Law in the Internet Society

A Digital God is Born

-- By JaredHopper - 05 Dec 2023


In a December 2023 interview between Elon Musk and Andrew Ross Sorkin of the New York Times, the topic of AI came up in a conversation about X/Twitter's content moderation in light of the recent advertising controversy (major advertisers have pulled out after the platform’s algorithm allegedly began boosting antisemitic conspiracy theories). X/Twitter has not only been using AI technology to shape the featuring of particular content, but Musk has started feeding tweets to his own AI program to "train it." Notably, in response to a question about the implications of copyright law and AI learning, Musk responded that "by the time these [copyright] lawsuits are decided, we'll have Digital God." Absurd diction aside, it's worth thinking about the nexus between the increasingly capable technology and the interest in developing it with the legal protections and principles of copyright law. If AI will become "Digital God," is copyright meant to bow down to this new, thieving creator? This question seems easier to answer as related to properly licensed software/other works than it does artistic materials, like theatrical set design, that don’t fit neatly into the bounds of copyright protection.

Wikipedia as a Training Ground

Wikipedia is licensed under Creative Commons and is thus available for commercial usage, meaning that users are relatively free to share and edit the text so long as it is attributed correctly.

No, you've got the license terms quite wrong. It's CC-BY-SA, not CC-BY, and that makes a world of difference. Let's correct for that and see where we are.

But how does this work with AI training? It seems that AI could freely use the information to educate itself, but in folding that knowledge into the commercial product that AI will increasingly become, it is not clear how companies can properly attribute Wikipedia’s contributions when the information is integrated into the Digital God’s mind without clear delineation? If it wanted to, Wikipedia could absolutely change their license and prohibit AI usage, but this seems counter to their open-source mission. Some people worry, furthermore, that such free training would “allow[] companies like OpenAI? to exploit the open web to create closed commercial datasets for their models.” This concern is less about copyright than it is the integrity of such knowledge, driven by the fear that AI interpretation upon AI interpretation will warp the information displayed on sites like Wikipedia.

In general, if we want to train AI to be a responsible source of knowledge, allowing free access to open-source sites like Wikipedia seems like a smart move. Monetizing Wikipedia isn’t the goal, but safeguards may be necessary for companies who work on a for-profit model, such as explicit restrictions and software that monitors and curbs AI behavior.

Current Case Study: AI Set Designs

A larger problem with AI and copyright can be found in the context of theatrical set design. Set designers have an important job, and as producers seek to cut costs more and more, designers are pressured (and probably required in some spaces) to create a less physical and therefore burdensome set and, instead, use digital screens for seamless and cost-effective scene transitions. Such "AI Set Design" began floating around on social media about a year ago, and many are excited about the visual outputs already created. Protecting set designs from widespread distribution was previously not a major concern, since theater is rarely, if ever, available on the internet. However, with the pandemic and the increase in the availability of theater “pro-shots,” a variety of works (such as “Hamilton” on Disney+) now circulate on the internet and are thus ripe for illegal copying and re-posting.

Unlike properly-licensed software, set designs are (1) rarely protected by copyright and (2) are more difficult contexts to monitor AI theft, even with explicit licensing restrictions. If AI is inspired by a design, how is that any different from a human artist being inspired by a work? If we could determine exact copying, we might be able to police this, but it is hard to say that ever-developing AI software could not quickly get around this.

Concern of copyright in this space, however, may be unfounded. Some set designers are actively encouraging using AI in their designs; they are so certain that their jobs will not be replaced and that AI really is just "computer assisted drawing" that will prove more helpful than harmful. But what about the source material for these AI interfaces? They aren't just able to compete with the Tony Award-winning set designers out there without reason, and that reason is that they are learning from the IP of preexisting designers. When asked about OpenAI? 's claims that their AI is not using copyrighted information to learn, Musk brusquely responded that "that's a lie." And unlike with human copyright infringement claims, an algorithm cannot (or should not be able to) claim that a work, though inspired by another's, is independently created to skirt liability.

Ramifications for Creatives on X/Twitter

Opening up these questions to Musk's new AI baby is necessary for those who consider the above AI use for set design plagiaristic. The problem in talking more specifically about using a platform like X/Twitter as creative fodder for AI is that it necessitates talking about many different kinds of copyright concerns. Not only do we have the same initial (ethical? legal?) concern about AI directly basing its designs off of preexisting work without the creative spark we have been told is innately human, but there is also the difficulty of copyright protections generally on the social media website. Even though you “retain the rights to anything you post,” if the artist is not the one who posts, then their work is not protected. Unlike Wikipedia, set designers may eventually start caring about copyright infringement. However, it may not be necessary to involve the law: can preexisting social norms be enough?

You are entitled to restrict access to your paper if you want to. But we all derive immense benefit from reading one another's work, and I hope you won't feel the need unless the subject matter is personal and its disclosure would be harmful or undesirable. To restrict access to your paper simply delete the "#" character on the next two lines:

Note: TWiki has strict formatting rules for preference declarations. Make sure you preserve the three spaces, asterisk, and extra space at the beginning of these lines. If you wish to give access to any other users simply add them to the comma separated ALLOWTOPICVIEW list.


Webs Webs

r4 - 21 Jan 2024 - 21:08:35 - EbenMoglen
This site is powered by the TWiki collaboration platform.
All material on this collaboration platform is the property of the contributing authors.
All material marked as authored by Eben Moglen is available under the license terms CC-BY-SA version 4.
Syndicate this site RSSATOM