Creator of impossible interviews

During the last event/performance we did in Calvi on AI and Creativity and which I told you about a couple of days ago, Gaia Reposati and I had fun with the audience interviewing virtual characters, we also created one together during the evening.
Today I’ll put you the link to the web app we used. The project is not entirely mine, I started from this repo on GitHub:

I’ve modified it a bit here and there, integrated gpt-3.5-turbo instead of LLAMA2 (but I’ll replenish it to create a blend between the two) and improved user and character management. If you are interested I will report the repo on GitHub with my changes.

In the meantime, however, I must tell you that this tool has proved to be very interesting, also from a “didactic” point of view. Mind you, there’s nothing you can’t do on your own with ChatGPT (for now though, because I’ll add a couple of features that will make it even more interesting and unique) but it does it in a fairly simple and elegant way, especially if you want to demonstrate something to the public.

The process is very simple, you create a character, write his story (writing in the second person, as if you were giving indications about the character to an actor) and add an exchange of jokes between “human” and character (this step serves to characterize it further, giving it a particular and unique way of speaking for example).

It may seem like a simple little game but it turned out to be a stimulating and useful process for creative approaches. Let me explain.

Talking to a fictional character, or maybe Albert Einstein, can be fun, but let’s try to go further. The fact of having to define a character through the description of his story, and trying to define his way of speaking is already in itself a creative process that is far from trivial.

If you waste some time, you get amazing results. For example, we created a “Cloud” character inspired by one of our installations to which we said that we could only express itself with flashes and animations of lights and colors. And chatting with this “Cloud” was very interesting, the AI tried to personalize the character, also suggesting interactions and animations that created a personality anyway. I don’t rule out creating a Cloud that is fully controlled by an AI anytime soon. And so, for example, you can indulge in creating, that is, a cook who responds only with recipes, trying to associate them with the reaction to your phrases or questions. Or create a character that is strongly focused on the use of a particular tool or make it interact in unusual ways. It’s funny Mona Lisa responding to interactions not with sentences, but describing elements of the painting, giving meaning to this or that “her self” detail. In short, it is definitely a stimulating process.

Play with it