Quick-n-Dirty noAI MIT license | Hacker News Reader

Quick-n-Dirty noAI MIT license

gist.github.com

4 pointslelanthran11 comments

There are these anti-AI licenses popping up en masse now, and I don't see the point. They get the intention across, yes, but at least under American law training AI either is fair use or it isn't.

If it is, your new license doesn't matter, they may use it. If it isn't, your new license doesn't matter, they're lready acting illegally. Is this argument wrong?

> If it is, your new license doesn't matter, they may use it. If it isn't, your new license doesn't matter, they're lready acting illegally. Is this argument wrong?

Well, yes. "It hasn't been tried in court yet" is not the same as "Precedent says this is explicitly allowed".

Add in the fact that the main AI players are negotiating commercial agreements to use publicly available content from the main publishing corporations tells you everything you need to know about how they think the lawsuits are going to turn out.

From the PoV of the main AI companies, they've already indicated that they know what the outcome is going to be.

> Well, yes. "It hasn't been tried in court yet" is not the same as "Precedent says this is explicitly allowed".

Please read my comment carefully, without simply assuming I must be on AI companies' side. I'm covering two exhaustive, mutually exclusive cases. In both cases I think your license does nothing, compared to a normal MIT license.

> they've already indicated that they know what the outcome is going to be

I agree with you! But even in the case where they are in the wrong, they are already in the wrong with a standard MIT license!

And in the other case: fair use trumps the license text. That's the point of it!

> In both cases I think your license does nothing, compared to a normal MIT license.

I've read it very carefully, and I'm still of the opinion that an explicit prohibition needs to be tested in court before it can be ignored.

IOW, until it is tested, it stays. After it is tested, then your comment about it being superfluous will apply. Right now neither of the two alternatives apply.

If someone wants to make a test case out of my little prohibition, they are welcome to.

I wanted a quick MIT based license that imposed restrictions on usage for machine training/machine learning.

This is what I came up with.

So not FOSS?

> So not FOSS?

What makes it not FOSS? After all, FOSS comes with restrictions too, you know, like the GPL.

Does the restriction on using the licensed content to train models in any way deter you from using the licensed content at all?

Yes? I can't use the licensed model to train content?

In the same way I can't use BSL to use the software in the cloud or Meta's model license when I can't be a big company.

* BSL - https://github.com/clockworklabs/SpacetimeDB/blob/master/LIC...

* Meta - https://github.com/meta-llama/llama-models/blob/main/models/...

> Yes? I can't use the licensed model to train content?

Well, that's the point; free software does not mean "no restrictions". Most of the FLOSS licenses come with restrictions too, just different ones. That doesn't make them "not FLOSS".

If you're arguing that licenses can't be FLOSS if they have restrictions, then you're going to have an uphill battle to convince people that the GPL isn't FLOSS.

I think there’s room for an AGPL for ai but I’m probably going to die on that hill between MIT/BSD/Apache2 and GPL.

Try proposing FSF. They might be interested

FOSS is usually defined by the FSF's Free Software definition, and the Open Source Initiative's Open Source Definition. I don't think either field-of-use restrictions nor per-user royalties are acceptable under either definition.

https://www.gnu.org/philosophy/free-sw.html https://opensource.org/osd/

Reply on news.ycombinator.com