Artificial Life: One Leg at a Time is a multi-channel installation designed to provide a different image of the technology and science behind AI. The project uses the machine learning paradigm called reinforcement learning (RL), which aims to make agents act according to their environment, to maximize the expected benefits.
The installation tries to represent the learning process of AI in a different way from how it is presented in science fiction films and the technology industries. Here AI is presented, not as an intelligent and performing machine since its creation, but as a slow and tiring learning process, subject to failures.
The protagonist of the work is a model trained as an athlete who performs the action of running. The training phase shown in the installation is slow and clumsy, but at the same time fun and moving. What emerges is a complex behavior that is built by artificial neural networks through a process of recursive decision-making, which involves many errors and failures.
The installation tries to raise interesting questions about the relationship between AI and biological life. Although the learning process of AI is based on principles similar to those of biological evolution, its methods still appear clumsy and alienating in relation to those used by the human being.