LeCun Distinguishes World Models from Video Generation
Replies debate distinctions between world models and video generation in AI thread.
Yann LeCun replied that world models per one definition should not be confused with video prediction or video generation models. He stated that understanding dynamics of a system to control it is not the same as producing cute videos. Lucas Beyer replied in the thread that the position differs from an earlier stance describing representations as capturing static semantics of a scene behind an image with no dynamics.
@JitendraMalikCV Exactly. We also should not confuse world models, as per your definition, with video prediction or video generation models. Understanding the dynamics of a system in order to control it is not the same as producing cute videos.
Combined views
LeCun Distinguishes World Models from Video Generation
Replies debate distinctions between world models and video generation in AI thread.
Yann LeCun replied that world models per one definition should not be confused with video prediction or video generation models. He stated that understanding dynamics of a system to control it is not the same as producing cute videos. Lucas Beyer replied in the thread that the position differs from an earlier stance describing representations as capturing static semantics of a scene behind an image with no dynamics.
@JitendraMalikCV Exactly. We also should not confuse world models, as per your definition, with video prediction or video generation models. Understanding the dynamics of a system in order to control it is not the same as producing cute videos.
Reactions
1 postLB 
Lucas Beyer (bl16)@giffmana@JitendraMalikCV @ylecun Yes but that's a very different stance than "These capture the static semantics of the scene behind an image. No dynamics." :)
- likes: 3
- replies: 1
- bookmarks: 1
- reposts: 0