Comment by MoonGhost

He didn't mention multi-modal models. Probably because they don't fit in the oversimplified picture.