How are newsrooms using generative AI for journalism?

The video opens by acknowledging the ongoing ethical debates surrounding AI in newsrooms and highlights practical applications observed by the Assistant Director at the Center for Cooperative Media. Newsrooms are primarily using AI to boost productivity and efficiency, reducing cognitive workload for journalists and allowing them to focus more on community needs.

Generative AI is being used for content summarization, generating key takeaways from long documents, suggesting alternative headlines, and crafting social media posts. It also serves as an editorial check for biases, reporting gaps, and source diversity, helping journalists understand public interest in their stories.

The Marshall Project utilizes ChatGPT to simplify complex bureaucratic text into summaries for public service journalism and for textual analysis to classify themes. They emphasize a 'human-machine hybrid approach' with robust human oversight to maintain editorial integrity and overcome resource constraints, while also stressing the importance of involving historically marginalized groups in the design process to ensure responsiveness to community needs.

ARL has been testing AI for newsletters, using it to condense information, create summaries and bullet points, simplify research, perform text correction and copy editing, and categorize articles. AI also generates two-minute audio news summaries, a 'thought of the day,' and a 'haiku of the day.' For back-office tasks, tools like 'yesio' (a Slack integration) reduce time spent on headline generation, SEO keywords, and ensuring content visibility.

The speaker personally uses AI to reformat text into various styles (Markdown, CSS, HTML), valuing its ability to save time on tedious tasks and reduce AI 'hallucinations' since it's reformatting existing information rather than generating new content. A more ambitious application involves using AI to scrape newsroom archives, offering both internal insights and a new way for the public to interact with reporting by asking questions across entire story histories using context embeddings.

The video concludes by stating that AI is a transformative way of interacting with information and data. As AI capabilities expand, direct natural language inputs are likely to translate into executable actions, allowing users to not just converse with AI but have it change code, edit content, generate user profiles, and create websites, marking a fundamental shift in our relationship with machines.