To repeat what I keep saying in these cases: there’s nothing special about the particular license you use, if it still has pretty much any condition. If Copilot is subject to copyright restrictions, MIT-licensed code is affected just as much as AGPL-licensed code: Copilot is still not meeting (and cannot meet) the condition of the license, attribution. Copilot depends entirely upon exemption from copyright restrictions under “fair use” exceptions.
But still, you’re quite welcome to issue takedown notices, though I don’t know if the regular process is actually appropriate (given that it’s a GitHub project rather than a project hosted on the GitHub platform). Their reaction might be interesting, though I expect they will just declare the notice invalid because of fair use exemptions and ignore it.
I'll approximate: Contract violations are okay if they result in no damages. Courts determine damages by harm done. This comes down to a lot of detailed issues, including things like intent. Licenses are more complex, and I'll ignore issues like statutory damages.
A lot of licenses are GPL-compatible because although the terms differ in phrasing, the GPL has provisions which are substantively the same (e.g. attribution versus copyright notice). Courts won't care.
MIT signals one intent, and AGPL signals another. That's a big difference. I have code under AGPL, GPL, and MIT, and in the case of the AGPL code being ripped off by copilot, it undermines a lot of the purpose of having licensed the code as AGPL if others can use it without that license. For the MIT code, a lack of copyright notice+attribution would typically be mild annoyance at best.
A court would pick up on that, and act accordingly.
It's interesting to see this statement. When it comes to Copilot, many HNers share the sentiment that it has copied code from open-source projects without their permission.
However, when it comes to a different group such as artists, the same people seem to believe stable diffusion/DALLE-2/midjourney et al. are "just weights in a neural network and the neural network is generating art based on its understanding of text and style".
I feel like that provides a valid ethical framework in both cases.
For example, if my AGPL code is taken and ends up in another open-source project, just without attribution, I'm okay with it. If it's used for a father-son duo making an awesome thing for Burning Man, unaware of where it came from, I'm fine with that too. If it's used by a large commercial organization to compete with my open0source project, I'm not okay.
I don't feel bad using SD at home for personal use. It's a lot of fun. I would feel bad, for example, using it to make advertisements or Hollywood films. I don't feel bad about most non-commercial and educational settings.
Your line might be a different place, but there's a line somewhere, and it's not all-or-none.
Are these truly the exact same people commenting? HN has a wide range of interests, so I think it's possible that you're combining two different groups of HNers into "the same people".
I, for example, have some views about Copilot, and think that there's a case for saying that it violates copyleft code licenses. I...do not have a view about art, mainly because I literally have no idea how copyright works in the art world, or if there's something similar to copyleft; so I usually don't comment on the AI-generated art stuff that seems to be popping up left and right.
It's interesting to see this statement. When it comes to Copilot, many people share the sentiment that it has copied code from open-source projects without their permission.
However, when it comes to a different group such as artists, also many people seem to believe stable diffusion/DALLE-2/midjourney et al. are "just weights in a neural network and the neural network is generating art based on its understanding of text and style".
You should. What’s the point of licenses if big companies ignore them?
Personally I’ve only ever released source as MIT, knowing it could be resold or used for nefarious things, or kept source private. I don’t trust ransoms on the internet to respect posted licenses, so anything I post publicly I post with as much permission as possible.
Well, supposedly, the next step is suing them ? This of course still requires some power and influence, but if the violation is egregious enough you generally can find a rights defence association willing to help you.
(Of course the DMCA is still problematic for grey areas and spurious claims, like here.)
One can sue, but it won't resolve the problem of double standards. As far as I know, you can't sue google for automatically acting on DMCA claims when they are sent by certain people of power and influence and leaving all other claims to the courts.
To repeat what I keep saying in these cases: there’s nothing special about the particular license you use, if it still has pretty much any condition. If Copilot is subject to copyright restrictions, MIT-licensed code is affected just as much as AGPL-licensed code: Copilot is still not meeting (and cannot meet) the condition of the license, attribution. Copilot depends entirely upon exemption from copyright restrictions under “fair use” exceptions.
But still, you’re quite welcome to issue takedown notices, though I don’t know if the regular process is actually appropriate (given that it’s a GitHub project rather than a project hosted on the GitHub platform). Their reaction might be interesting, though I expect they will just declare the notice invalid because of fair use exemptions and ignore it.