Authors Express Mixed Reactions to Chatbot Training
Prominent authors have expressed mixed reactions to the revelation that their books are being used to train AI chatbots. The books in question have been pumped into the electronic brain of LLaMa, a chatbot developed by Meta Platforms (formerly known as Facebook) as a competitor to OpenAI’s ChatGPT bots. Some authors, such as John Grisham, George R.R. Martin, and Scott Turow, have expressed outrage and even filed copyright infringement lawsuits against OpenAI. Others, like Margaret Atwood and Lauren Groff, have taken to social media to voice their irritation. However, there are also authors, like Ian Bogost and Stephen King, who see the use of their books as an opportunity and have adopted a more relaxed attitude towards it.
The Nature of AI Chatbots and their Training
Before delving into the legal issues at hand, it’s important to understand the nature of AI chatbots and how they are trained. These chatbots, including LLaMa and ChatGPT, fall under the category of “generative AI.” Contrary to the term “artificial intelligence,” they are not truly intelligent in the way humans and animals are. They are designed to appear intelligent by using probabilistic strings of words and phrases to produce responses that resemble human thought. However, they do not possess true understanding or creativity. Chatbot developers train their systems by feeding them vast amounts of text from the internet or specialized databases, which the chatbots then use to generate responses.
The Significance of the Books3 Database
The disclosure that the Books3 database has been used to train LLaMa highlights a crucial aspect of chatbot technology. It reveals that everything produced by chatbots ultimately originates from human sources. While OpenAI has not confirmed its use of Books3, it is highly likely that ChatGPT has undergone similar training. This raises questions about the value and ownership of the content used to train chatbots. Authors question whether their contributions should be compensated and worry about the impact on their ability to make a living from their works.
The Legal Concerns Surrounding Chatbot Training
The reactions from authors reflect the confusion and complexity of copyright law, particularly when it comes to “fair use.” Fair use allows for the use of copyrighted material without permission in certain contexts such as reviews, summaries, and parodies. However, the emergence of new technologies like chatbot training challenges the boundaries of fair use. Authors and the Authors Guild argue that the use of their works in training chatbots constitutes wholesale copying without permission or compensation, amounting to what they deem as “systematic theft on a grand scale.” They claim that the chatbots can create derivative works that harm the market for their original works. Interference with the marketability of copyrighted material is a significant factor in court cases involving fair use defenses.
The Impact on the Marketability of Authors’ Works
Authors are concerned that chatbots like LLaMa and ChatGPT, which can generate texts resembling their works, diminish the demand for their original creations. This threat to their marketability poses a significant challenge to their ability to earn a living. The encroachment of AI into various professional fields, as seen in recent strikes by Hollywood writers and ongoing negotiations with actors, highlights the importance of establishing limitations on the use of AI technologies.
The Lawsuits and Meta’s Defense
Authors, including Sarah Silverman, have filed lawsuits against Meta Platforms, mirroring the Authors Guild’s case against OpenAI. Meta contends that its use of the Books3 database is transformative and falls within fair use. However, the alleged piracy involved in creating Books3 may weaken Meta’s defense. The outcome of these lawsuits will depend on whether the court deems the chatbots’ output as reproducing substantial portions of the authors’ works and whether the interference with marketability is considered in favor of fair use.
Authors Deserve Compensation for their Contributions
While some argue that using published and copyrighted material for chatbot training does not require authors’ permission, this perspective fails to acknowledge the fundamental difference between recommending a book and copying it. OpenAI and Meta are commercial enterprises that stand to profit from their chatbots, making it crucial for them to recognize the value of the content they use to train their systems. Authors deserve compensation for their contributions, especially if their works were used without payment or acknowledgment.
