Tomo Kihara & Playfool
How to avoid being spotted by AI in the city
How (not) to get hit by a self-driving car
Commentary
How does AI recognize what people look like?
Having learned from a vast amount of image data, AI can distinguish in an instant between a human and a dog.
Automated driving, which is expected to see increasing use, aims to improve safety even further through the use of this AI technology.
However, what if a child was hiding in a cardboard box? Pretending to be a dog? Could AI detect that these were humans?
This work is a game where you have to get over a pedestrian crossing without being detected by the AI as a pedestrian.
If the AI can detect you as human, you lose.
To win, you need to trick the AI’s eyes, finding ways to avoid being identified as a pedestrian, and reach the goal.
What happens is that each time the player wins, the results are stored as data where the AI could not detect a pedestrian.
Through collecting the many different variations of human actions through play, this work provides AI with fresh perspectives that conventional data do not allow, and could contribute to further technical development.
Profile
Tomo Kihara & Playfool
A collaborative team between Tomo Kihara and Playfool (Daniel Coppen). Using “play” as a theme, they design tools that foster creativity and exhibit art projects in Japan and overseas that intervene in society and cities. Recently, they have exhibited at the Victoria and Albert Museum and at Ars Electronica, among others. Photo : Tada (Yukai)
