I don't like LLMs

Fowler feels both fascination and fear about AI. But when he talks to LLMs directly, the reaction that dominates is dislike of their voice: a grating, uncanny-valley imitation of human conversation, confident even when it’s making things up.

He separates the discomfort, a personal signal, from a technical objection. Jessica Kerr’s case still holds for him: ignoring LLMs is irresponsible, given the gains in usefulness and speed.

Fichamento

#

• Fowler holds fascination with AI’s potential (productivity gains, even cures) alongside fear of what it could enable, from infrastructure takeover to designed bioweapons.
• The feeling that dominates comes from direct use: dislike. LLMs talk in an uncanny valley of human conversation, and confabulate with the same confidence they show when they’re right.
• Avoiding them anyway strikes him as irresponsible; he agrees with Jessica Kerr that they’re useful, thorough, fast.
• He traces the dislike mainly to the Silicon Valley tech-bro culture that shaped these models’ values.
• Anthropomorphizing AI agents is the trap he warns against (Morgan’s canon applies almost directly here): their behavior isn’t explicitly programmed, but it’s nurtured by the people and corporations that build them.
• A personal rule closes the piece: he avoids people he doesn’t like or trust, and extends that same rule to LLMs, treating the discomfort as a signal worth acting on.