11 thoughts on “To the three different people who sent me chatbot emails today”

  1. Tangentially relevant: I read today about how the court system is struggling to cope with the onslaught of AI-generated submissions. The problem is that resources to deal with submitted texts have always been, and always will be, limited. Lamentably, it goes against social (and often legal) norms to submit lists of bullet points in court filings, submissions to social services agencies, academic or term papers, or even emails to random professors on the internet. It would save us all a lot of reading time — and I’m not being sarcastic here.

    • Interesting, I’ve always considered bullet points the best option for communicating with representatives of officialdom for exactly those reasons.
      I’m also ambivalent about the AI-for-officialese pathway. My experience is that LLMs have massively lowered the barrier for submitting official paperwork, which seems a potentially good thing for democracy and access: presumably those lacking the skills, knowledge and time have been excluded from numerous bureaucratic processes in the past and LLMs are removing some of that barrier for some people (I’m not Panglossian about this).

      • Most readers are probably familiar with Tufte’s “Cognitive Style of Powerpoint,” but for those who are not, here are a few of his comments on bullet points:

        “For the naive, bullet lists may create the appearance of hard-headed organized thought. But in the reality of day-to-day practice, the PP cognitive style is faux-analytical, with a bias towards promoting effects without causes.”

        “As analysis becomes more causal, multivariate, comparative, evidence-based, and resolution-intense, the more damaging the bullet list becomes.”

        And, quoting Feynman, “Then we learned about “bullets” – little black circles in front of phrases that were supposed to summarize things. There was one after another of these little goddamn bullets in our briefing books and on slides.”

        Much more where that came from.

  2. Not that I’m endorsing folk sending Andrew random chatbot output, but, to play Devil’s advocate, the same prompts can yield quite different responses, even for the same model, depending on personalisation and account chat history (and probably other things I’m unaware of).

Leave a Reply

Your email address will not be published. Required fields are marked *