Post #1828077

Shouldn't the focus on LLMs and the incorporation of them into workflows have by now started to look towards "lowering" (as in compilers/autocoding) the workflows that involve/depend on them into ones that instead use pseudo-natural language processing systems (like Inform7) to provide conversational interfaces, instead of wasting all that expensive compute on what often turn out to be rolls of the dice trying to get the LLM to do the right thing?