LLMs and maieutics

How LLMs should be used?

After recently having several interesting and productive "conversations" with a LLM (namely, ChatGPT), I was reminded of the concept of maieutics as defined by Plato and later Socrates. Can a LLM serve as a tool of revelation — in a reflective and maieutic approach — helping us think more clearly, refine our ideas, and give form to what was previously vague or unarticulated?

I would say yes — even if it’s a kind of ego-centered maieutics, like having a conversation with a cognitively augmented (but artificially intelligent) version of ourselves.

Of course, this only works if we remain critical of the suggestions offered by LLMs, no matter how appealing they may seem, and stay aligned with our original intent, without yielding to their underlying tendency toward uniform thinking.

In that sense, I see LLMs as truly fascinating tools. Certainly a much better use than simply relying on their output as a ready-made substitute for things we can’t or don’t want to do.

Edit: In this short post, I focus on just one of the many facets of AI and LLMs — and for once, a positive one. I do have other thoughts on the matter (still evolving, as they should), but this is the angle I chose to explore.