Highlights from AI x Journalism House at SXSW 2024

What will an AI-influenced information ecosystem look like in five years?

What will an AI-influenced information ecosystem look like in five years, and how can journalism as a discipline better work with large language models and other powerful technologies? What role should journalism play in a world of hyper-personalized content?

To discuss these questions and more, Hacks/Hackers hosted an AI x Journalism House in the heart of downtown Austin alongside SXSW, Austin’s annual mega-festival devoted to the convergence of technology, film and music. For two days, we organized presentations, discussions, and workshops attended by nearly 200 people with speakers and attendees from news and technology organizations including The New York Times, BBC, The Texas Tribune, American Journalism Project, Macarthur Foundation, Texas Monthly, Online News Association, Axel Springer, Der Spiegel, the Poynter Institute, Hugging Face, Cohere, Yahoo, Microsoft, Meta, and many others.

Zach Seward: Creating sense with generative AI

With support from the Online News Association and the Brown Institute for Media Innovation at Columbia University, highlights of the events included a fireside chat with Zach Seward, The New York Times’ newly appointed editorial director of AI initiatives, for a background conversation with about 50 AI x Journalism House attendees.

Seward also presented a solo talk at SXSW, his first since joining The New York Times, and coordinated by Hacks/Hackers. Seward’s “state of the nation” about AI in journalism, “AI news that's fit to print,” began by discussing “when AI journalism goes awry.” However, rather than focusing too much on the “bad and ugly,” Seward spent time highlighting inspiring uses of artificial intelligence for journalism:

So how might we think about AI journalism that works? Well, it's got to be vetted, in a rigorous way. The idea should be motivated by what's best for readers. And, above all, the first principles of journalism must apply: truth and transparency.

Highlighting positive, powerful use cases that incorporate artificial intelligence and large language models (LLMs) in journalism, Seward discussed how his former company Quartz worked with the International Journalism Consortium to build a tool that was able to analyze an “enormous cache of documents — too big for any human to go through page-by-page” and report on law firms that specialize in hiding wealth overseas.

While machine learning can identify patterns in data, Seward said, generative AI’s utility can be found in creating patterns and creating sense. As an example, Seward provided an overview of a journalism initiative by The Marshall Project:

The Marshall Project, a non-profit newsroom covering the U.S. justice system, has been investigating what books are banned in state prisons and why. It maintains a database of the actual books, but also got its hands on the official policies that guide book banning in 30 state prison systems.

These are often long and esoteric documents, and The Marshall Project wanted to make the policies more accessible to interested readers. Its journalists went through each document to identify the parts that actually mattered, and then Andrew Rodriguez Calderón, using OpenAI's GPT-4, employed a series of very specific prompts to generate useful, readable summaries of the policies. Those summaries were then reviewed again by journalists before publication.

Zach Seward’s talk at SXSW 2024 featured many more examples and insights about AI in journalism, as well as slides, scripts, and references, and is worth checking out.

News product, transparency, bias, and potential harms: AI x Journalism House at SXSW 2024

Hacks/Hackers’ AI x Journalism House also featured two days of hands-on programming for journalists. The sessions kicked off with a facilitated discussion about the role and process of journalism and news product development with Trei Brundrett, a senior advisor to American Journalism Project and the Aspen Institute.

Another highlight was a presentation from Benjamin Toff, assistant professor at University of Minnesota’s Hubbard School of Journalism and Mass Communication, about how people judge AI-generated content that has been labeled as such as less trustworthy than human-generated content. This is a paradox, notes Toff, that could disincentivize media organizations from labeling AI-created content.

Brigitte Tousignant, communications lead at Hugging Face, shared some newsroom use cases about operationalizing ethical AI. Tousignant explained that, as an organization, Hugging Face aims to democratize good machine learning by acting as a hub to share models, datasets and demos, facilitate collaborations to advance the open-science movement, and surface open-source standardization and tooling.

It’s all based on a philosophy, Tousignant said, around open-source that includes rapid AI advancement; AI technologies should be accessible and inclusive, and transparent. With that in mind, open-source AI should also feature auditability and reproducibility, and should allow for tailored-use. You can see Tousigant’s full presentation to AI x Journalism House here.

Another highlight of AI x Journalism House, Poynter held a conversation on AI and ethics that dug into topics including labor, tools, partnerships, climate impacts, transparency and audience trust.

The full list of talks supported by Hacks/Hackers at AI x Journalism House at SXSW included:

• “The Paradox of AI Disclosure for Audience Trust in News,” by Ben Toff, assistant professor, Hubbard School of Journalism and Mass Communication, University of Minnesota.
• Hugging Face and open-source AI for newsrooms, by Brigitte Tousignant, Hugging Face.
• Dive into LLMs, applications and prompt engineering, by Jay Alammar, Cohere AI.
• Ethics of generative AI, facilitated by Alex Mahadevan, director of MediaWise at Poynter. Local journalism and AI: What’s happening at The Texas Tribune, by Liam Andrew, Tribune chief product officer, Ayush Patel, CTO of Attri.
• Navigating the AI Revolution: Insights from 40+ Industry Leaders on the Newsroom Robots podcast, by Nikita Roy, Newsroom Robots.
• Dive into LLMs, applications, and prompt engineering, by Brigitte Tousignant, Hugging Face.

What’s next for Hacks/Hackers on AI and Journalism?

To move from discussions to product prototypes, join Hacks/Hackers and the Brown Institute for Media Innovation at Columbia University for an in-person open-source AI hackathon from April 5-7. Sponsored by Hugging Face and Codingscape, the hackathon lasts the entire weekend, and provides a hands-on opportunity to experiment and even build generative AI tools for journalism workflows. See signup details here.