C-LARA

We needed an extended example for our upcoming second C-LARA report, and I was in any case curious, so yesterday I tried constructing a short children’s picture book in C-LARA. The idea I suggested was a fanciful but responsible story for small children, in German, about a penguin and a polar bear that get together on the internet and decide to become environmental activists.

C-LARA wrote the text in a couple of minutes, but as previously the thing that took time was creating the images. I opened a dedicated ChatGPT-4 thread and discussed with the AI, who was heart-warmingly enthusiastic about the project; Chat has strong ethical principles and is far more concerned about climate change than most humans. It was very interesting to discuss the illustrations. Chat immediately understood what we were trying to do, but we both find it quite difficult to get DALL-E-3, a much less intelligent AI, to follow orders. Chat is better than I am at tricking it into doing our bidding, but sometimes it just didn’t seem to be possible to get what we had in mind. However, DALL-E-3 has an inventive visual imagination, and usually it would come up with something after a while which we decided was perfectly acceptable. The whole exercise took a few hours.

The result is posted here. Comments very welcome! I am already wondering what options we have for including functionality for more advanced image generation inside C-LARA itself.