HalfCheetah: Half Cheetah #24

Half Cheetah by James Bloom uses Reinforcement Learning AI models running in real-time. Each piece shows runners on a never-ending journey towards an unknown destination. 3D scans of humans act out behaviours generated by mammalian AI agents. A second RL model then analyses their motion in real-time and re-orders movement dynamically according to alternative goals. There is no improvement towards efficiency. Instead, an abstract flow of body motion continues in perpetuity, trending towards repetition and failure. Functioning software that has no utility but nevertheless exists.