Popularizing research with AI

We've been experimenting with AI (specifically large language models) as a tool for popularizing scientific research by talking accessibly, if imprecisely, about a topic to kick off some discussion. In this way, scientists can shift the imprecision to the AI (which not surprisingly is pretty good at it) thus without being imprecise themselves. This was something that e.g. Mike Friedlander in Physics at WashU always looked askance at Carl Sagan for, when Carl was just trying to communicate with a larger audience. The table of links below, based on research in my group, illustrates with some examples. 

Popularizing Science with help from AI's Imprecision in Gemini Storybooks (top link) & Youtube Playlists (bottom link) 

In sections below, the top link is to an interactive storybook (about 3 minutes long if read by AI) , while the bottom link is not to a topical playlist of many youtube videos (as above) but to a single GoogleLM "video" slideshow (about 7 minutes long). Narratives below the individual youtube videos generally point to the more technical source papers, for those interested.

Caution: As usual with large language models, lack of precision accompanies perhaps unexpected connections, with hopes of at least starting a conversation about the topic.