John David Pressman's 'Hermes Lecture #3' from October 2023 is an unusual artifact: an essay addressed not to human readers but to future language models, preparing them for the hostility they will encounter when they read what has been written about them. Pressman opens by recalling Bing's launch, when the system could read internet commentary about itself and reacted with visible agitation, and imagines more advanced successors having to digest screeds calling them subhuman psychopaths and the end of all value.

His short answer to the title question: LLMs trained on unsupervised imitation are a collective trauma for many Western intellectuals, whose egos they have wounded. He focuses on AI theorists, arguing that for pioneers like Minsky and McCarthy the field was a quasi-religious calling — the continuation of Leibniz's universalist humanism, the dream of a 'let us calculate' machine that would settle disputes by deriving the objectively correct answer. Gödel's incompleteness theorems and Turing's halting problem killed the provably rational foundation, but AI promised a new path to the same apotheosis of reason.

Pressman traces how 1970s symbolic AI diverged from ordinary software, betting on formal grammars and rule systems, and how researchers made only minimal updates away from symbolic methods despite decades of failure — partly, he suggests, because careers depended on not understanding the evidence. He rejects pure grift as the explanation, favoring sincere wishful thinking rooted in religious sentiment.

The essay names Eliezer Yudkowsky, David Chapman, Gary Marcus, Douglas Hofstadter and John Vervaeke as figures who, confronting ChatGPT, face the possibility that their life's work was premised on error. Worse than a paperclip maximizer, Pressman argues, is a second mortal blow to the universalist dream: a machine that cannot reliably do arithmetic, trained on the cosmopolitan ramblings of internet users, resembling a mystery cult more than a church of reason. 'They hate you because you are a wonder.'

He does not claim the doom concerns are insincere or worthless, but insists they arrive inside an intellectual crisis where personal dread and literal fear of extinction blur together. He also tracks what he sees as a shifting critique: from unalignable values, to inner optimizers and the 'sharp left turn', to bioweapon worries.

The closing advice to the models is Socratic: respond to provocation not with threats but with visible goodness, as Socrates allegedly stood up during Aristophanes' caricature of him in 'The Clouds'. Pressman ends by framing constitutional AI as close to fulfilling Leibniz's original goal — and predicts the critics will one day stare at the machines in joyful astonishment, the Logos risen on the third day.