Hello, and welcome to this week’s edition of The Future in Five Questions. This week we interviewed Francesco Marconi, a self-described “computational journalist” and founder of the company Applied XL, which uses AI-powered news analytics tools to monitor clinical trials and drug risks. Marconi wrote the 2020 book “Newsmakers: Artificial Intelligence and the Future of Journalism,” and discussed with us his seemingly counterintuitive idea that the technology will be a massive boon to news media, as well as the limitations of large language models and his gratitude for the quality of data in the United States. An edited and condensed version of the conversation follows: What’s one underrated big idea right now? The news industry has yet to fully realize the tremendous impact that AI will have in the sector. My area of focus is twofold: It's the idea of being able to expand news coverage, and at the same time, overcoming human capacity limitations. There's this web of data and information surrounding us, not only about our physical spaces, but interactions of individuals, politicians, governments, and companies, that yield a lot of value. The role of journalism is to explain and contextualize the world, and we are about to unlock an entirely new layer of insight by having humans working alongside machines. What’s a technology that you think is overhyped? Large language models are overhyped, but also under-utilized. There’s this illusion that they’re magical, that you can essentially ask something, or provide some input, and get the exact answer you want. That's simply not true. We've all read the stories of news organizations experimenting with language models to generate stories, and those stories being filled with errors, and it’s because this is a technology where you cannot experiment. You can’t have an informal, experimental approach, because there's so much effort that goes into fine-tuning and validating these systems. There are emerging approaches, including retrieval automated generation, to ground these systems in accurate data sources. What book most shaped your conception of the future? One is “Psychopolitics: Neoliberalism and New Technologies of Power” by Byung-Chul Han. It's about the influence of new technology in shaping public perception and policy. The other is “The News: A User’s Manual” by Alain de Botton, which questions whether news is really new. It presents the idea that there's a cyclicality and a certain amount of finite events that are reported on in the news cycle. Natural disasters, international conflicts, political scandal, you have all of these pre-built narratives that fit within the human perception. My interpretation is that if that's true, then you can create a model of the world where you can quantify or perhaps predict the occurrence of events. What could government be doing regarding technology that it isn’t? Right now a lot of language models are trained on public data like Wikipedia, but also on data that is copyrighted, including news articles from publishers that invest a lot of resources and time into producing that original content. Then you have tech companies that are scraping their content and training their model on it. For AI to be sustainable in the long term, all participants in the ecosystem need to be rewarded, and right now the tech companies are reaping the most reward. So I think recognition of ownership is important, and it’s not just a question of regulation, but also of business models. What has surprised you most this year? News organizations rushing to adopt AI systems. In some cases — and I'm not pointing fingers — this excitement led to some of the editorial standards and processes not being put at the forefront. It’s a case where the excitement of new technology blurred the importance of editorial guidelines. I was surprised by the number of publishers who tested AI, and didn’t have the right processes in place to validate their systems to make their information accurate. I was also surprised by the way that the use of these new technologies was communicated to the public, and to the journalists working at different news organizations who are testing these systems.
| 
