Runway and Nvidia debate whether AI can take over an actor’s performance
Runway co-CEO Cristóbal Valenzuela argued at Turin’s View Conference that AI could replace human actors. Nvidia’s Richard Kerris emphasized the spontaneous work between a director and performer, while both speakers discussed the creative skills filmmaking still requires. Their exchange puts different expectations about controllable digital performances alongside the unresolved question of material used to train generative models.
Artificial Intelligence··Evening
Spontaneous performance divides the speakers
Runway, a company developing generative video tools, and Nvidia executives debated AI’s place in filmmaking at View Conference, the visual-effects gathering in Turin, Italy. In Variety’s account, Runway co-CEO Cristóbal Valenzuela argued that AI could replace human actors. Nvidia’s media and entertainment executive Richard Kerris emphasized the spontaneous exchanges between actors and directors. He recalled a director’s example of Marlon Brando improvising with a cat during The Godfather. Valenzuela responded that models could eventually become directable in a similar way. Their exchange concerned expectations for future performances rather than a measured comparison between an actor and a generated character.[1]
Filmmaking still calls for character and composition
Kerris also stressed that filmmaking requires an understanding of character, composition and story, rather than a single button that makes a movie. Valenzuela’s published essays describe a future in which the cost of generating images falls and audiences produce personalized entertainment as quickly as they consume it. Those are his forecasts about production economics. His essay on the cost of a frame focuses on high-production-value content, the area in which many conference professionals work. Kerris compared resistance to generative tools with reactions to earlier changes such as computer-generated imagery.[1]
Training material remains a point of dispute
Visual-effects artists and animation professionals at the conference raised concerns about the material used to train generative models. Asked about training data, Valenzuela described models as learning representations rather than functioning as databases. Runway is among the companies named in a proposed class action by visual artists over the alleged use of their copyrighted work without permission or compensation. The conference discussion therefore included both company expectations for creative control and artists’ objections to the use of their work in developing those tools.[1]
