Interestingly though for AI, this doesn’t necessarily mean we need a different model architecture. A single large multimodal transformer might be capable of a l

Interestingly though for AI, this doesn’t necessarily mean we need a different model architecture. A single large multimodal transformer might be capable of a lot that an LLM is not (besides the multimodality).