He/him

  • 0 Posts
  • 8 Comments
Joined 1 year ago
cake
Cake day: June 14th, 2023

help-circle

  • AI very provably does use other peoples’ art more than any other artist. It needs huge amounts of media that’s used as a basis for training material — far, far more than your average artist will consume. You can teach a person how to draw, sculpt, paint, model, etc. without ever showing them another artist’s work. You really can’t do that with ML tools we have currently. It’s not completely impossible, but you would be relying on getting a lot of training data in another way and it would probably require a lot of input from humans on the output end to make a model that can come up with something reasonably comprehensible. A

    We don’t have much in terms of laws about this kind of usage because it’s not like in the past a company like DC comics has decided that they want to make Jim Lee’s style to become the “official” style of DC comics, but they don’t want to pay Jim Lee, so they hire a Chinese art factory to mimic his style and cut him out. Something like that wouldn’t be illegal in the sense of current laws, but probably would have been substantially more expensive than simply hiring Lee himself. However, it definitely would have been unethical. It also would likely have caused a legal challenge that might have affected how our laws deal with replication of a “style”. Even in cases where a company establishes their own style guide based on an art style of a specific artist as is common in animation (where it’s understood that the usage of that style is part of the concept art), there is typically an evolution in how that style as it standardizes- See “Steamboat Mickey” versus current versions of Mickey Mouse, or the changes from the first season to the current season of the Simpsons for example.

    This isn’t about using AI tools for your average DM to make art resources for their home campaign. That’s a perfectly reasonable use-case. It isn’t as though your average DM is likely to be commissioning custom art every time there’s a new character in the campaign - they’ll do what we’ve always done: Find reference material that’s “close enough” from copyrighted works and say “something like this.” But if a company is going to start digging into AI, then we as the audience have the right to say, “No, I’m not going to support that and won’t buy a product produced in that way. I assign value to art made the ‘traditional’ way” The obsolescence of industries due to technology is not an inevitability - by all rights it’s entirely possible that an automated process to make perfect, nutritionally balanced food bars that are both cheaper and healthier than a McDonald’s burger could have been produced by now - but no one wants that. Very few people have a diet that consists entirely of Soylent. Just as there’s more to food than nutrition and value, there’s more to art than pictures. The so-called “free hand of the market” goes both ways.

    I’m a digital artist. I’m in an interesting position in this debate, because I see the value and the power of tools like MidJourney and Stable Diffusion and the like. The prospect of training an AI tool on my own work and giving it to the public to be able to make their own art using my style is exactly the kind of artsy-fartsy “concept” thing I dig. I use things like “content-aware fill” tools and special brushes in my work that are basically cousins to these systems and they help me immensely. But also I think that artists should have the right to choose whether their work is used in this way and that if a company is profiting from the usage of an AI model that’s been trained from mass scraping of the internet there should be some legal consideration for that.



  • AI discourse has way too much “Throwing the baby out with the bathwater,” especially from a lot of people who have no idea what they’re talking about. AI, as a thing, is not a perfect system. It’s not a magic panacea that will cure all. There are legitimate concerns about how much it infringes on creative spaces and how it may put people out of work. There are also legitimate concerns about the AI training data scraping web-hosted content indiscriminately without permission. However, these are not the same as AI just being “bad”.

    Do I think a D&D campaign led by a ChatGPT-like DM would be “good”? Probably not as it stands. I’ve played a lot with ChatGPT and its limitations are pretty obvious. Could it get better in the future? Probably. Is it an interesting possible way to get to play D&D if you can’t get a group together? I mean, it’s gotta be better than nothing, right? But the real interesting prospect to me is machine-learning powered tools for the DM. A System that’s trained on WotC-owned resources that lets you just choose a paintbrush that’s labeled “cave” and draw out a series of tunnels and have it automatically populate with crystals and mushrooms and visual points of interest, which lets you sketch out a good-looking map in minutes. Then, as your party is in the cave, the system knows what type of “biome” you used so it has a button to let you generate a random encounter, which it takes from your character levels and where your players are. There’s a lot of ways that “smart” tools could take a lot of work off the DM’s shoulders that would be great. I don’t know if they’re in the pipeline, but the point is that AI isn’t a boogeyman that’s just out to steal jobs and IP.




  • I own it. It’s fun to play with but ultimately suffers from the same problem that almost every tool in this style does. The resources you get to use are limited to the ones that they’ve thought to include. If you want to make a jail, that’s fine, you can make it work. A tavern? Easy. An ancient Greek temple? Eh, you can get there with a bit of imagination. A bathroom? Sorry, bud, you’re on your own.

    I’ve traditionally used Dungeon Painter Studio for my maps, and while it has similar limitations, it has the benefit of being able to import other art, and you get a whole dimension to hide your crimes in. That vaguely bookshelf- looking rectangular thing on the map? It’s an armory cabinet in the barracks. Now in the bathroom it’s a medicine cabinet. In the bedroom? A wardrobe. You can’t see what’s in it, can’t see how tall it is or how high it’s mounted on the wall, so you get to fill in the details with my description. 3d limits your ability to do that because everything looks like what it is. So if you don’t have a model of what you’re looking for, it’s more obvious when you’re making do.