Théo Louvel

For a while now, I've been thinking about how LLMs push us to question our questions, thus improving our thinking. We used to search the web by stringing together a query composed of keywords (what I like to call hashtag thinking) to get a list of sorted results. We sort of understood the role of context, but were only able to express it through keywords.

With LLMs though, articulating ideas together matter just as much as the ideas themselves. Framing the problem influences the outcome. This open form allows us to include all relevant parts of a problem, not just a list of related keywords, and express the relationships between them.

I can't help but notice how the conversational nature of those tools pushes us to refine and rephrase further, propelling us into a virtuous feedback loop that ends up making us better problem solvers at the end of the day... I have this romantic belief that a well-defined problem is indistinguishable from the solution, as having no more questions to ask is equivalent to having turned over every stone.

A third way LLMs enhance our thinking is by distancing ourselves from our ideas, enabling us to become more of a critic, curating the good ones and throwing out bad apples. As Leslie Lamport puts it, "writing is nature's way of telling us how lousy our thinking is." It's hard to throw away bad ideas when you've tied them to your identity. Externalizing them helps untie this Gordian knot.

Those mental tools aren't exactly news, though, and neither is the need to improve our thinking by questioning our questions and beliefs altogether: The Socratic method has been around for a while. But when was the last time we came up with a tool enabling so many people to simultaneously work on asking better questions?

In my view, those three factors come together to produce somewhat of a Lollapalooza effect (the convergence of multiple factors that multiply each other to produce a result much greater than any of them would have produced individually). I can't help but think this will affect the future in many ways, and I'm curious to see what society will look like 20 years from now if we keep working on those systems.

“We become what we behold. We shape our tools, and thereafter our tools shape us.” ― Marshall McLuhan

What will you ask?

PS: There's a gigantic blindspot in this analysis. Some people will focus more on the outcome and taking it for the absolute truth, leading to compounding ignorance, stupidity and surface-level thinking. The remaining question is what will the majority do, and how we can shape those tools to nudge users in the right direction.