The Models is an interactive installation exploring truth, wonder and the quirks of generative AI through the improvised theatrical tradition of Italian Commedia dell'Arte. The installation consists of an endless series of scenes generated in real-time with a video game engine, and featuring a cast of characters acting in response to audience input. Characters embody both the archetypal, popular and exaggerated traits of Commedia dell'Arte masks, and tendencies observed in generative AI, such as making things up, being antagonistic, or being overly servile.

Through their phones, audiences prompt characters to act out these tendencies, praising or booing their performances, and dropping controversial props on the virtual stage that symbolise superstitions, falsehoods or popular Commedia gags known as Lazzi. The characters' dialogues and voices are generated with a system of large language models and text-to-audio models assembled by the artists using the Leonardo Supercomputer, housed in Bologna, Italy. Sixty-four different scenographies exist with bespoke lighting design and music composed by Hero Image.