OpenAI and others will likely face more copyright lawsuits as systems like DALL-E produce copyright-infringing materials without attribution or informing users
including movie, TV, and computer game scenes and characters. Reductio ad absurdam: trademark-infringing output from a two word prompt ("animated toys"): https://garymarcus.substack.com/ ... … Neil Blevins / @artofsoulburn@mastodon.art : Really interesting article showing how a user can violate copyright laws using genAI without naming the IP. This means a user could honestly want to not violate copyright but the dataset is likely to do it anyway and then doesn't tell them. What a mess! https://garymarcus.substack.com/ ... [image] Baldur Bjarnason / @baldur@toot.cafe : “Things are about to get a lot worse for Generative AI” — Like I and quite a few others have been saying for ages, generative models are very prone to verbatim copying both text and image. People in tech pretend it doesn't happen but it's much more prevalent than people think https://garymarcus.substack.com/ ... Bluesky: Revan / @docrevan.xyz : This article documents how Dall-E and other image generators will spit out copyright infringement, but I am going to point out it's funny that with the generic Mario prompt ("video game plumber") Dall-E generates Rule 63 Mario. https://garymarcus.substack.com/ ... [image] X: Eliezer Yudkowsky / @esyudkowsky : All IP law took a giant wrong turn at the first point anyone envisioned an exclusive license, rather than a compulsory license (anyone can build on the IP without asking, but pays a legally-determined fee). Gary Marcus / @garymarcus : If OpenAI does collapse, WeWork style, it will likely be seen as a tale of hubris. They knew infringement was going to be a major issue. They proceeded anyway. That will not play well in front of juries. In the end it could devour them. Jonathan Zittrain / @zittrain : @ylecun FWIW, the choices need not be exclusive - more than one party involved can be liable. And, of course, sometimes there are defenses like fair use to what is otherwise clear infringement. I'm not sure how much prior doctrine helps here, tbh. Seems a largely new situation. Paul Jeffries / @pauljeffries : @ylecun Under U.S. law, facilitation of infringement is actionable. That is distinct from infringement. If a person uses a tool to make an infringing work, the person is the infringer. If someone then publishes that work, that is a further infringement by the publisher. In most cases in the U.S., a platform is not regarded as a publisher. I used to be head of legal operations at Facebook (where you are now — as Meta) and would be happy to chat about this stuff sometime! Antonio García Martínez / @antoniogm : This whole AI training vs. copyright issue is such a fascinating collision of tech vs. media, creation vs. remix, plus questions of what it means to own intellectual property, or even be human. And it's at the core of how we'll navigate our hive mind. Feels like *the* issue. Drew Olanoff / @yoda : it's funny how the argument for AI is around creativity and being tools to create, when they would be absolutely worthless without the zillion things created already Cecilia Ziniti / @ceciliazin : 🦜OpenAI seems to have fixed verbatim content parrot-backs, at least since NYT put together Exhibit J. Some copyright-aware answers from ChatGPT ... “I'm sorry, but I can't provide verbatim excerpts from copyrighted texts” “I can't complete the paragraph” “I can summarize or... [image] Peter Hense / @peterhense : A year ago in January 2023 I wrote: “The training of large language models represents the most significant expropriation of creators in the history of mankind. And all of this is happening without any legal foundation. Where is the outcry from the victims?” It's coming.... Cecilia Ziniti / @ceciliazin : “Copyright is not about the money. It's about ... ALL of the money.” -overhead in a sleeper hit class in law school - a law & economics copyright seminar - from the guest speaker, a lawyer for a big rapper at the time, like P. Diddy. Marge Nelk / @nelkmarge : @Cornubot ... One thing is evident from these frantic attempts at damage control - both OpenAI and Midjourney are visibly concerned about infringing on the intellectual property of major companies, lacking a deal for training data. That means big companies can sue. Ben Landau-Taylor / @benlandautaylor : >NYT wins the OpenAI lawsuit >AI companies purge all training data that hasn't aged out of copyright >Chatbots retrained on 1880s guys who'd call a debate club an “athenaeum” >Renaissance of culture, wisdom, and civilization Ethan Mollick / @emollick : AI represents a novel enough thing that trying to put it in existing categories will often cause issues We need to make new policies to address the questions AI raises, rather than lawsuits & courts. How do we want IP to work in an AI world? How about privacy? Use in 🏥 or 🏫 ? Bob E. Hayes / @bobehayes : “Unless and until somebody can invent a new architecture that can reliably track provenance of generative text and/or generative images, infringement - often not at the request of the user — will continue.” ~ @GaryMarcus #GenerativeAI #ArtificialIntelligence Justine Moore / @venturetwins : We're definitely winning the copyright battle, guys. These Italian brothers look nothing like Mario and Luigi. [image] Gary Marcus / @garymarcus : OpenAI is in a heap of trouble, and it's not just text. Long thread why (1/n), based on work with @Rahll [image] Yann LeCun / @ylecun : Genuine question about image generation: If someone uses a generative AI tool to produce an image that is substantially similar to a copyrighted piece (a drawing, painting, movie screenshot, etc), who should be liable for copyright infringement? Should it be: A. the company... Andriy Burkov / @burkov : It's unlikely that OpenAI will win against The NY Times. The reason for this is simple: they don't know how ChatGPT works and thus will have a hard time answering the judge's question: “Is it possible that your model reproduces the copyrighted content verbatim? If yes, can you... LinkedIn: Rafael Brown : Generative AI is built on copyright infringement. As we go back to a broader moment of Cloud AI, generative AI wil gave its Napster moment … Daniel : It's no longer a matter of “if.” Its now a matter of “when” the top companies in America stop using these tools. — Fortunately we have a solution. … Frith Tweedie : IP infringement issues with generative AI aren't going anywhere - the New York Times case is likely just the tip of the iceberg. … Jon Lam : Amazing Substack by Gary Marcus, as he and Reid Southen (Concept artist for film Marvel Studios, DC, Hunger Games, Matrix etc) … Forums: Hacker News : Things are about to get worse for generative AI See also Mediagazer