If you ground the LLM consistently in a context (for instance by constrained them with sensory data, a personality model and memories of all previous interactions), does your argument change?

→ View original post on X — @plinz,