In the last years, there have been major advances in the field of machine learning and Artificial Intelligence (AI). The fact that machines are getting smarter raises an obvious question: Are we humans still smart enough to fool intelligent systems? If we are able to deceive smart systems, what does that say about their security and robustness? If, on the other hand, we cannot fool them, what does this say about our control and agency in a society where smart machines are ubiquitous?
In this interactive installation, festival visitors are invited to deceive an AI. Thery are asked to upload a selfie to an AI system that has been trained to identify humans. The goal is simple: Fool the AI. Participants have to disguise themselves so that the AI does not recognize them as a human being. To do so, they can use physical props or makeup. By participating and sharing their thoughts about AI in general and our system in particular, visitors can contribute to our research into the “public understanding of AI”.
The Misidentify.me installation was exhibited at the InScience science film festival in Nijmegen in November 2018 as part of a joint research project with Yana van de Sande, Gabi J. Schaap and Tibor Bosse.
We also published the following paper and abstract:
Schraffenberger, H., Sande, Y. van de, Schaap, G., and Bosse, T. (2019). Investigating people’s attitudes towards AI with a smart photo booth [extended abstract] In BNAIC/BENELEARN 2019. CEUR Workshop Proceedings, 2491.
Schraffenberger, H., Sande, Y. van de, Schaap, G., and Bosse, T. (2019). Can you fool the AI? – Investigating people’s attitude towards AI with a smart photo booth. In Proceedings of the Seventh Conference on Computation, Communication, Aesthetics & X, xCoAx’19.
Here some impressions from the exhibit:
