Having worked with AI, I can, for the most part, confirm what the article says; AI lies--and I wager more than 10%.
AI is garbage at creating anything novel. If you use it to write, you have to hold its hand through the entire process. It's not worth it, unless you're completely swagless and without panache. If you write orcs and medieval sword-wielders, or mafia bosses making love with a gunholster still around their hip, you can certainly be replaced by AI. But since novelty is one of the most important characteristics of great art, humans will always be there.
Much of the information in this article is wrong or misleading. People who decide to write about LLMs should partner with someone who actually understands the technology. Lots of good stuff in this piece, but next time run it by someone who works with these things at a more intimate, technical level.
Speaking as an amateur artist, your average artist really is extremely neurotic and in some way unwell. People will absolutely pay an artist for good work as opposed to using an AI if said artist can prove they will deliver. That becomes a lot riskier ground to tread when 80% of artists spend more time on Twitter whining about how badly their last piece performed than actually working on anything or trying to understand what happened. The best artists are the best people in general, they keep complaining to a minimum and are grateful about whatever reach they do have.
Once AI fixes it's art-dementia, comic artists are in trouble. If AI could consistently reproduce the Hypergamouse style, say, an episode could be illustrated in seconds instead of hours.
I haven't even begun dabbling in AI art yet because the whole thing intimidates and confuses me.
Interesting to see if an AI will be able to capture the nuances of each SSH type for something like the Hypergamouse characters. There's a level of understanding about men that you have that makes your illustrations resonate. That's why the comic works.
I assume AI can learn to replicate patterns of expression and posture per type. Question is, how well. May work as a tool to speed up your process, though.
Having made a few PC games, there is nothing less reliable than an artist. I will exclusively rely on pre-made art packs, because I dont want to waste my time dealing with artists flaking and turning in crap work, when they turn in amazing work to someone else for less money because they prefer things that way today.
Artists are their own worst enemies in the commercial world. They also like to spend too long on details that don't matter when they should be in pipeline mode, it's just not a talent that lends itself to making money or happy customers.
On the flip side, sound guys are rock solid, and get things done fast and cheaply, I've had a lot of good success with audio guys, buy unfortunately there isn't an incentive to pay them very much for their great work, because they can do it fast and so can 100K other people.
I definitely second the stupidity of artists...bless their hearts. Having had minor success with some European freelance artists, I turned to an American digital artist and offered a market rate commission of a single piece and their agent said they weren't interested. Given that their known work time on a piece is 8-10 hours, I found it very strange that they'd flat out reject the commission - one days work. So, I turned to a few different AI engines and with some of this tips for art generation SDL has dropped here, I had a piece that was "good enough" within an hour. For free.
Any artist who's any good will find an audience. There will always be people who want the artisenal. The mediocre ones will be replaced, but that's nothing new. A lot of commies are gonna lose their nepo-jobs, which is good.
I have several talented artists in my family - each of them has gone the "I'm toooo edgy to do what customers/managers/agencies/producers want - that would be selling out." So they make just enough to scratch by, sliding into obscurity because they can't be GREAT. AI is going to make even scratching by impossible.
Great post! It offers me a new perspective on AI. In fact, I believe AI will neither surpass nor replace humans. The pictures clearly show the difference. AI doesn’t seem to capture the same magic that a real artist does.
I understand that if you're looking for reliability and consistency, AI might be better, but as the market becomes more saturated with superficial AI art, people will start searching for something unique again. The true magic is only found in human art, as far as I’ve seen.
I think the digital dementia is due to token size. The Ai can only "think" until the token is full, then it suddenly becomes incoherent.
The question then is whether token size will increase according to Moore's law. If it does, then in another decade or two the AIs will be able to remember your conversations, and in a few decades after that they will remember everyone' conversations.
But I am not certain because I do not understand how AI "thinks." If the thinking is analyzing every piece of information in relation to every other piece of information in the entire conversation, then the increase in token size will be exponential, and it will take far longer for AI to remember long detailed conversation series.
That AI image of the Spring Queen is impressive. I’m guessing you created it with a tool other than Bing. I wonder if AI will reach the point where it can reliably replace comic book artists.
Having worked with AI, I can, for the most part, confirm what the article says; AI lies--and I wager more than 10%.
AI is garbage at creating anything novel. If you use it to write, you have to hold its hand through the entire process. It's not worth it, unless you're completely swagless and without panache. If you write orcs and medieval sword-wielders, or mafia bosses making love with a gunholster still around their hip, you can certainly be replaced by AI. But since novelty is one of the most important characteristics of great art, humans will always be there.
