1 Matching Annotations
  1. Last 7 days
    1. VL-JEPA: Joint Embedding Predictive Architecture for Vision-language

      for - from - Youtube - LLMs are dead! - https://hyp.is/iRe3QuxFEfCGYzMyXPsieQ/www.youtube.com/watch?v=BrNn1TcNK5s

      SRG comment - Yann Lecun's thesis is that current LLMs only mimick half of humans cognitive capacities - the purely linguistic, and this is quite an abstraction. Human reasoning depends on the other important part, embodied experience. - A more accurate AI would take into account the embodied aspects of human learning that are the necessary context for linguistic affordance to develop