So the reason I’m getting more and more worried about the gen-AI debate-- and how volatile many of the opinions are getting, even among people who don’t understand or agree on what the terms mean–

– is pretty much the same reason I’m worried about some of the views getting expressed on what types of fictional sex scenes AO3 should and shouldn’t allow.

Before I get attacked about this: Yes, I do believe there is a significant difference between the two issues! The lines between different things that can be called “AI” are somewhat easier to draw than the lines between different fictional themes. And there is a lot more evidence of real-life harm done by some of the “AI” technologies than real-life harm caused by any type of fiction.

But the debates still worry me in similar ways.

In both cases, the reason for the concern isn’t just that people are concerned about the harm that something could do. It’s that they express this by wanting the thing eliminated, through laws and policies– and that they haven’t thought about whether it’s possible to make and enforce those laws and policies without banning the things they actually want to keep.

In both cases, there are even some people who haven’t thought at all about how definitions are written (and how the definitions of what they want to ban actually already include the things they like.)

In both cases, if you point out how difficult it is to define, some people will accuse you of trying to use hypothetical border-cases to distract from the extreme, egregious examples that are their real point.

But in both cases, people are using those extreme examples to push hard for laws and policies that could not be enforced, or even written, without being able to draw the line on those border-cases.

For example:

There’s already a law that “AI-generated art can’t be copyrighted.” So far, the law only encompasses art that’s made entirely by entering prompts into a gen-AI program. It still allows copyright on any material that was made with the assistance of such a program.

When I first heard about this law, it worried me a lot, because I was immediately thinking about where the line would be drawn– if not in this version of the law, then in other, more restrictive versions that some people are currently demanding.

And, I mean– my own art and writing are about as far on the safe side of the line as you can get! I don’t even use spellcheck in my writing. (Or at least, I turn it off as soon as I can figure out how, with every new software update.)

And I completely gave up digital drawing in favor of colored pencils on paper, a long time ago.

But the “AI” companies are making it harder and harder to avoid, even so. Smartphone cameras have already been so thoroughly infiltrated by generative AI that they will detect you taking a photo of the moon (sometimes incorrectly! it’s happened with a photo of the sun at least once!) and fill in the circle with what their datasets say the moon looks like.

This type of feature gets harder to find and turn off, every time the phone updates. Eventually, of course, they’ll start charging lots of money to use it. (And I hope that’ll make it easier to avoid– but realistically, it’ll probably just mean it’s going to make it harder to find and turn off the setting that’s going to double the cost of your phone bill.)

So, anyway:

if I make a piece of art in my usual style, entirely with pencils and paper– and then use my phone to take and share a photo of that art– it’s already very difficult to guarantee that the picture I share isn’t “made with the assistance of gen-AI.”

And any way you could write the definition will bring up similar problems.

Now– in that first link, I mentioned that digital drawing programs have been using the term “AI” for some of their features for decades.

And that the features they call “AI” include things that range from smoothing your brushstrokes in real-time, to adding a filter that adjusts the appearance of the whole image (color, sharpness, brightness, line thickness, or combinations of these details to give a different overall impression, such as making the image look more like a drawing or painting or mosaic).

You couldn’t ban all of this, not without without banning a large percentage of what modern digital art is.

Could you write a law to regulate which of these features would and wouldn’t be allowed, or which of them would make your work uncopyrightable?

Well… you’d have to know a lot more about the computer science of how they work than most people do.

(Including– maybe especially– the lawmakers who’d be deciding this.)

For example: If your definition involves “entering prompts in typed or spoken language,” then this fails to cover the problem in both directions.

1. A gen-AI company could evade it by putting some of the functions in buttons that don’t fit the law’s definition of “prompts,” and become legally indistinguishable from any other buttons in a digital art interface.
2. The language banning “typed or spoken prompts,” if worded without enough understanding of the technology, might also ban digital artists who do any of their work using voice commands or keyboard shortcuts, even if those are nothing more than direct translation of a function like “change entire image to grayscale and increase contrast 30%.”

Point is, if you’re pushing for a law to be made– any law, about anything– you have to consider the people who will be interpreting and enforcing that law.

And if you’re not okay with those people including some of the most ignorant bureaucrats and corrupt cops that you’ve ever heard of?

–then you’re gonna have to push for a redesign of the whole legal system, before you push for any more laws.

And even if you do someday reach that point, you still have to be able to express the law you want in the form of a definition that the enforcers can understand, not to mention “prove beyond reasonable doubt” whether something fits within it.

And a lot of people in these debates are not there yet.

Anyway.

So far, the aforementioned copyright law only encompasses art generated entirely by AI prompts (however you define “AI” and “prompts.”) So the human creator just has to be able to convince you (or whatever authority is challenging the copyright) that they contributed something to the project besides just entering prompts.

Some of the people arguing about this don’t think that law is enough.

Some of them, if you asked what would be enough, would tell you the law should prohibit copyright on any art made with ANY assistance from a gen-AI program.

And if you bring up any of the above-mentioned questions about how they would define “a gen-AI program,” and probe deeper into exactly what art their definitions would and wouldn’t allow? …Then, sometimes, they’ll just get mad and start accusing you of being in league with the enemy.

And these accusations start sounding a lot like the Wattpad antis who start yelling “pedophile” as soon as they realize their definitions are breaking down.

I mean. Yeah, we’ve got to be able to define these things sometimes! Especially when they’re important!

So, if all questions about “how you’re gonna write the definition” just get shut down with accusations labeling you as the worst category in the proposed definitions…. that is active sabotage of any real attempts to regulate anything.