FTC Continues to Wade into Copyright Issues in AI Without Understanding Anything
techdirt.com
techdirt.com
Am I reading this right? That posturing is frustrating.
Still... The argument that developing these tools under the assumption that publicly available data is fair use is what allows upstarts to develop LLM AI tools instead of giant companies is very astute. I think it's 100% correct. It makes me understand why LLMs are a battleground for copyright reform.
I appreciate that the article makes the stance of both sides understandable. It's enough to make me begrudgingly say that I'm glad for having read it.
Similarly, statements like: "they held with a ridiculously one-sided panel of folks from the creative industries, who more or less all agreed that everyone should be forced to give them money at every opportunity." just reflect Techdirt's own bias as to what they believe the outcome should be.
I also think that Tech Dirt's analysis of the FTC's position is conclusory regarding the downstream effects of increasing wealth concentration and reducing competition. It seems like Tech Dirt came to this in the absence of any thought as to what the implications would be for competition in a world where training data of copyrighted materials don't need to be licensed. I'm not so sure that such an outcome would improve competition in creative fields, and I'm not seeing anything from Tech Dirt explaining how they got there.
I cannot currently search the web for any sort of art without drowning in a swamp of AI results. Even if the SEO-tag-laden midjourney slop wasn't ugly as sin (it is), it's drowning out the art made by humans. A 'disaster for creativity' is already happening and if we don't get copyright protections for artists and creators against their works being made into the grist for AI content mills, this will only get worse.
I've found no searches where AI made art is "drowning out the art made by humans". Do you have some example searches where you find this to be true?
Also AI tools will let more people make art they like, and perhaps sell some to others that like it. I suspect this will be like the claims recording music would kill off musicians since up to then music was live. If anything, it expanded music by making it available cheaply and easily to more people. (See the Luddite story on textiles for yet another example).
I work in advertising and it's becoming common to hear that concept artists aren't hired anymore. People are using Midjourney and other tools. The thing is, they are likely not asking for a "random" style, there is a list of 400 or 4000 artists (I can't recall off the top of my head) that Midjourney is deliberately set to imitate. That means oh - I like this Star Wars concept design, can I get more in the same style of This Artist in Particular (AI) is a real thing and it's costing people real money.
My understanding of fair use is that this might be a violation, since one of the factors to be considered is the effect of the use upon the potential market for or value of the copyrighted work.
edit: Here is evidence for the list (4000 artist names) https://twitter.com/JonLamArt/status/1741545927435784424
I don't see anything supporting this claim so I asked for how it was obtained.
Also, styles aren't copyrighted. Specific works are. I can paint in the style of anyone and sell it no problem. I can use whatever tools I want for that. If I clone a specific work, then I may be in violation.
Are we having the same discussion or are you simply injecting your own discussion with yourself into a this thread?
Next, your use of the word "clone" is sloppy - those are clearly not clones in any use of the word I have seen. They are very similar, but clone would imply, well, a clone. Indistinguishable.
As to whether a tool can make a copy of something when the user controls the tool, courts have ruled consistently and repeatedly that when a tool has significant non-infringing uses, such as copiers, tape recorders, cameras, VCRs, CD burners, etc., that when a user infringes copyright using the tool the company is not liable. Its how we all can have cameras and recorders and DVRs and the Internet and a whole host of modern tech - it can all be used for infringement, but it has significant uses outside that. A user can infringe.
Good luck calling this infringement. When the twitter user is asking the tool to infringe that will almost certainly not be MidJourney's liability.
I doubt MS indemnified users of it's tools from copyright claims that their lawyers didn't look long and hard at this. Just because most of the internet is ignorant of the nuance of copyright doesn't mean big law groups (and me too) are also ignorant.
Now, please if you feel the need to respond, stay on the original topic. No sense in you hijacking this thread further.
Err. It expanded the music industry. It did not expand music. Arguably, it killed music just as people feared. All the money and attention is focused on the top 1% artists who make 77% of the money[1], while other artists have to also get a "real job". The only reason many of my favourite artists make music at all is because they're subsidized by the Canadian government[2].
Think about the effect on innovation when every local garage band with a fresh sound has to "compete" with U2 and The Rolling Stones (and their multi-million marketing budgets). This is the kind of consolidation that the likes of Walmart or Amazon can only dream of.
