From Publishing Perspectives:
In a statement issued from San Francisco today (May 9), the subscription service Scribd has “clarified how its data may be used in an update to its terms of service.”
This update, according to the company, “emphasizes that Scribd’s users, subscribers, and partner companies may not utilize the company’s data for monetization or to train large-language models without Scribd’s explicit consent.
“Additionally, Scribd confirmed that it has not allowed any companies that train large-language models to use full content provided by its publishing partners, which is only available through its digital subscription service.”
This is just the latest, of course, in quickening reactions and evaluations of “artificial intelligence” in the publishing and content realm, several points about which were addressed on Monday (May 8) in the Association of American Publishers’ annual general meeting.
During that live event, AAP president and CEO Maria A. Pallante laid out a gratifyingly comprehensive overview of issues that the US and international publishing industry needs to consider amid the popular giddiness and occasional doomsday chatter around systems such as ChatGPT introduced by OpenAI.
Among the most pressing questions Pallante poses—each having bearing on Scribd’s unusually broad, sector-crossing offerings. From Pallante’s message to the United States’ publishers:
- “Consider academic publishing. Each year more than two million articles are published in more than 26,000 research journals following peer review and curation that is painstaking, but essential to ensure integrity and confidence and research results. How can AI tools help with this mission? What threats does it pose?
- “Consider education publishing. There’s an old saying that people are entitled to their own opinions, but not to their own facts. What are “facts” in the context of AI? A percentage of truth? How will learning be amplified or cheating be contained?
- “Consider trade publishing. Do we as a society want AI-generated works flooding the Internet, potentially depressing the value of human authorship? If we can’t contain AI-generated works, what should be the ethics about disclosing their provenance?”
Link to the rest at Publishing Perspectives
As PG has mentioned previously, based on his understanding of how AI programs utilize written works of all kinds, he doesn’t think they’re violating US copyright law because AI doesn’t reproduce the text protected by copyright.
During his experiments with AI writing programs, the closest PG has seen direct references to the written works of others is a prompt that asks for the AI to write something in the style of Hemingway, Fitzgerald, Nora Roberts or Lucy Score. The AI writing from those prompts presents no danger to the future royalties earned by Ms. Roberts or Ms. Score.
(PG notes that academic publishing generally produces the most turgid collections of words known to humankind.)