I've tried to hire a broke artist for three easy projects, paid partial in advance twice, and yet to see any result from it. Just useless.
You need a good editor. I'm not available.
Much of the information in this article is wrong or misleading. People who decide to write about LLMs should partner with someone who actually understands the technology. Lots of good stuff in this piece, but next time run it by someone who works with these things at a more intimate, technical level.
Speaking as an amateur artist, your average artist really is extremely neurotic and in some way unwell. People will absolutely pay an artist for good work as opposed to using an AI if said artist can prove they will deliver. That becomes a lot riskier ground to tread when 80% of artists spend more time on Twitter whining about how badly their last piece performed than actually working on anything or trying to understand what happened. The best artists are the best people in general, they keep complaining to a minimum and are grateful about whatever reach they do have.
Once AI fixes it's art-dementia, comic artists are in trouble. If AI could consistently reproduce the Hypergamouse style, say, an episode could be illustrated in seconds instead of hours.
I haven't even begun dabbling in AI art yet because the whole thing intimidates and confuses me.
Interesting to see if an AI will be able to capture the nuances of each SSH type for something like the Hypergamouse characters. There's a level of understanding about men that you have that makes your illustrations resonate. That's why the comic works.
I assume AI can learn to replicate patterns of expression and posture per type. Question is, how well. May work as a tool to speed up your process, though.
Having made a few PC games, there is nothing less reliable than an artist. I will exclusively rely on pre-made art packs, because I dont want to waste my time dealing with artists flaking and turning in crap work, when they turn in amazing work to someone else for less money because they prefer things that way today.
Artists are their own worst enemies in the commercial world. They also like to spend too long on details that don't matter when they should be in pipeline mode, it's just not a talent that lends itself to making money or happy customers.
On the flip side, sound guys are rock solid, and get things done fast and cheaply, I've had a lot of good success with audio guys, buy unfortunately there isn't an incentive to pay them very much for their great work, because they can do it fast and so can 100K other people.
I definitely second the stupidity of artists...bless their hearts. Having had minor success with some European freelance artists, I turned to an American digital artist and offered a market rate commission of a single piece and their agent said they weren't interested. Given that their known work time on a piece is 8-10 hours, I found it very strange that they'd flat out reject the commission - one days work. So, I turned to a few different AI engines and with some of this tips for art generation SDL has dropped here, I had a piece that was "good enough" within an hour. For free.
I learned that day why artists are starving.
Any artist who's any good will find an audience. There will always be people who want the artisenal. The mediocre ones will be replaced, but that's nothing new. A lot of commies are gonna lose their nepo-jobs, which is good.
I have several talented artists in my family - each of them has gone the "I'm toooo edgy to do what customers/managers/agencies/producers want - that would be selling out." So they make just enough to scratch by, sliding into obscurity because they can't be GREAT. AI is going to make even scratching by impossible.
Great post! It offers me a new perspective on AI. In fact, I believe AI will neither surpass nor replace humans. The pictures clearly show the difference. AI doesn’t seem to capture the same magic that a real artist does.
I understand that if you're looking for reliability and consistency, AI might be better, but as the market becomes more saturated with superficial AI art, people will start searching for something unique again. The true magic is only found in human art, as far as I’ve seen.
“Clown world has been attempting to come up with reproductions that replace the real thing”
Clown world is hoping tech can come up with more docile slave reproductions of humans. Hence the race to transhumanism or just robots
I think the digital dementia is due to token size. The Ai can only "think" until the token is full, then it suddenly becomes incoherent.
The question then is whether token size will increase according to Moore's law. If it does, then in another decade or two the AIs will be able to remember your conversations, and in a few decades after that they will remember everyone' conversations.
But I am not certain because I do not understand how AI "thinks." If the thinking is analyzing every piece of information in relation to every other piece of information in the entire conversation, then the increase in token size will be exponential, and it will take far longer for AI to remember long detailed conversation series.
That AI image of the Spring Queen is impressive. I’m guessing you created it with a tool other than Bing. I wonder if AI will reach the point where it can reliably replace comic book artists.
That was OpenArt.ai. It's very impressive and excellent for cover images, but not reliable enough to serve as a comic book illustrator.
So AI is going to replace pajeets?
Moving information from one place to another is a great description of HR. I am going to use that.
"So if you ever wondered why it takes so long to finish a comic book project, now you know. "
"All artists are flakes." - artist Alex Ross.