Transformers are probably AGI, at a mathematical level. But that’s just an architecture (likely a wildly inefficient one), not an implementation of that architecture (modern LLMs).