Reading Schelling is fascinating because he has many of the same verbal tics as AIs. Maybe the single most AI-like writer of any famous + high-quality dead writer I could think of!
em-dashes, “genuinely,” “not X but Y,” regular structural parallelisms[1], Sentences like “We did not, and it cost us something.”
But he’s still clearly a much better writer than AI, for somewhat abstract and hard-to-precisely-define reasons.
One aspect is just structural and information density. Every page is guaranteed to teach you multiple things in addition to being in the furtherance of his overall point.
But it’s more than that. It’s not just teaching, exactly, it’s more the emphasis on finding true claims that are ex ante surprising. Consider this topic sentence of a paragaph: “But just as one cannot incur a genuine commitment by purely verbal means, one cannot get out of it with cheap words either.” [2] AIs (and indeed most human writing I see on Substack today) tend to do a much worse job of understanding and managing reader expectations, with the supporting paragraphs often not delivering much utility beyond the thesis statements.
Reading Schelling is fascinating because he has many of the same verbal tics as AIs. Maybe the single most AI-like writer of any famous + high-quality dead writer I could think of!
em-dashes, “genuinely,” “not X but Y,” regular structural parallelisms[1], Sentences like “We did not, and it cost us something.”
But he’s still clearly a much better writer than AI, for somewhat abstract and hard-to-precisely-define reasons.
One aspect is just structural and information density. Every page is guaranteed to teach you multiple things in addition to being in the furtherance of his overall point.
But it’s more than that. It’s not just teaching, exactly, it’s more the emphasis on finding true claims that are ex ante surprising. Consider this topic sentence of a paragaph: “But just as one cannot incur a genuine commitment by purely verbal means, one cannot get out of it with cheap words either.” [2] AIs (and indeed most human writing I see on Substack today) tend to do a much worse job of understanding and managing reader expectations, with the supporting paragraphs often not delivering much utility beyond the thesis statements.
That said, one thing he absolutely does not do is eyeball kicks.
This is surprising because somebody else writing about commitments might instead think of verbal commitments as highly asymmetric.