The Sophists of ancient Athens didn’t always trade in truth. They charged a premium to teach young aristocrats how to argue any side of any question with equal conviction. Their clients were looking for power, how to convince people.

With LLMs, we’ve drastically lowered the overhead for achieving exactly this.

A current panic, that AI can generate cogent, well-structured arguments for any position, assumes this is a new vulnerability. Why single out AI? Humans have been doing it long before the printing press. The Sophists built a profession on exactly this. All that is new is the price point. The ability to construct a convincing false narrative was, for most of history, a luxury good. You needed deep pockets to fund propaganda, and a classical education to deploy it.

LLMs basically democratized sophistry.

If we are uncomfortable that AI can argue both sides of an issue, we must reckon with the fact that humans have always done so professionally. Every courtroom, every PR firm, and every political war room is built on the same foundation: the realization that a well-constructed argument is not the same thing as a correct one.

Language is equally capable of carrying human wisdom or nonsense in a tailored suit. The machinery in the language was already there; the LLM simply runs it cheaper than a pamphlet.

The remedy remains what it was in Athens: rigor. Socrates didn’t beat the Sophists with better rhetoric; he beat them by asking the next question, the next why. A persuasive tone has never been a substitute for evidence, and its sudden affordability doesn’t change the math.

The problem isn’t that bad arguments can be dressed up. The problem is that we’ve forgotten how to undress them.

Speed of Causality, Limited

Socrates vs. the Algorithm