[1] https://www.synthtopia.com/content/2014/03/07/the-long-tail-...
[2] https://www.canada.ca/en/canadian-heritage/services/funding/...
Imagine ie eventually you could get new episodes for any podcast in the world through AI. It would ingest the data from the hosts, it would ingest the data from everyone who's ever been interviewed, and you could ask it for say Marc Maron interviewing Abraham Lincoln. Or if you are into conspiracy theories, you could ask it to produce new episodes of Alex Jones on Joe Rogan, using current headlines as data. You could do this forever.
Sounds awesome
It's perfectly reasonable for the FTC to be involved here, and what constitutes unjust enrichment and anti-competitive practice when covered works are used as components in a training set is within their purview.
The USPTO will be involved in issuance questions, and the legal system will be involved in determining single cases.
On this issue, the following is my opinion (not a lawyer, not legal adv.).
A training set is a collection of representative input, which may or may not be covered works. The training set is a simple aggregation, it does not alter the works contained in any significant way, so it is not transformative. Therefore, the training set is as an omnibus of other works.
The weights of an system are a product of the input training set (if DL is used), and the parameters of the training environment. 1 is an omnibus of potentially covered works, the other is likely non-covered (but maybe patent-eligible in some cases). It is unclear if this is transformative, but it is certainly not covered by any of the existing exemptions (such as covers of songs).
Therefore, it is possible that the product of the above is a non-transformative derivative work based on one or more covered works, and therefore would have required a license or grant from the author(s) of those covered works. Note that authors often retain their ownership of UGC, and grant a service provider the ability to make that available to others in limited capacity for specific purposes. Some service providers (Like stack overflow) require that contributors license contributions under a specific license, and yet others transfer ownership as part of the terms of service.
If the training set was built from only entirely public domain works, then there wouldn't be a question, but the extension of copyright term over the years has robbed the public domain of anything even remotely useful for these kinds of cases.
On a side note, StackExchange contributions are CC BY-SA, which means that if the training input included anything from stackoverflow, the weights could be be affected by the Share-Alike requirement, which could lead to some interesting results if someone ever tried to request them. (However the attribution would probably be pretty easy, because you could pretty much say "all stackoverflow contributors" and be correct)
It's not out of the question that model weights would be <that hypothetical term for "created from works but doesn't copy expression">.
That's "de minimis" copying.
>It's not out of the question that model weights would be <that hypothetical term for "created from works but doesn't copy expression">.
Model weights would be a derivative work that might be a fair use because of its transformative nature.
The issue with derivative works isn't necessarily the extent of the copying so much as that it seems logical that we'd let the creator of Seinfeld have control over whether or not a Seinfeld movie exists.
I feel like many people here struggle with copyright because they try to treat it like a math problem, but it's not a math problem. it's a political and social problem that we try to resolve by applying principles and reason (like much of the law). so if you aren't applying reason and principles, but instead treating it like a computer system and trying to blow it up by stuffing the inputs, etc, you aren't really getting it and you aren't engaging with the actual law, which doesn't work that way and doesn't suffer the problems that computer systems do, in part, due to their lack of ability to reason.
> "created from works but doesn't copy expression"
This is what I'm getting at. Copyright only applies to expressions. So you can't be a copy of something without having copied its expression. For the same reason you can't be a derivative of another work, without having copied some of that expression. To put it another way, if you have a copyrighted work, and you subtract out the expression - there is nothing left of the copyrighted work. There might be things left, but they were never what was captured by the copyright, because copyright only pertains to expressions fixed in a tangible medium.
Having said all that, I don't buy the framing of "big bad creators vs. UwU small bean AI companies" that this article puts forward[1]. For example, OpenAI is Microsoft as far as I'm concerned. And a lot of the training data came from small-time creators on sites like Deviant Art, not from "The Disney Vault", because the former is publicly accessible while the latter is not. Meanwhile, the first movers in AI are already "salting the earth" for any AI newcomers since AI content is now flooding the internet and is inferior for training.
[1] > Last month, it published an FTC “staff report” about “AI and Creative Fields,” based on a roundtable they held with a ridiculously one-sided panel of folks from the creative industries, who more or less all agreed that everyone should be forced to give them money at every opportunity.