Marcus has used chess several times in his articles about AI on stderr and in comments on Pharyngula and it got me thinking about whether there is something valuable we can learn from how the ascend of AI of sorts has impacted chess. And I think there is. First about the state of affairs as far as AI in the chess world goes.
The good:
The chess-playing AI’s are getting better and more accessible very quickly. What once needed a supercomputer the size of a wardrobe that probably used enough power to heat a household, can now be easily done by a pocket computer running on a battery. This accessibility of high-quality game analysis to anyone with a smartphone has led to a relative chess boom. Today’s young generation has unprecedented access to learning about chess games. Websites like chess.com and Lichess.org are thriving. As a result, new chess masters and grandmasters are getting younger and younger. AI has contributed to humans getting better at the game and has led to more people enjoying said game.
The bad:
Wide and easy access to AI that can easily beat even the best chess player of all time has its dark side too. Cheating both in online and OTB tournament chess is at an unprecedented level. I am not a bad chess player and not an excellent one either. But I am good enough to occasionally be paired with really good players online. And also with cheaters who like to pretend they are good. I do not know the exact number, but I have reported probably over a dozen people for suspiciously good play. One report was rejected at the time but said player was confirmed to be a cheater about a month later. One report was a mistake on my part. All the rest were confirmed to be cheaters, sometimes after a short delay, sometimes nearly immediately. And every year there is a talk about cheating in high-ranking OTB chess tournaments, occasionally even with physical proof – a few years ago a chess grandmaster lost his title after being caught analyzing his current game with a phone hidden in the restrooms.
The ugly:
The rampant cheating online and some prominent cheating scandals OTB foster a culture of paranoia. Former world champion Vladimir Kramnik embodied this paranoia last year, when he publicly hinted that GM Hikaru Nakamura is cheating, without outright saying so. The only proof that Kramnik provided for his allegations proved only that a high understanding of the game of chess does not automatically translate to a high understanding of maths and statistics and how proofs work. But Kramnik is not alone. Allegedly the talk about cheating is behind the scenes all the time at the highest echelons of chess and suspicions are not uncommon. Rarely names are dropped and proofs are provided, but the suspicions are there all the time. I observed this paranoia in myself after losing a game egregiously and my high ratio of correct to false reporting of foul play is because I do my best to analyze the games afterward and look at some data before reporting someone. I also know that I have been myself probably twice reported for foul play (at least my opponents told me they were reporting me). Both of those reports would of course be mistaken. Funnily enough both of those instances I did not play particularly well and subsequent analysis found really sub-par gameplay on my part.
So, what to do with it, is there something to learn about how to deal with AI overtaking the arts? I think there is.
If anything, chess teaches us that the ascend of highly capable AI into a field does not automatically mean the death of said field. Chess tournaments still exist, and amateur chess players still enjoy the game of skill. People do not want to just see and admire good chess games, they want to see and admire good chessgames played by other people. And I think the same applies to art. Using AI as I tried (and failed) to do is equivalent to a chess player using AI to learn a new strategy or analyze their games. If done properly, it could help a lot of people to learn new skills faster and better than before and unleash an unprecedented boom of art. But people still want to see other people’s creations, not just slop churned out by algorithms.
However, chess avoided destruction by implementing and enforcing strict regulations. That is more difficult to achieve in arts than in chess because there is no overarching authority like FIDE and I do not know how to implement this in the real world. But an effort should be made. If someone uses AI to create a picture and then passes it off as their own creation, they should be dealt with the same way as if someone is caught cheating at chess. No galleries should display art by said artist, no auction houses should sell it and their reputation should be forever tarnished and the community should shun them and ridicule them (the last one appears to be happening, at least). They might not be plagiarizing in the sense the word is understood right now, but they definitively are not creating in any sense of the word.
