It was tempting to skip wrapping up 2023, because at first glance large language models seemed so thoroughly dominant (and boring to revisit), but bringing the net.wars archive list up to date showed a different story. To be fair, this is partly personal bias: from the beginning LLMs seemed fated to collapse under the weight of their own poisoning; AI Magazine predicted such an outcome as early as June.
LLMs did, however, seem to accelerate public consciousness of three long-running causes of concern: privacy and big data; corporate cooption of public and private resources; and antitrust enforcement. That acceleration may be LLMs’ more important long-term effect. In the short term, the justifiably bigger concern is their propensity to spread disinformation and misinformation in the coming year’s many significant elections.
Enforcement of data protection laws has been slowly ramping up in any case, and the fines just keep getting bigger, culminating in May’s fine against Meta for €1.2 billion. Given that fines, no matter how large, seem insignificant compared to the big technology companies’ revenues, the more important trend is issuing constraints on how they do business. That May fine came with an order to stop sending EU citizens’ data to the US. Meta responded in October by announcing a subscription tier for European Facebook users: €160 a year will buy freedom from ads. Freedom from Facebook remains free.
But Facebook is almost 20 years old; it had years in which to grow without facing serious regulation. By contrast, ChatGPT, which OpenAI launched just over a year ago, has already faced investigation by the US Federal Trade Commission and been banned temporarily by the Italian data protection authority (it was reinstated a month later with conditions). It’s also facing more than a dozen lawsuits claiming copyright infringement; the most recent of these was filed just this week by the New York Times. It has settled one of these suits by forming a partnership with Axel Springer.
It all suggests a lessening tolerance for “ask forgiveness, not permission”. As another example, Clearview AI has spent most of the four years since Kashmir Hill alerted the world to its existence facing regulatory bans and fines, and public disquiet over the rampant spread of live facial recognition continues to grow. Add in the continuing degradation of exTwitter, the increasing number of friends who say they’re dropping out of social media generally, and the revival of US antitrust actions with the FTC’s suit against Amazon, and it feels like change is gathering.
It would be a logical time, for an odd reason: each of the last few decades as seen through published books has had a distinctive focus with respect to information technology. I discovered this recently when, for various reasons, I reorganized my hundreds of books on net.wars-type subjects dating back to the 1980s. How they’re ordered matters: I need to be able to find things quickly when I want them. In 1990, a friend’s suggestion of categorizing by topic seemed logical: copyright, privacy, security, online community, robots, digital rights, policy… The categories quickly broke down and cross-pollinated. In rebuilding the library, what to replace it with?
The exercise, which led to alphabetizing by author’s name within decade of publication, revealed that each of the last few decades has been distinctive enough that it’s remarkably easy to correctly identify a book’s decade without turning to the copyright page to check. The 1980s and 1990s were about exploration and explanation. Hype led us into the 2000s, which were quieter in publishing terms, though marked by bursts of business books that spanned the dot-com boom, bust, and renewal. The 2010s brought social media, content moderation, and big data, and a new set of technologies to hype, such as 3D printing and nanotechnology (about which we hear nothing now). The 2020s, it’s too soon to tell…but safe to say disinformation, AI, and robots are dominating these early years.
The 2020s books to date are trying to understand how to rein in the worst effects of Big Tech: online abuse, cryptocurrency fraud, disinformation, the loss of control as even physical devices turn into manufacturer-controlled subscription services, and, as predicted in 2018 by Christian Wolmar, the ongoing failure of autonomous vehicles to take over the world as projected just ten years ago.
While Teslas are not autonomous, the company’s Silicon Valley ethos has always made them seem more like information technology than cars. Bad idea, as Reuters reports; its investigation found a persistent pattern of mishaps such as part failures and wheels falling off – and an equally persistent pattern of the company blaming the customer, even when the car was brand new. If we don’t want shoddy goods and data invasion with everything to be our future, fighting back is essential. In 2032, I hope looking back shows that story.
The good news going into 2024 is, as the Center for the Public Domain at Duke University, Public Domain Review and Cory Doctorow write, the bumper crop of works entering the public domain: sound recordings (for the first time in 40 years), DH Lawrence’s Lady Chatterley’s Lover, Agatha Christie’s The Mystery of the Blue Train, Ben Hecht and Charles MacArthur’s play The Front Page. and the first of Mickey Mouse. Happy new year.
Illustrations: Promotional still from the 1928 production of The Front Page, which enters the public domain on January 1, 2024 (via Wikimedia).
Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon
