Topics / AI products

Attributed viewpoint · Not a direct quote

Interactive models need realistic counterfactuals

Anastasis Germanidis argues that interactive models should produce equally realistic outcomes for different actions. He sees this as an important gap between video models and world models.

Behind the viewpoint

Translations are for reading; original excerpts remain the evidence.

Runway’s WorldPrompt and the Engineering of Real-Time Worlds

Original excerpt

if you have an interactive model, you want it to be able to generate counterfactuals. Like, if I take this action versus this action, you want it to generate equally realistic outcomes. so that’s, I think, the big gap between video models and world models

Additional context is not included in this excerpt. Read the original conversation.

Open the episode and seek to 1:08:33.

Start time comes from the supplied transcript. Playback alignment is awaiting review.