I do not understand why some people cheat in a game of skill even when there is nothing tangible of value to be gained. But people still do it, my understanding, or lack thereof is inconsequential. Apparently, they do get the dopamine hit after a won game, even though they did not, as a matter of fact, win the game – a machine did that on their behalf. And there are people to be found online who consider themselves to be artists because they write elaborate prompts to stable diffusion. But they are no more artist than a teenager who uses Stockfish is a chess grandmaster.
In my opinion, just as it is not morally (and in a sense legally as far as online chess sites and FIDE go) OK to “commission” your game of chess to an AI and then pretend that you are the one who won, it is not OK to commission an art piece and then pretend you were the one who created it.
Addendum: One interesting thing about AI in chess is that whilst the AI does play better than humans, it is generally lousy at mimicking human play. I have won games in a lost position because my opponent resigned – the position was winning for them, but the winning move was so obscure and difficult to find that they could not find it in time. I also lost winning games because I lost my nerves. AI cannot (so far) mimick the time distribution of moves that people have etc. So far even AIs that are deliberately dumbed down to have a lower level corresponding to human players of some strength for the purpose of training or entertainment feel a bit “off” and there are signs that show that they are not human.
Addendum 2: AI is to art what ultra-processed fast food is to nutrition. And if unchecked, it will have the same consequences on our societal mental health as fast food had and continues to have on our physical health.
chess is inherently a competition and art is not. that’s all im going to say for now because i “can’t even with this,” in the parlance of our times.
@Great American Satan, I do not think that difference is very relevant (point 1), it is not even a difference in principle, merely in degree (point 2).
1 -- people cheat in chess even in games where the outcome does not have any real significance whatsoever, as I mentioned in the article. The cheaters know they are bad, they know they won’t get money or fame or reputation, and they cheat anyway.
2 -- the world of art certainly is competitive, especially for people for whom art is not a mere hobby but a livelihood. They compete with each other and with their environment for attention, for customers and patrons, and for reputation. The competition is not as direct as in chess, but it is there. And in the market, an artist who has spent years mastering a technique and then spends days painstakingly drawing a picture is at an inherent disadvantage against one who merely puts prompts in an AI and clicks on results.
It is relevant. Marcus isn’t trying to compete with anybody when he posts AI art on his blog and I’m not either. If either of us was in an art competition, we’d list our materials and techniques. And if an artist who spent years training in art is disadvantages in art, it’s no different from a guy who spent years mastering kung fu being disadvantaged by the invention of guns. It’s going to be a brutal adjustment, but there is zero way to roll this back. Even regulating it won’t save any jobs that it ends. Better for people to get ready for the inevitable on this deal.
Marcus is not claiming that he has drawn the pictures and is not trying to pass them off as such. In all his posts he clearly states that they were created by Midjourney. I did not read your posts containing AI art, to be honest, they passed me by completely.
I think you took the article as an attack on you and Marcus and thus grossly misunderstood what I was trying to say. My bad for not writing more clearly. I have no problem with using AI to create things for personal use, neither do I claim that AI should be abolished or that it is useless. In fact, I wrote in the article that I consider generative AI to have a huge potential.
However, regulating it is necessary not only because it will endanger the livelihoods of artists but also because of its extreme potential for abuse (it is already abused).
That regulations fail to be perfect is not an argument against discussing and/or implementing them.
I suppose I do feel affronted by the anti-AI art side of things, which influences my rhetoric, but I meant Marcus and myself as examples of non-competitive artists. Why do we do what we do? It’s illustration or entertainment, it’s whatever we want to give freely of our own creative lives. I used the word “inherent” about the competitive nature of chess because it is true, and I think the reason most of your comparison doesn’t work regarding art. In an ideal world art isn’t competing against art. There’s a whole other paradigm available to art that will never be possible for chess because chess can never be anything but a contest.
And in that paradigm is the only salvation of art for its own sake -- getting away from the “dream job” view, from art careerism. Art’s been a fucked up wasteland career for ages (specifically because of its siren allure as a dream job) and this tech is the coffin nail. But it’s also a liberator. An artist who uses AI tools in conjunction with traditional techniques to increase their productivity can create more outside of their day job.
