Any spoiler leads to Yann Lecun’s push against current architectures?

So, Lecun announced the public that he believed that the LLMs would never heal the gunshot because, in fact, the way to token each step toward the wrong ways.

Instead, he offers the idea that we should have something to be an ‘energy minimization’ architecture; As I understand it, it has a concept of ‘strength’ in a whole response, and training will try and reduce that.

Which is to say, I don’t fully understand it. That said, I was shocked to hear what ML researchers think about Lecun’s acquisition, and if any engineering was made around it. I didn’t find much after the release of Ijepa from his group.