Pigs sign partnership with the meat grinder
Condé Nast is the latest company to capitulate to OpenAI with a deal to have its content chewed up and regurgitated.
Previously: Newspaper coverage of A-level results day? It's a failing grade
One of the biggest lies in technology is that a particular development is inevitable. It’s at the heart of the story that OpenAI likes to tell about its brand of generative AI. Having plundered content across the web to train its models, it’s now engaged in a campaign of signing deals with publishers to gain access to their archives and new work as it’s published. This week’s agreement with Condé Nast to use content from brands including Vogue, Wired and The New Yorker in ChatGPT and SearchGPT follows similar deals signed in recent months with the Financial Times, News Corp, Le Monde and Axel Springer.
Condé Nast CEO Roger Lynch’s email to staff, reported by The New York Times, reveals the mix of fear and misguided hope driving the strategy. He wrote:
Generative AI is rapidly changing the ways audiences are discovering information. It’s crucial that we meet audiences where they are and embrace new technologies while also ensuring proper attribution and compensation for use of our intellectual property. But as
of noted: ChatGPT only has around 100 million weekly users, while Snapchat (which is hardly a source of constant buzz) has 410 million daily users. Companies are indeed pushing AI features on users with increasing impatience — unwelcome AI overviews now intrude at the top of every Google search — but their level of interest simply doesn’t match the hype.Lynch claims that Condé Nast has teamed up with OpenAI to “meet audiences where they are” but the numbers suggest they aren’t there yet and may never be. OpenAI is a cat burglar that has persuaded its targets to open their front doors wide rather than have it creep in the back or sneak in through a skylight. It wants them to believe that there is no other option, even as The New York Times and The Intercept continue to pursue legal action against the company for scraping their articles.
What makes Condé Nast’s capitulation so foolish is that it’s entirely contrary to how it usually tries to sell its brands. Vogue is about luxury, Wired purports to bring its readers visions of the future early, and The New Yorker’s entire reason for being is to produce in-depth, original, and human reporting. The glossiness of print does not easily translate online but allowing OpenAI to chew up that content and spit it out further devalues outlets that have always traded on exclusiveness.
OpenAI is haemorrhaging money (some estimates suggest it could post losses of $5 billion by the end of this year) and doesn’t have anything like a sustainable business model. It also needs much more organic data to train its models as the alternative of using synthetic data produced by other models is likely to result in the AI equivalent of mad cow disease. OpenAI needs more data and persuading publishers that they’ve no other option than to collaborate with with it is one way to feed that hunger.
The promise/threat OpenAI is holding over fearful and desperate media executives is that their audiences will desert them for the ease of a chatbot so they might as well sign a deal now and get in on the action. Those media companies are also seduced by the dream of using AI to generate most of their content in the first place, doing away with swathes of those inconvenient human journalists with their demand for pay rises, holidays, and benefits.
The media companies’ greed will come back to bite them as it did when they were lured into retooling their strategies and cutting their workforces to cater towards the vagaries of Facebook traffic in the ‘pivot to video’ era. Taking original material and offering it up to be chewed up and regurgitated by a large language model which is highly unreliable at crediting its sources devalues that work. Rather than serving steak produced by a familiar farm, Condé Nast will give up its product to be minced and pumped out as one ingredient in the information equivalent of pink slime.
Generative AI can’t produce original reporting or seek out things that are new and interesting. It’s a purely regurgitative technology. Journalists and anyone who values what journalism can do at its best should be fiercely opposed to the idea that it can somehow provide a future for publications. OpenAI’s deals with media companies are reminiscent of the Borg in Star Trek: “You will be assimilated. Your uniqueness will be added to our collective. Resistance is futile.” But, in this case, resistance is not futile. Publications and individual journalists should fight to retain their uniqueness and avoid becoming just another node for the AI model to draw upon.
Publishers who make a deal with OpenAI aren’t just failing to defend the value of the work their journalists produce, they’re also trading what little credibility they have for cash from companies that intend to make them obsolete. They’re like pigs who’ve been persuaded they’re engaged in a mutually beneficial partnership with the abatoir.
Thanks for reading.
You can also follow me on Twitter, Threads, BlueSky and/or TikTok.
If you haven’t yet, please consider upgrading to a paid subscription.
It helps and allows me to spend more money on research and reporting. Buy a t-shirt if you’d like to make a one-off contribution and get a t-shirt.
Speaking in a purely personal capacity here. And I realise it's a niche query in the grand scheme of things. But as a former freelance contributor to titles owned by at least two of the publishers who have signed deals with Open AI and/or their competitors, I'm unaware of any attempt being made by either AI firms or publishers to ensure that only the "content" owned by the publisher will be used under these agreements, and/or to ensure freelance contributors are paid if and when their work is re-used for this purpose. While absence of information doesn't confirm anything, it would come as a bit of a surprise to find that both sides had taken the time and trouble to design suitable mechanisms to correctly identify freelance contributions and ensure they're not ingested during this process. Certainly, there is no indication as yet that publishers are going to pay freelances for such re-use. There is precedent here in the newspaper sector, with the attitude shown by the Newspaper Licensing Agency (a body established by newspaper publishers to collect reproduction royalties from the likes of libraries) to freelances, which, in essence, is: "So little of what we publish is freelance-generated that it is not economically viable for us to account for it properly, so what we'll do is take a percentage of our income and donate it to charity". Again, to stress, this is just guesswork on my part and I have no evidence to suggest this is what's happening: but I would be entirely unsurprised if the deals cut between publishers and AI companies have been built around a similar view, and will, in essence, deny freelance contributors any share of income from the deal, and not give those contributors an option to refuse to allow their work to be swallowed and digested by the LLM. So not only are these deals likely to prove self-defeating for the reasons you eloquently lay out, Mic, but in the process the rights and incomes of freelance contributors are presumably being further eroded as well. And this in an era where, I would wager, the proportion of media "content" being produced by freelances is increasing, thus making it less attractive or viable a prospect for a freelance to consider working for companies who have agreed deals with LLM developers.
Thanks for this Mic. Could be the lateness of the hour/old age & decrepitude, but I’m still struggling to keep up with AI, or even to know what it’s really about. Have difficulty preventing stories going into the same bin as Bitcoin or Y2K. Bad attitude, I know...