The U.S. Copyright Office solicited comments on “Copyright Law and Policy Issues Raised by Artificial Intelligence Systems'' and received thousands of responses. Comments from many authors and organizations such as the Authors Guild and the American Publishers Association argue that large-scale language models rely on large-scale piracy to work. The Authors Association says:
Today's leading LLMs were “trained” on a dataset of approximately 200,000 books, plus millions of news articles and books collected from the internet. All book datasets that we know of have been copied from pirated e-book sites, and nearly all other books have been collected by the authors of datasets compiled from web crawls. (e.g. Common Crawl collected 60 billion domains in 12 years) are also copied from pirated sites.
AAP agreed, saying, “Several soldiers…[erative] “AI systems include pirated books in their training datasets,” it concludes, and “fair use should not encourage large-scale piracy.”
The response from tech companies like Google, Meta, and OpenAI that want to exploit this technology is instead defending their methodologies. OpenAI states that “the factual metadata and fundamental information that AI models learn from training data is not protected by copyright law,'' and Meta also “extracts unprotectable facts and ideas from copyrighted works.'' “That in itself is not copyright infringement.” Whether the extraction is done by humans (e.g. by learning from books) or by technological processes. ”
Of course, to extract that metadata, they mined thousands of illegally obtained copyrighted texts, but OpenAI says they had no choice. OpenAI has certainly made some concessions, stating that the company's chatbot “ChatGPT recognizes certain prompts that appear to be intended to reproduce significant portions of potentially copyrighted works.” “They are taught in post-training to refuse to respond.” This “post-training” follows numerous lawsuits from authors who discovered that ChatGPT actually generates copyrighted material on demand.
You can search these comments and many others.
One notable response to the Copyright Office was filed by another government agency, the Federal Trade Commission, and raised a number of issues. The FTC further commented on this matter in a press release. It reads in part:
The way companies develop and release generative AI tools and other AI products has raised concerns that they can harm consumers, workers, and small businesses. The FTC has investigated risks associated with the use of AI, including violations of consumer privacy, automated discrimination and bias, proliferation of deceptive practices, scammer schemes, and other types of fraud.
The FTC is concerned about the “scope of rights and liability under copyright law.” For example, not only can a creator's ability to compete be unfairly compromised, but also consumers can be deceived if the creator does not match the consumer's expectations. If the product is created by AI, consumers may think that the work was created by a particular musician or other artist. ” The FTC notes that “conduct that may violate copyright law may also constitute unfair competition or unfair methods.” Unfair or deceptive conduct, especially if copyright infringement deceives consumers, exploits the reputation of creators or diminishes the value of existing or future works, reveals personal information, or in any other way , which causes serious damage to consumers. ”
In a press release, the FTC is using its existing legal authority to take action against AI-related misconduct, saying that Amazon and Ring used its existing legal authority to take action against AI-related misconduct, and that It cites examples of consumer protection, including claims that it used confidential data it collected.” Although they acknowledge that AI is still evolving, “AI already has the potential to transform many industries and business practices. Therefore, the FTC will vigorously use the full scope of its authority to protect Americans from deceptive and unfair practices and to maintain open, fair, and competitive markets. I plan on keeping it.”
while you are here Please support us Trajectory With a one-time or recurring donation. We rely on reader donations to keep our magazine and site running, and we'd like to keep our site paywall-free. we need your financial support Continuing to provide quality coverage of the science fiction and fantasy field.
©Locus Magazine. Copyrighted material may not be republished without permission from LSFF.
