I think much like with a basic Turing machine definition compute is possible on a variety of substrates that some kind of intelligence can be created with a whole class of implementations, transformers included. Indeed the video and image input of LLMs is one of the most exiting use cases.
No comments yet.