I wish we could start arguing about the ethics of compensation for training data and requiring a concrete way to both protect opt-out, as well as compensate those who contribute, rather than argue about a product that absolutely does have a user base (as is continually proven). I don’t think there’s a win against the demand, but you can win the ethics battle and force better regulations.
GenAI advocates would rather get rid of IP altogether, though. They claim they’re all running ethical models already and it’s perfect, but they also want artists’ right to opt-out to not exist. Nevermind compensation, or the need for opt-in, we can’t even agree on the importance of consent.
And robocallers/spam callers would rather get rid of bans on automated call systems and enforcement of Do Not Call lists. Doesn’t mean we have to do what they want, and it would be an extremely ineffective argument to argue for a ban on phones or even just a ban on automated call systems connected to phones. Both are tools with extremely legitimate use cases that can and have been exploited for malicious and unethical means. Welcome to the complexity of modern living. I see you as an ally, but I warn that we’ll need to be specific in our language and our desires in order to shape the discussion properly, else you’ll just end up categorized with the “nutjobs wanting to ban phones” (embellished simile I’m using to give you an example in a different context) and you’re going to lose a lot of momentum from the legitimately ethical people who are on your side.
I don’t think there’s really a “demand” per se. It seems to me like the vast vast majority of AI “art” and text is spam. Many of these users seem to be using cheap/free versions of whatever LLM or image generators.
That viewpoint is extremely short sighted. You’re missing the field for the trees. Open source models that people run on their local hardware with open weights absolutely do exist and function well. As an example of demand, I personally have a DnD group that uses it for token generation. It gives a far deeper sense of immersion for our custom campaigns where we would otherwise not be able to afford to commission custom imagery, and yes these are generated locally on an m1 mac mini. People viewing it as a replacement for custom commissioned art are, at least with its current and foreseeable capabilities, incorrect in their assumption. It’s merely an augmentation and tool that fills niche low-cost low-“risk” voids. I assure you, for example, that there is absolutely some kid out there who has generated an image of either their imaginary friend or custom super hero. This has likely brought them great joy, especially if they’re unable to otherwise embody their idea due to lack of skill or funding. You have to look at the tool from all angles. A car, in isolation, is a multi-ton inheritia machine capable of unspeakable atrocities, yet we cohabitate with them every single day because we understand life is complex, there are benefits to doing so, and a single view of a tool does not reflect it’s reality.
I don’t think “some kid” experiencing “great joy” at AI slop is worth the spam from scammers and environmental impacts.
Also, the majority of people using AI are not running it locally. If people were running it locally, on low power to preserve the environment, with content that artists consented to have trained, to use exclusively for non-commercial use, I wouldn’t have a problem with it.
But that isn’t reality.
Cars are dangerous, yet we live in a society that…
Cars are dangerous, that’s why they can only be used by licensed operators (both for safety and environmental impact), are heavily regulated, and still have problems coming from overreliance that would have been easier to solve before we built infrastructure around them.
It’s why we’ve seen things like walkable cities and public transit come back as popular ideas.
I think you’re aiming for perfect over better, and honestly I don’t think you’re going to make much headway in that effort, especially when you’re disregarding the joy of children but you do you.
So a couple of things. One, he’s right and I agree with him on his first point. There is no such thing as a “ai artist” or a prompt director or whatever you’d like to call it. The machine is not complex enough in use to need a specialized person like that, and I wouldn’t say they were an artist even if it were. Second, I literally follow artists who use ai just for finishing details on their work, sometimes it’s as simple as fur renders that they don’t want to add by hand so they involve an ai renderer to apply the finishing layer, and these are artists I’ve been following since before ai “art” (image generation) existed. So he’s just straight up wrong about there not being a single real artist using ai. It’s a tool, like any other. You can have your negative opinion on it, but it’s honestly useless to be so hostile to something just because it scares you and you don’t understand it, so I’m not going to watch the video past that.
If current models never changed again - none of what’s happening would “die.” We already have programs that can turn any image you provide into any image you describe, even if you provide solid noise.
What people do with that tool can be trivial… or it can take immense effort and thought. I don’t understand how an iterative process lasting days could be anything but art. Objecting to where the tools came from can’t change that.
I wish we could start arguing about the ethics of compensation for training data and requiring a concrete way to both protect opt-out, as well as compensate those who contribute, rather than argue about a product that absolutely does have a user base (as is continually proven). I don’t think there’s a win against the demand, but you can win the ethics battle and force better regulations.
