The rapid rise of artificial intelligence in recent years has touched every aspect of our lives.

There are clear potential applications, including in healthcare, scientific research and education.

The risks are also all too obvious. Deepfakes and AI-driven disinformation dominate the headlines, yet the potential impact of AI deployment on society, government, employment and the environment is still not understood.

Cosmos has a long tradition of communicating science to a wide audience. Our scientifically trained journalists report on the latest research and interview the people behind the stories. Cosmos prides itself on fact-based reporting, including coverage of AI research and deployment.

We decided to bridge the gap and become part of a scientific experiment to answer important questions about the use of AI in the media. Our goal is to report on our experiences with the use of AI in our newsroom and provide evidence-based examples of the risks and opportunities for science communication.

Our first realization in hindsight is that we needed to communicate the project more clearly before testing it and publishing test articles. We could have done better. We have paused the project briefly while we review the feedback and requests we have received so far, and remain committed to ensuring responsible and ethical AI practices. We value your feedback, positive and negative, as well as your questions.

What is the Cosmos AI project?

We have more than 16,000 stories in our archive from writers, including current and former staff members, as well as a number of freelancers who have been commissioned to write articles for Cosmos. This content has long been used by our journalists as part of their research for newer articles or explainers, such as how we might travel through time. Our archive gives us a solid source of facts to create more explainer articles.

Explainer videos are important because they help our readers understand science. They have been popular on our website for many years. They are an educational tool for all ages. They are fundamentally different from news and features.

Our journalists have written explainers in the past, but this has prevented them from reporting on breaking news and interviewing researchers. We decided to test whether an AI system could help us create explainers to support our journalists’ work. The explainers must be factually accurate to promote facts rather than misinformation.

The Cosmos AI system uses a retrieval-augmented generation (RAG) model. First, it identifies articles in the archive that are relevant to the question or topic. This saves the journalist from having to manually cross-check the archive at the beginning (though important human editorial involvement is required later in the process – see below). RAG reduces the risk of “hallucinations” where the AI ​​can make things up.

The second step is the writing phase, where a large language model (LLM) is used. After researching various platforms, Open AI’s GPT-4 (GPT-4) was selected. GPT-4 helps write the explanatory texts. Our story archive is not used to train GPT-4 or other LLMs. We do not use ChatGPT, the publicly available product from Open AI that many readers will be familiar with.

The last step in our process is the most important. We use trained science communicators to fact-check and edit the first drafts of the statements created with the help of the AI ​​system. Nothing will be published without at least two real people, including our journalists, reviewing, editing and finalizing the statements. We will monitor how long this work takes and compare it to the time it takes a journalist to write and fact-check the statement from scratch.

Cosmos received a grant from the Walkley Foundation’s Meta News Fund 2023 to carry out this project. The project will run from March 2024 to February 2025. The Cosmos AI system was created by an external developer who has been working with Cosmos for many years. The system is not publicly available, but we will continue to share our project results and experiences publicly.

If we want to ask the question, “what is the point of AI in media?” we need to explore how it can be used and question the results. If we cross that line, it needs to be underpinned by an unwavering commitment to ethical and responsible journalistic practices.