> I don't foresee AGI arising out training bigger LLMs
I agree that AGI will involve tool usage but not only involving domain specific AI models.
But lets try to find the discriminating point in the discussion - do you believe AGI will necessarily involve training bigger LLM's or not?
I believe they are necessary. WBU?
No, I don't think LLMs are necessary for AGI at all. I think there are multiple paths to AGI, some of which involve LLMs and some which don't.