I also wonder whether the chess example is misleading because it is so inherently limited, so very inside the box. To go back to the above fight example: I’ve armwrestled with much stronger people I have beaten in actual fights. If they are much stronger, I inevitably lose the armwrestling. I am just not strong enough, and while I can set my arm with determination until the muscle rips… well, eventually the muscle just rips, and that is that. If I were to use my whole body for leverage like I would in a fight, or chuck something in their eyes to distract them, I would maybe get their hand on the table—but it would be cheating, and I can’t cheat in an armwrestle match, you can’t win that way, it is not allowed. Similarly, acting unpredictably during chess is of very limited advantage compared to the significant disadvantage from suboptimal moves, in light of the limited range of unpredictable move. A beginner being so foolish as to act random in chess is trivially beaten; I am notably terrible at chess. If there are only a few positions your knight can be in before jumping back to prior configurations, then you randomly choosing an inferior one will doom you.
There is ultimately only so much that an AI during a chess game can do. It can’t shortcut via third dimension via an extra chess board above. It can’t put a unicorn figure on the board and claim it combined the queen and a knight. It can’t convince you that actually, you should let it win, in light of the fact that it has now placed a bomb on the chessboard that will otherwise blow up in your face. It can’t screw over your concentration by raping your kid in front of you. It can’t get up for a sip of water, then surprise club you from behind, and then claim you forfeited the match. There are only so many figures, so much space, so many moves, so much room for creativity.
But the real world is not a game. It will likely contain winning opportunities we might not be aware of at all. A very smart commander may still be beaten in a battle by a less smart commander who has a lot more of the same type of weapons, troops and other assets. But if the smarts of the smart commander don’t just extend to battlefield tactics, but also, say, to developing the first nukes/nanotech/an engineered virus, or rather whatever equivalent transformation there may be that we cannot even guess at, at some point, you are done for.
It seemed to be key that I fight well by some metrics
That couldn’t be the case—that would leave you, even after having a black belt, vulnerable towards people who can’t fight, which would defeat the purpose of martial arts. Whichever technique you use, you use when responding to what the other person is currently doing. You don’t simply execute a technique that depends on the person fighting well by some metrics, and then get defeated when it turns out that they are, in fact, only in the 0.001st percentile of fighting well by any metrics we can imagine.
(That said, I’m really happy for your victories—maybe they weren’t quite as well-trained.)
This has me wonder whether an AI would have significant difficulties winning against humans who act inconsistently and suboptimally in some ways, without acting like utter idiots randomly all the time
I’m thinking the AI would predict the way in which the other person would act inconsistently and suboptimally.
If there were multiple paths to victory for the human and the AI could block only one (thereby seemingly giving the human the option to out-random the AI by picking one of the unguarded paths to victory), the AI would be better at predicting the human than the human would be at randomizing.
People are terrible at being unpredictable. I remember a 10+ years-old predictor of a rock-paper-scissors for predicting a “random” decision of a human in a series of games. The humans had no chance.
The “purpose” of most martial arts is to defeat other martial artists of roughly the same skill level, within the rules of the given martial art.
Optimizing for that is not the same as optimizing for general fighting. If you spent your time on the latter, you’d be less good at the former.
“Beginner’s luck” is a thing in almost all games. It’s usually what happens when someone tries a strategy so weird that the better player doesn’t immediately understand what’s going on.
The other day a low-rated chess player did something so weird in his opening that I didn’t see the threat, and he managed to take one of my rooks.
That particular trap won’t work on me again, and might not have worked the first time if I’d been playing someone I was more wary of.
I did eventually manage to recover and win, but it was very close, very fun, and I shook his hand wholeheartedly afterwards.
Every other game we’ve played I’ve just crushed him without effort.
About a year ago I lost in five moves to someone who tried the “Patzer Attack”. Which wouldn’t work on most beginners. The first time I’d ever seen it. It worked once. It will never work on me again.
The “purpose” of most martial arts is to defeat other martial artists of roughly the same skill level, within the rules of the given martial art.
Not only skill level, but usually physical capability level (as proxied by weight and sex) as well. As an aside, although I’m not at all knowledgeable about martial arts or MMA, it always seemed like an interesting thing to do might to use some sort of an ELO system for fighting as well: a really good lightweight might end up fighting a mediocre heavyweight, and the overall winner for a year might be the person in a given <skill, weight, sex> class that had the highest ELO. The only real reason to limit the ELO gap between contestants would be if there were a higher risk of injury, or the resulting fight were consistently just boring. But if GGP is right that a big upset isn’t unheard of, it might be worth 9 boring fights for 1 exciting upset.
The “purpose” of most martial arts is to defeat other martial artists of roughly the same skill level, within the rules of the given martial art.
This is false—the reason they were created was self-defense. That you can have people of similar weight and belt color spar/fight each other in contests is only a side effect of that.
“Beginner’s luck” is a thing in almost all games. It’s usually what happens when someone tries a strategy so weird that the better player doesn’t immediately understand what’s going on.
