AI Will Transform the Business Model of Publishing

Although we are barely over a year into the generative AI revolution, it is already having a significant effect on content creation. Publishers are increasingly using large language models (LLMs) to research, organize, synthesize, and write. This technology has already found a profitable foothold in many high-volume/low-stakes use cases (such as email marketing), but more interesting is the growing use of AI for serious professional writing and communication. Most publishers are now evaluating generative AI, and an increasing number are actively using it across their organizations. A recent report noted that 11% of the finalists for the Pulitzer Prize in Journalism used AI in some way https://www.editorandpublisher.com/stories/five-of-this-years-pulitzer-finalists-are-ai-powered . As it becomes clear that this technology can make professional writing and communication significantly more productive, adoption numbers will skyrocket.

But AI is more than just the next turn of software-driven efficiency in the workplace. It is also a catalyst for changing how content will be monetized across digital channels. Publishers have been working for years to wrest control of their customers and revenue back from search and social media companies. Paywalls, subscriptions, and pay for view channels have started to shift some value back to content creators. AI presents an entirely new opportunity.

The large language models at the heart of the generative AI revolution are trained on content, and researchers are finding that higher quality content leads directly to better models. But what rights do the model trainers have to this content? High profile lawsuits https://www.nytimes.com/2023/12/27/business/media/new-york-times-open-ai-microsoft-lawsuit have already begun to define the boundaries. At stake is whether training an LLM on copyrighted content is “fair use” or whether it must be compensated. While the legal issues are new and still to be decided, the industry is not waiting. Licensing deals between publishers and the companies behind these models are becoming more and more common, a strong signal that this will become the legal and business standard.

If licensing for training purposes becomes a new revenue source for publishers, what about the inference phase of the LLM lifecycle, when the fully trained model is used to generate new works? One of the unique aspects of generative AI is that source material can influence model output in many ways. A generated article fed by a source document may pull directly from that source (such as a quote). In that case, normal fair use and cite/source rules apply. But the derivative story may also be wholly influenced by the source document (whether or not a direct portion is replicated). In such a case, it seems equally inevitable that the owner of the source material should be compensated.??

Compensating rightsholders for the use of their work in inferences, however, presents some major practical challenges. While LLM training involves the wholesale use of fixed amounts of readily attributable content, compensating inference requires the ability to do attribution and compensation via microtransactions at the very moment the model is being prompted to generate output. The only way to do inference-time attribution with compensation is via an end-to-end platform. A marketplace for content and LLM generations can enable writers/consumers to find and use relevant content with appropriate payment to rights holders for each use. At Symbolic.ai , we are working with Publishers in news, communications, and research to build just such a platform.

Will these new revenue sources be meaningful? While we are at the very beginning of this revolution, forward-looking publishers are embracing the model and building for the future. It’s likely to be a self-reinforcing feedback loop. AI will become more and more useful, and adoption will grow. This will create more demand for high-quality content, and more value for the publishers who support the creation of that content by participating in the marketplace.? Over time, this will be transformational for the industry, and change the way they do business at a scale not seen since the early days of the internet. The result of this should be publishers that are more profitable and sustainable, and therefore capable of creating more high quality professional content such as journalism. It is down to the industry itself to be a participant in the transformation and help build this future.?