By Porter Anderson, Editor-in-Chief | @Porter_Anderson
Bläsi: ‘Remaining Problems Won’t Go Away’Heating up right along with international temperature gauges this summer, the issues around “artificial intelligence” and publishing seem to become more contentious weekly.
The United States’ Authors Guild updated its information on Wednesday (July 19) to say that more than 10,000 “writers and their supporters” have signed an open letter to CEOs of AI companies including OpenAI; Alphabet (parent of Google); Stability AI; IBM; and Microsoft.
As frequently happens in the Guild’s approach—which is not unlike that of many NGOs issuing their statements on various issues—there’s an impressive list of big names being rolled out here, the advocacy organization having attracted signatories including Dan Brown, James Patterson, Jennifer Egan, Margaret Atwood, Jonathan Franzen, Roxane Gay, Celeste Ng, Louise Erdrich, Viet Thanh Nguyen, and George Saunders.
With the arrival of large language models comes both a need and opportunity to better focus the core purpose of publishing and determine “what will be the specific role for humans?”Christoph Bläsi at Readmagine
At the core of this protest is a very real alarm that the source material on which a large language model might be “trained”—the texts used in advanced statistical algorithms’ collection of linguistic content patterns—may well be copyrighted works. Franzen is quoted by the Guild saying that the organization is “advanc[ing] the rights of all Americans whose data and words and images are being exploited, for immense profit, without their consent—in other words, pretty much all Americans over the age of six.”
Certainly on its face, this copyright challenge is immediately and urgently part of a deepening and widening body of alarm now being reflected by elements the actors’ and writers’ strikes in Hollywood. While writers’ vulnerability might be closer to that of the writing corps in book publishing, the parallel threat to actors is unmistakable: their likenesses and voices can be artificially captured and manipulated, giving the broader AI controversy an easily understood visual component. The crisis of residual payments from many streamers may be the immediate money-ask in those labor actions, but as Andrew Dalton has written for the Associated Press, “Artificial intelligence has surged to the forefront of Hollywood’s labor fights. … The technology has pushed negotiations into unknown territory, and the language used can sound utopian or dystopian depending on the side of the table.”
At a national governmental level, in the States on Friday (July 21), Cat Zakrzewski writes at the Washington Post, “the Biden White House on Friday took its most ambitious step to date to address the safety concerns and risks of artificial intelligence, announcing that seven of the most influential companies building AI have agreed to a voluntary pledge to mitigate the risks of the emerging technology, escalating the White House’s involvement in an increasingly urgent debate over AI regulation.”
International Regulatory Explorations
In international policy, the European Parliament’s announcement of an EU AI Act was made in June. Michelle Nichols at Reuters, in her coverage of the UN Security Council’s first meeting on artificial intelligence, wrote that China’s UN ambassador Zhang Jun said Beijing agrees with a concept of central coordination on ground rules for AI, cautioning that the world must “prevent this technology from becoming a runaway horse.”
“The lack of consensus on what to include in datasets is a reflection of the lack of societal consensus on what we want the capabilities of generative AI to be.”GenLaw: 'The Devil Is in the Training Data,' Lee, Ippolito, and Cooper
Many in world book publishing might be forgiven for feeling that a squadron of start-ups and vendors is forming just outside the door, ready to swarm the subject (as happened during the “digital transition” just years ago), offering the inevitably described “tools” and “solutions.” These may be of scant help to a book business still struggling to understand the weightier challenges of AI’s developments–such as how to protect the intellectual property on which publishing and its authors are entirely dependent.
One early case has had oral arguments before the US District Court for the Northern District of California. A a class-action suit of Stability Ltd. asserted that “AI companies’ decision to include their works in the dataset used to train their image generator models is a violation of their copyrights,” as Shanti Escalante-De Mattei writes for ArtNet. “Because their work was used to train the models, the artists argue, the models are constantly producing derivative works that violate their copyrights.”
