At the second meetup of the AI for Media Network, a significant milestone was achieved: The AI + Automation Lab of BR introduced a tool for verifying AI-generated content using AI itself – and released it as open-source. The reception was overwhelmingly positive.
On October 9th, over 100 AI practitioners from various media sectors gathered at BR for the second AI for Media Network meetup. Organized by BR and Ippen Digital, the event focused on reducing hallucinations and enhancing factual accuracy in Large Language Model (LLM) outputs.
Second Opinion: When AI Verifies AI Output
Marco Lehner and Max Brandl from the AI + Automation Lab presented “Second Opinion”, an application that allows one AI to verify the content generated by another. On the Second Opinion website, users can input a statement, which is first summarized by ChatGPT. In the second step, ChatGPT checks the summary against the original text for any discrepancies.
Second Opinion Released as Open-Source
What makes this unique is that BR is releasing “Second Opinion” under an open-source license. This allows the code to be integrated into existing news production workflows. For more details on how Second Opinion works, you can read the report on the BR Next blog.
Uli Köppen, Chief AI Officer at BR, explains the rationale behind the open-source release of Second Opinion: “In the AI for Media Network, we not only want to exchange learnings, we also want to share methods and code. ” She hopes this initiative will set a precedent.
dpa and Krone Hit Integrate AI into News Workflows
In addition to this, there were two other noteworthy presentations: Digital consultant Marcel Tuljus showcased the workflow used by the Austrian radio station Krone Hit to generate unique news content for its four new DAB+ stations using generative AI.
Sara Ghaeizadeh and Arne Beckmann from the news agency dpa demonstrated two AI tools employed in their newsroom: Radiomat, a prompt library that helps staff generate text suggestions for audio news pieces from dpa texts within seconds, and the AI Workbench, a system independent of the editorial system designed for experimenting with prompts and prompt chains and sharing them with others.
New format “AI Spotlight” well received
Following the presentations, the three teams engaged with the audience, addressing numerous (technical) questions in detail.
In the new “AI Spotlight” segment, three participants had the opportunity to briefly present a current AI project from their work – an idea that was very well received by the audience.
The meetup concluded with a networking session. In the foyer outside the main conference hall, many personal connections were made between AI leaders from various German, Austrian and Swiss media outlets.
Meetup Documentation
- The presentations for the three projects can be found here.
- The video recording is available here. (password-protected).
- And this was the agenda of the meetup.
- The team from Ippen Digital interviewed some of the guests, see their reactions here:
AI for Media Symposium in November
On November 13th and 14th, the AI for Media Network will convene for a symposium at the Akademie für Politische Bildung in Tutzing. The event will focus on how quality media can maintain their standing in an information landscape increasingly shaped by AI. The next regular meetup is scheduled for the first quarter of 2025.
To stay updated on all news and events from the AI for Media Network, you can subscribe to our email newsletter here.
