If we can teach college freshman to cite their sources, why can’t we fix A.I. art generators so that they do the same?

The way that creators of A.I. art generators and the way that artists both talk about A.I. art generators is wrong. And the proof emerges in calculators. 

Does a calculator own the rights to mathematical proofs upon which its calculations are predicated?

Does a mathematician dictate the future forms and tools that employ his proofs?

No and no. Quite obviously: neither posits the locus of control for ideas. Indeed, ideas as such cannot be controlled.

However, there is the issue of justice — particularly when we’re talking about the minds and careers and livelihoods of the mathematicians in question. Part of the problem comes from our erroneous thinking about what does and does not constitute thinking. These programs aren’t thinking, as Hart says elsewhere.

We, not the A.I., endow it with meaning and purpose. We use calculators to help us think. And we use A.I. tools to help us make art. That much seems clear and someone like Ursula Vernon showed how many of these programs she needed in order to make a single, coherent multi-paneled graphic novel page. 

The issue emerges, as many have said, when the artists used to train these A.I. get zero compensation for this dragon hoard of art the programmers have amassed.

It’s the same problem as Facebook and user data. Or Google / Amazon and books. Who owns the intellectual ideas, activities, and power that these A.I. have stolen?

The artists and the users, that’s who.

And yet, the programmers have indeed created a new tool.

Why not split it, 50/50?

I’m working currently with the center for the commons in Alaska on a small film project concerning these issues as applied to every abstract and concrete resource, so I wanted to run an experiment.

Can you play a fun petition game with me?

There’s a simple solution: either the government or each individual company creates a sovereign wealth fund that invests in the broader economy, out of which is paid a royalty for every instantiation of an image. If an A.I. used an artist to train — or if an artist was invoked in a creative prompt — that artist gets paid out of the fund. If we can teach college freshman to cite their sources, why can’t we teach A.I.?

Here’s the thing: even calculators do this. When you click the button for sines and cosines, you’re actively citing Pythagoras.

PETITION: FIX A.I. ART

We, the undersigned, call our representatives and A.I. art companies to:

1. Reserve a full 50% of every monthly subscription or any other revenue generated from A.I. art in a fund.
2. Invest that half in a sovereign wealth fund in the world economy.
3. Pay royalties out of the earnings of that fund every time an A.I. uses an artist’s work — whether for training or by user prompting, whether in whole or in part, including the large repository of images used to train towards a specific prompt. This royalty — or royalty portion — will pay out in proportion to the use of every other artist, per quarter.
4. This amounts to training an A.I. art generator to cite its sources in the same way we teach college freshman.
5. If this cannot be tracked, the A.I. art company suspends business until it can do so.
6. If a policy along these lines is not adopted, we the undersigned will take our business elsewhere.

This petition is set to deliver to A.I. art companies in order to fix A.I. art generators. Share it far and wide and make sure you tag your representatives (local, state, and national) and any employees of A.I. art companies you know on social media:

30 signed of
100 goal