Botcasts – a combination of the terms “bot” and “podcast” – are AI-gen - erated radio broadcasts that take topics from real news sources in areas ranging from politics to culture to entertainment and continue them in the form of AI-generated speculative fiction put together by an algorithm.

So-called “alternative facts” seem to have become a strong political instru - ment of disinformation and pose a serious threat to democracies. At the same time, more than half of the global data traffic on the Internet is now produced by automated programs called bots, and both the production and dissemination of misinformation and half-truths is also increasingly done by bots. Botcast aims to explore the infiltrating and disruptive aspect of AI’s influence on digital media.

Technical Write up

The algorithm then “reads” the article and generates an “intelligent” summary using the newspaper3k module of Python. This summary is used as a prompt for the GPT-Neo model, which creates the news stories for the Botcast. While there are many text generation machine learning modules available, GPT-Neo is used here because of its open-source nature. It can easily be substituted with another ML model. In comparison to other models, such as the Python module aitextgen, GPT-Neo generated the most coherent output.

The project was exhibited as part of the Intelligent museum porject at the Biomedia exhibition at ZKM.