Professor Retweets Critique of Diffusion Models

Kosta Derpanis, Associate Professor at York University with ties to Samsung AI and the Vector Institute, retweeted a post by @KL_Div. The post states that diffusion and flow matching represent a standard generative AI approach yet remain slow, sample inefficient, and complicated. It asks whether such methods are truly required. The retweet forms part of visible AI-related conversation on the platform, with no further details or responses confirmed in the supplied evidence.