Think about webcomic artists who were never going to make a living at what they did, who only did it to share their creative spirit, who flamed out because of how difficult it can be to make art on a schedule without income. Comic artists love drawing people, hate drawing backgrounds, but sometimes come up with stories that really need backgrounds. AI fills in the stage and they can focus on the characters like they always wanted. Shit like that.
I think about the worst results of AI right now, and most of them come back to problems with society that need to be addressed head on, not by regulating the tools used to achieve those bad ends. For example, AI used to tell lies for political propaganda or in furtherance of a conjob, or deceptive advertising. The root issue is that there are no legal penalties for dishonesty, because allowing dishonesty protects moneyed interests. We could change that. The way BS and terrible things are spread by social media is because social media has always been in the hands of the techbroish libertarian private sector, and we’re well past the point that conduct on those platforms needs to be enforced by law. That’s something that could change.
I don’t have a lot of faith anything will change from either side tho, whether it’s regulating tools or regulating conduct, because we’re kinda hurtling toward the collapse of civilization, which will render the whole discussion irrelevant, but that’s my take, for now.
I am not anti AI, I am anti unregulated AI. And AI use in art will be regulated in the future.
Humans are social creatures, every aspect of human life is regulated, either by laws, traditions, customs, or by peer pressure and societal consensus etc. Generative AI is currently running wild because it is new. Whether some regulations are desirable or not, or effective, or needed or too early/too late is a discussion to be had, but they will emerge, eventually.
The things you mentioned I do consider to be OK use of generative AI -- context fill, backgrounds, to generate clues for inspiration, to do away with repetitive, time-consuming tasks like cross-hatching etc. Those are the AI equivalent of rulers, compasses, and stencils -- job-easing and time-saving.
Also, context matters. I cannot read your posts, I glanced at them, they are long and I do not have the time. But I did read some of Marcus’s posts on the topic and I do consider them OK too. They are a blog post written by Marcus about the pictures that Midjourney created on behalf of Marcus. Marcus comments on the capabilities and functions of the AI and sometimes contrasts it with what people do. I consider the writing to be the art and the AI pictures to be what the art is about.
But if Marcus let Midjoujney make the pictures and left the text to be written by ChatGPT, I would consider him a hack and the blogposts a stain on both his and FtB’s reputation.
A creation incorporating AI generative content can be art. Something created wholly by AI without human input except writing some words and clicking “generate” until it randomly spits out an aesthetically pleasing output with correct dentistry and fingers is not art. I do think that current copyright laws are in need of an overhaul, but I do agree with one aspect -- for a person to hold a copyright to something, that person must have a significant input to create it.
_____
I am actually using Grammarly when writing in English, not only for spell check but also because otherwise I forget quite often definite and indefinite articles which my native language does not have, and punctuation rules in English are wildly different from Czech too. I still write every word and sometimes ignore Grammarly suggestions because I want my writing to be mine, warts and all.
I do not think that difference is very relevant
Of course there is!! Chess has specific winning conditions. A “good game” is one that you win, and winning and losing are clearly defined and measured.
Art has subjective conditions and most of us would say something like that it is a matter of taste. For example, I won’t like a Jackson Pollock painting, or an AI version of the same technique. Or a Picasso or AI rendering of a cubist artwork. I just don’t like them. But we cannot say with any certainty that one is not good and the other is good.
I cannot read your posts, I glanced at them, they are long and I do not have the time.
TL;DR is one of the original asshole moves of the internet, and I am unimpressed in the extreme that you’re playing it. It’s disrespectful as all hell. In cases where you don’t read something someone put effort into writing, maybe you should take your mother’s advice and “if you can’t say something nice (or even moderate) stay silent.”