That doesn’t work in chess if the difference in skill is large enough—if it did, anyone could simply make up n strategies weird enough, and without any skill, win any title or even the World Chess Championship (where n is the number of victories needed).
If you’re saying it works as a matter of random fluctuations—i.e. a player without skill could win, let’s say, 0.5% games against Magnus Carlsen, because these strategies (supposedly) usually almost never work but sometimes they do, that wouldn’t be useful against an AI, because it would still almost certainly win (or, more realistically, I think, simply model us well enough to know when we’d try the weird strategy).
“Even after having a black belt”? One of the people I beat is a twice national champion, instructor with a very reputable agency and san dan in karate. They are seriously impressive good at it. If we agreed to do something predictable, I would be crushed. They are faster, stronger, have better form and balance, know more moves, have better reflexes. I’m in awe of them. They are good. I do think what they do deserves to be called an art, and that they are much, much, much (!) better than I am.
But their actions also presuppose that I will act sensibly (e.g. avoiding injury, using opportunities), and within the rule set in which they were trained.
I really don’t think I could replicate this feat in the exact same way. Having once lost in such a bizarre way, they have learned and adapted. Many beginners only have few moves available, and suck at suppressing their intentions, so they may beat you once, but you’ll destroy them if they try the same trick again. It might work again if they try something new, but again, if you paired the experienced fighter with that specific beginner for a while, pretty quickly, they would constantly win, as they have learned about the unexpected factor.
But in a first fight? I wouldn’t bet on a beginner in such a fight. But nor would I be that surprised by a win.
And I definitely would not believe that having a black belt makes you invulnerable towards streetfighters, or even simply angry incompetent strangers, without one. Nor do I know any martial art trainer who would make such a claim. Safer, for sure. Your punches and kicks more effective, your balance and falls better, better confidence and situational awareness, more strength, faster reflexes, ingrained good responses rather than rookie mistakes, a knowledge of weak body parts, pain trigger points and ways to twist the other person to induce severe pain, knowledge of redirecting strength, of mobilising multiple body parts of yours against one of theirs, all the great stuff. But perfectly safe, no.
This has me wonder about a related point.
I’m not a well-trained martial artist at all. But I have beaten well-trained martial artists in multiple fights. Apparently, that is not an unheard of phenomenon, either. It seemed to be key that I fight well by some metrics, but as a novice, commit errors that are incomprehensible, uneven and importantly: unpredictable to an expert because they would never do something so silly. I fail to go for obvious openings, and hence end up in unexpected places; but at that point, while I am underestimated because I have been foolish, I suddenly twist out of a grasp with unexpected flexibility, then miss being grabbed again because I have moved randomly and pointlessly, fail to protect against obvious threats, but don’t drop due to an unexpectedly high pain tolerance despite having taken a severe hit, and then take a well-aimed hit with unexpected strength.
This has me wonder whether an AI would have significant difficulties winning against humans who act inconsistently and suboptimally in some ways, without acting like utter idiots randomly all the time—because they don’t take offers the AI was certain they would take, fail to defend against threats the AI was certain they would spot and that were actually traps, stubbornly stick with a strategy even after it has proven defective but hence cannot be budged from it even when the AI really needs them to, etc.
Yet I also wonder whether the chess example is misleading because it is so inherently limited, so very inside the box. To go back to the above fight example: I’ve armwrestled with much stronger people I have beaten in actual fights. If they are much stronger, I inevitably lose the armwrestling. I am just not strong enough, and while I can set my arm with determination until the muscle rips… well, eventually the muscle just rips, and that is that. If I were to use my whole body for leverage like I would in a fight, or chuck something in their eyes to distract them, I would maybe get their hand on the table—but it would be cheating, and I can’t cheat in an armwrestle match, you can’t win that way, it is not allowed. Similarly, acting unpredictably during chess is of very limited advantage compared to the significant disadvantage from suboptimal moves, in light of the limited range of unpredictable move. A beginner being so foolish as to act random in chess is trivially beaten; I am notably terrible at chess. If there are only a few positions your knight can be in before jumping back to prior configurations, then you randomly choosing an inferior one will doom you.
There is ultimately only so much that an AI during a chess game can do. It can’t shortcut via third dimension via an extra chess board above. It can’t put a unicorn figure on the board and claim it combined the queen and a knight. It can’t convince you that actually, you should let it win, in light of the fact that it has now placed a bomb on the chessboard that will otherwise blow up in your face. It can’t screw over your concentration by raping your kid in front of you. It can’t get up for a sip of water, then surprise club you from behind, and then claim you forfeited the match. There are only so many figures, so much space, so many moves, so much room for creativity.
But the real world is not a game. It will likely contain winning opportunities we might not be aware of at all. A very smart commander may still be beaten in a battle by a less smart commander who has a lot more of the same type of weapons, troops and other assets. But if the smarts of the smart commander don’t just extend to battlefield tactics, but also, say, to developing the first nukes/nanotech/an engineered virus, or rather whatever equivalent transformation there may be that we cannot even guess at, at some point, you are done for.
