My theory on why the default LLM writing is tiring: LLMs are very optimized for software engineering and this leaks into writing: every sentence needs to be correct, hedged, and defensible. 1/
It would have been funny if it were not sad. Apparently many academics do not care about progress and discovery of truth but care more about the preservation of the profession. If the profession cannot keep up, we should turn Amish.
It's worth being explicit about this.
AI written submissions will break peer review as it currently exists. They also undermine one of the key functions of academic publishing, ie to identify talented researchers.
So we have to create some kind of bright line distinction that
If “writing is thinking” then job interviews would have been written exams. There is a reason the vast majority of job interviews require people to talk, not to take written exams.
Papers should ship with data, code, and validated references. Do that and the AI-authorship panic disappears. Fake results cannot survive verification. Real results do not care who typed them.
[Pangram score: 100% AI]
Norse Atlantic delayed my family’s flight overnight in Oslo. To soften the blow, they offered a $25 meal voucher and ignored all my emails.
So I used ChatGPT-Pro to navigate Norwegian civil procedure and sue them from my couch in NY.
This week, they wired me $4,760.36. 🧵
