I think there were many unthinkable things achieved in the past that one can hope that we can find a way to train LLM-style “AI” a lot faster with way less recourses and thus achieve self trained assistants, that interact exactly as one expect from their personal helper.