I suspect that the domain of martial arts is unusually susceptible to that problem because
Fights happen so quickly (relative to human thought) that lots of decisions need to be made on reflex
(And this is highly relevant to performance because the correct action is heavily dependent on your opponent’s very recent actions)
Most well-trained martial artists were trained on data that is heavily skewed towards formally-trained opponents
That couldn’t be the case—that would leave you, even after having a black belt, vulnerable towards people who can’t fight, which would defeat the purpose of martial arts. Whichever technique you use, you use when responding to what the other person is currently doing. You don’t simply execute a technique that depends on the person fighting well by some metrics, and then get defeated when it turns out that they are, in fact, only in the 0.001st percentile of fighting well by any metrics we can imagine.
(That said, I’m really happy for your victories—maybe they weren’t quite as well-trained.)
I’m thinking the AI would predict the way in which the other person would act inconsistently and suboptimally.
If there were multiple paths to victory for the human and the AI could block only one (thereby seemingly giving the human the option to out-random the AI by picking one of the unguarded paths to victory), the AI would be better at predicting the human than the human would be at randomizing.
People are terrible at being unpredictable. I remember a 10+ years-old predictor of a rock-paper-scissors for predicting a “random” decision of a human in a series of games. The humans had no chance.
The “purpose” of most martial arts is to defeat other martial artists of roughly the same skill level, within the rules of the given martial art.
Optimizing for that is not the same as optimizing for general fighting. If you spent your time on the latter, you’d be less good at the former.
“Beginner’s luck” is a thing in almost all games. It’s usually what happens when someone tries a strategy so weird that the better player doesn’t immediately understand what’s going on.
The other day a low-rated chess player did something so weird in his opening that I didn’t see the threat, and he managed to take one of my rooks.
That particular trap won’t work on me again, and might not have worked the first time if I’d been playing someone I was more wary of.
I did eventually manage to recover and win, but it was very close, very fun, and I shook his hand wholeheartedly afterwards.
Every other game we’ve played I’ve just crushed him without effort.
About a year ago I lost in five moves to someone who tried the “Patzer Attack”. Which wouldn’t work on most beginners. The first time I’d ever seen it. It worked once. It will never work on me again.
Not only skill level, but usually physical capability level (as proxied by weight and sex) as well. As an aside, although I’m not at all knowledgeable about martial arts or MMA, it always seemed like an interesting thing to do might to use some sort of an ELO system for fighting as well: a really good lightweight might end up fighting a mediocre heavyweight, and the overall winner for a year might be the person in a given <skill, weight, sex> class that had the highest ELO. The only real reason to limit the ELO gap between contestants would be if there were a higher risk of injury, or the resulting fight were consistently just boring. But if GGP is right that a big upset isn’t unheard of, it might be worth 9 boring fights for 1 exciting upset.
This is false—the reason they were created was self-defense. That you can have people of similar weight and belt color spar/fight each other in contests is only a side effect of that.
That doesn’t work in chess if the difference in skill is large enough—if it did, anyone could simply make up n strategies weird enough, and without any skill, win any title or even the World Chess Championship (where n is the number of victories needed).
If you’re saying it works as a matter of random fluctuations—i.e. a player without skill could win, let’s say, 0.5% games against Magnus Carlsen, because these strategies (supposedly) usually almost never work but sometimes they do, that wouldn’t be useful against an AI, because it would still almost certainly win (or, more realistically, I think, simply model us well enough to know when we’d try the weird strategy).
“Even after having a black belt”? One of the people I beat is a twice national champion, instructor with a very reputable agency and san dan in karate. They are seriously impressive good at it. If we agreed to do something predictable, I would be crushed. They are faster, stronger, have better form and balance, know more moves, have better reflexes. I’m in awe of them. They are good. I do think what they do deserves to be called an art, and that they are much, much, much (!) better than I am.
But their actions also presuppose that I will act sensibly (e.g. avoiding injury, using opportunities), and within the rule set in which they were trained.
I really don’t think I could replicate this feat in the exact same way. Having once lost in such a bizarre way, they have learned and adapted. Many beginners only have few moves available, and suck at suppressing their intentions, so they may beat you once, but you’ll destroy them if they try the same trick again. It might work again if they try something new, but again, if you paired the experienced fighter with that specific beginner for a while, pretty quickly, they would constantly win, as they have learned about the unexpected factor.
But in a first fight? I wouldn’t bet on a beginner in such a fight. But nor would I be that surprised by a win.
And I definitely would not believe that having a black belt makes you invulnerable towards streetfighters, or even simply angry incompetent strangers, without one. Nor do I know any martial art trainer who would make such a claim. Safer, for sure. Your punches and kicks more effective, your balance and falls better, better confidence and situational awareness, more strength, faster reflexes, ingrained good responses rather than rookie mistakes, a knowledge of weak body parts, pain trigger points and ways to twist the other person to induce severe pain, knowledge of redirecting strength, of mobilising multiple body parts of yours against one of theirs, all the great stuff. But perfectly safe, no.