don't tell me, he only needs few billions to continue the research     See Less

Great success! I was spending my last month trying to understand the direction of AI in general and I came to realize the principles of JEPA before knowing it exists as a project. Which mean     See More rrived at the same conclusions with Yann about the architecture but from a different road and a different philosophy. That is validation to me!    See Less

This seems to be about understanding and i'm looking forward for the reasoning based on understanding. Interesting times.     See Less

This will kill prompt engineering     See Less

I think this model is much closer to true AI compared to current LLMs, which often generate outputs without understanding their correctness. This approach also appears significantly more eff     See More     See Less

The woman at the end seemed to be reading off notes and gave the impression of not deeply understanding the ideas     See Less

yeah no progress to see here, its gon be behind paywall or worse, in some CEO type desk drawer, never to be seen by free public.. if it was released as open source, then this would mean that     See More rices could go down, because it does not need the memory thing that much..    See Less

??
1) A paper is not a model
2) Publishing instructions is not gonna help him "start his own thing"

thanks for giving out free stuff for everyone i guess.     See Less

Sure, LLMs are not the end of AI evolution. There will come much more and powerfull with dramatic less energy consumption     See Less

I do like the discussion at the end there. Surely it's not necessary to train the AI to drill down on the details of what makes a cucumber, it just needs to be able to detect what a cucu     See More like    See Less