But if Marcus let Midjoujney make the pictures and left the text to be written by ChatGPT, I would consider him a hack and the blogposts a stain on both his and FtB’s reputation.
Speaking of being a hack and a stain. Charly, I urge you to rethink your attitude.
Something created wholly by AI without human input except writing some words and clicking “generate” until it randomly spits out an aesthetically pleasing output with correct dentistry and fingers is not art.
I see. Is throwing paint at a canvas art, enquiring minds want to know?
I am not anti AI, I am anti unregulated AI. And AI use in art will be regulated in the future.
Consider the possibility that regulatory frameworks around AI are intended to maintain capitalist ownership of an important technology, and have nothing to do with making the world a better or safer place.
I’ll let you return to waving your cane angrily at the clouds.
@Marcus, it was you who brought the chess into these comparisons, I thought that you saw it as pertinent and I expressed my thoughts on it. I am talking about parallels in the impact of AI on a field previously dominated by humans. I do know -- and I said so under one of your posts -- that the fields are different. But I also thought I agree with you that there are some usefull parallels in there. And now you say that there are no useful parallels after all? Color me confused.
Good art vs bad art is irrelevant to the issue at hand. Whether something is good or bad art is highly subjective. No two persons, either experts or laymen agree on what is good or bad. And many laymen confuse “I do not like it” with “it is not art” which is nonsense. But in my experience almost all people -- and definitiely all my tutors from university -- do agree that for art to be art it has to combine human skill and experience and the human desire to express and evoke emotions through the application of said personal skills by creating something. It is also, AFAIK roughly the definition of art in the dictionary and many legal systems.
I was not intending on being an asshole, I was intending to do the exaxt opposite. I really could not read all that GAS has written about AI on his blog right and I merely did not want him to continue the discussion under the misapprehension that I am talking about or responding to his posts. I would consider saying nothing at all in that scenario to be counterproductive. Should I let him to continue to write here thinking I am responding to his blogposts? I completely fail to understand why it is an asshole move and what attitude am I supposed to rethink? Being honest?
Throwing paint on canvas is art, if a human does decide on what paint to throw, how to throw it, on what surface to throw it, when to throw it, how much to throw, and when to stop throwing because the work is done etc. There is a human decision to create, how to create, what to create and there is skill and experience involved. There is someone behind each step of the creation with whose decisions it is possible to empathize.
Having a mass-produced machine throwing paint at the canvas with the push of a button by any random passerby without them having any active role whatsoever in deciding the canvas size and type, the positioning, the colors and the amount of paint thrown is not art. An AI equivalent of that would be a “generate Jackson Pollock style picture” prompt as the button, Midjourney as the machine, and the internet users as the passersby. There is a decidion to push the button, I grant you that -- but that is all.
I am considering the possibility of some regulations being bad, I am not saying that all regulations are automatically good and need not be challenged. Indeed I am expressly saying:
So I fail to see why you felt the need to write that as if it were relevant.
So, is in this context using AI in art a skill? Sometimes yes, sometimes no. I think there is a continuum from a lot of skill being involved to no skill being involved at all. Writing some phrases into AI and hitting generate repeatedly is firmly on one on the latter end and I am concentrating on that. But I am expressly saying -- and now repeating -- that there are legitimate uses of AI in the arts.
Honestly, I did not expect you two to get this offended and defensive like I shot your dog or something. I thought I wrote pretty anodyne stuff. Do you feel personally attacked because you do consider yourselves to be the creators of the pieces you let Midjourney make? Because you definitively are not, even legally you are not AFAIK. You are commissioning Midjourney to create pieces on your behalf. You are no more creating them than you would be if you approached an artist with a portfolio and said to them “I want this style to be used to depict this object instead of that one” and then let them run through several sketch feedback loops until they have made something you like.
Specific example to finish it: When you commissioned your tribal badger logo by Andreas, you did not create it. If you give Midjourney a prompt “create tribal badger head logo in the style of Andreas Avester” you would not have created it either.