If the case is dismissed, as some expect, she writes, she writes, it could be because “current copyright law is not equipped to address the potential injustices engendered by AI.”
And Katherine Lee, Daphne Ippolito, and A. Feder Cooper make clear in the title of their GenLaw paper, The Devil Is in the Training Data, “The lack of consensus on what to include in datasets is a reflection of the lack of societal consensus on what we want the capabilities of generative AI to be.”
Christoph Bläsi at Readmagine:
Perhaps what made Christoph Bläsi’s presentation so effective at last month’s Readmagine program at the Fundación Germán Sánchez Ruipérez‘s (FGSR) Casa del Lector complex in Madrid was the fact that he didn’t skirt these broader, more fundamental issues in favor of the gee-whiz, look-how-handy-this-is approach of some.
Very much along the lines of the GenLaw assessment, publishing, in Bläsi’s understanding, publishing is a formidably creative industry imbedded in a society that has no idea what it actually wants from AI.
Devised by FGSR director general Luis González with FANDE managing director José Manuel Anta, Readmagine allowed Bläsi a free hand in assessing what—at least this summer—is a keen overview of the potential crises that AI applications could deliver to publishing amid evident potential advantages.
These, in Bläsi’s purview, range more broadly across the literary landscape than in the realm of potential copyright infringement and commercial chore management to include a new, needed valuation of what is not possible to automate: human creativity.
Bläsi is a professor of book studies at Johannes Gutenberg University Mainz, and has a background of work with German publishing houses. His presentation roamed with unusual comfort that gray area between the potential day-to-day utilitarian uses of AI in publishing and these much broader questions of at what cost the faux “intelligence” of these systems is gained by the marvelous speed and dazzling reach of their ability to vacuum up great swaths of data.
Bläsi refers to the work of Michael Bhaskar, whose forthcoming book is Mustafa Suleyman’s The Coming Wave: Technology, Power, and the 21st Century’s Greatest Dilemma (Penguin Random House UK/Bodley Head, September 7). Bhaskar’s writings in the past have pointed to four key areas of impact predictive in these technologies on publishing:
Even heightened productivity, Bläsi noted, might lead to overproduction–in our international markets that already are annually saturated with books. More worrisome, of course, are real and still-gathering threats of censorship, climate-crisis challenges, and inadequate levels of diversity “in the industry and its products,” something many in the business are studying.
The industry, as he noted, already “is working on a huge number of titles in very low numbers.” AI’s impact might, he said, create even “more echo chambers and filter bubbles” than we see now in various subject- and issue-driven sectors of the business.
Despite the fact that Bläsi is a congenial and upbeat personality among colleagues in travel, he took care not to let some rosy, uninformed assumptions stand unchallenged.
“Conglomeration will not go away,” he told the Readmagine audience, not least because the leading AI models are being developed by American mega-corporations. Likewise, he said, “Censorship and attacks on freedom of speech won’t go away.”
As Bläsi’s sums up his top-level points:
- AI applications in the book-media system are a reality, and they stand at different stages of the value network; these span the industry from authoring and “gatekeeping” (editing) to predictive pricing and recommendations. In other words, they run from the point of creativity to the book business’ success or failure in the marketplace.
- With the arrival of large language models, we see both a need and opportunity to better focus the core purpose of publishing, and to define job profiles around the least automatable and most complex value contributions, together with a confident use of algorithms. That is to ask, where will humanity really count in a newly refocused mission and potential for publishing?
- For publishing education, the moment shouldn’t be treated as an application add-on, but as a new focal construct of its own.
Here is Christoph Bläsi’s presentation from Readmagine, “AI in Our Industry”:
More from Publishing Perspectives on artificial intelligence and its debate relative to publishing is here, more on digital publishing is here, more on Readmagine is here, more on the work of Luis González is here, more on world publishing conferences is here, and more on Spain’s publishing market is here.