GenAI advocates would rather get rid of IP altogether, though. They claim they’re all running ethical models already and it’s perfect, but they also want artists’ right to opt-out to not exist. Nevermind compensation, or the need for opt-in, we can’t even agree on the importance of consent.
And robocallers/spam callers would rather get rid of bans on automated call systems and enforcement of Do Not Call lists. Doesn’t mean we have to do what they want, and it would be an extremely ineffective argument to argue for a ban on phones or even just a ban on automated call systems connected to phones. Both are tools with extremely legitimate use cases that can and have been exploited for malicious and unethical means. Welcome to the complexity of modern living. I see you as an ally, but I warn that we’ll need to be specific in our language and our desires in order to shape the discussion properly, else you’ll just end up categorized with the “nutjobs wanting to ban phones” (embellished simile I’m using to give you an example in a different context) and you’re going to lose a lot of momentum from the legitimately ethical people who are on your side.
I don’t think there’s really a “demand” per se. It seems to me like the vast vast majority of AI “art” and text is spam. Many of these users seem to be using cheap/free versions of whatever LLM or image generators.
OpenAI is by far the most popular, but also said that even on the most expensive $200/month plan, they are losing money.
Is this “demand” going to exist if and when they inevitably raise the price?
If and when Facebook makes changes to how they monetize posts, will the shrimp Jesus spammers move on to the next scheme?
Will the businesses using AI for customer service and data entry keep using it if it costs more than using human employees?
This whole “industry” is teetering on a knifes edge.
That viewpoint is extremely short sighted. You’re missing the field for the trees. Open source models that people run on their local hardware with open weights absolutely do exist and function well. As an example of demand, I personally have a DnD group that uses it for token generation. It gives a far deeper sense of immersion for our custom campaigns where we would otherwise not be able to afford to commission custom imagery, and yes these are generated locally on an m1 mac mini. People viewing it as a replacement for custom commissioned art are, at least with its current and foreseeable capabilities, incorrect in their assumption. It’s merely an augmentation and tool that fills niche low-cost low-“risk” voids. I assure you, for example, that there is absolutely some kid out there who has generated an image of either their imaginary friend or custom super hero. This has likely brought them great joy, especially if they’re unable to otherwise embody their idea due to lack of skill or funding. You have to look at the tool from all angles. A car, in isolation, is a multi-ton inheritia machine capable of unspeakable atrocities, yet we cohabitate with them every single day because we understand life is complex, there are benefits to doing so, and a single view of a tool does not reflect it’s reality.
I don’t think “some kid” experiencing “great joy” at AI slop is worth the spam from scammers and environmental impacts.
Also, the majority of people using AI are not running it locally. If people were running it locally, on low power to preserve the environment, with content that artists consented to have trained, to use exclusively for non-commercial use, I wouldn’t have a problem with it.
But that isn’t reality.
Cars are dangerous, that’s why they can only be used by licensed operators (both for safety and environmental impact), are heavily regulated, and still have problems coming from overreliance that would have been easier to solve before we built infrastructure around them.
It’s why we’ve seen things like walkable cities and public transit come back as popular ideas.
I think you’re aiming for perfect over better, and honestly I don’t think you’re going to make much headway in that effort, especially when you’re disregarding the joy of children but you do you.
Local models already won.
I agree with the points about ethics and compensation, however.
So a couple of things. One, he’s right and I agree with him on his first point. There is no such thing as a “ai artist” or a prompt director or whatever you’d like to call it. The machine is not complex enough in use to need a specialized person like that, and I wouldn’t say they were an artist even if it were. Second, I literally follow artists who use ai just for finishing details on their work, sometimes it’s as simple as fur renders that they don’t want to add by hand so they involve an ai renderer to apply the finishing layer, and these are artists I’ve been following since before ai “art” (image generation) existed. So he’s just straight up wrong about there not being a single real artist using ai. It’s a tool, like any other. You can have your negative opinion on it, but it’s honestly useless to be so hostile to something just because it scares you and you don’t understand it, so I’m not going to watch the video past that.
If current models never changed again - none of what’s happening would “die.” We already have programs that can turn any image you provide into any image you describe, even if you provide solid noise.
What people do with that tool can be trivial… or it can take immense effort and thought. I don’t understand how an iterative process lasting days could be anything but art. Objecting to where the tools came from can’t change that.