Yes, I think you are right. I didn't get a chance to get into this with OpenAI, but it seems connected to the idea of multimodal models.

→ View original post on X — @willknight,