Prof. Diepold on improvisation, creativity and Artificial Intelligence

"When AI produces amazing pictures, is it really art?"

Prof. Diepold, on October 23 you'll be on stage as a scientist during an evening of improvisational theater. What do you expect?

I hope that we'll be able to use the interaction on the stage to trigger interaction with the audience. AI evokes very contradictory reactions – people are either thrilled by it or afraid of the implications of this technology. I'd be pleased if we could use this approach to initiate exchange on the topic of Artificial Intelligence.

One of the topics will be the role of improvisation and creativity for Artificial Intelligence – isn't that a contradiction in terms?

There's really no general answer to the question of whether or not computers can be creative. For example, if we take a look at explorative creativity, computers can indeed produce amazing results and can support humans in creative tasks.

What exactly do you mean by explorative creativity?

Explorative creativity refers to creating works within existing rulesets –the exploration, so to speak, of a clearly defined space. One example would be the fugues of Johann Sebastian Bach, which follow precisely defined mathematical rules. A computer can try out all the possibilities allowed within the scope of such rules and can return impressive results. But when it comes to anything beyond that – "Thinking Outside the Box" – well, then computers have a harder time. Maybe we could put it this way: AI can beat humans in games like chess and Go, but it can't spontaneously cheat.

Right now, a lot is going on in the area of computer-generated images. Programs like DALL-E 2 and Stable Diffusion can create impressive images based on a text prompt. Would that still be explorative creativity?

That's certainly an interesting development. It may also be good that art changes as a result – Art is always influenced by the respective available technologies. But now we once again have the reactions to AI that we mentioned before – the euphoria and the fear. We should always perform a reality check: Where are we really? And when AI creates amazing pictures, is it really art? Art exists in a social realm, in a space where people communicate and decide jointly what is and is not art.

Wouldn't it be possible to program an algorithm in such a way that humans conclude that all images or pieces of music it generates are art?

And thus pass a kind of Turing test for art? Yes, perhaps in the future. But is it really enough to fool us with something, or is there more involved? In the meantime, the expert community agrees that the classic Turing test is no real test for intelligence.

Getting back to improv: You've been looking at the topic for quite some time now. Actually, an improv theater workshop is mandatory for participants in your seminar on computers and creativity, "Komputer und Creativität". Why?

In order to find out whether or not computers can be creative, the students first have to develop a feel for what human creativity really is. Improvisational theater is an excellent tool for doing that which doesn't have any additional prerequisites. Furthermore, improvisation is an important part of the engineering profession – we constantly have to improvise in our careers and, more than anything, we have to communicate.

Will you be nervous before you take the stage on October 23?

I won't be nervous, no. But there will certainly be tension – working with that is part of the improvisation concept. That's also part of the fun, and the Fastfood Theater team members are real pros. It's sure to be a great evening with plenty of laughs