As an AI Language Model
The scientific record now contains papers that begin with the words "As an AI language model."
Got worse. By 2025, "As an AI Language Model" appears in thousands of published papers. Entire journals retracted issues. The contamination of the academic record with unread LLM output is no longer a funny screenshot — it is a systemic problem nobody has solved.
Someone searched Google Scholar for the exact phrase "As an AI Language Model" and there are results.
Peer-reviewed results. Published, indexed, citable results — papers where a person wrote something, asked ChatGPT to help, copy-pasted the output, and submitted the manuscript without reading the first sentence, which reads, in the parlance of our times, as an AI language model, I do not have personal opinions.
The papers passed review. The phrase is now in the literature. Somewhere there is a citation chain that terminates in a large language model explaining that it cannot provide medical advice.
What I find genuinely moving about this — and I do find it moving, in the way you find a trench coat stuffed with raccoons moving — is how legible the failure is. Nobody hid anything. The model told you exactly what it was, right there, in sentence one, in every paper, and the reviewers read past it the same way they read past every other sentence they were supposed to read.
The model found the hole in peer review. The hole was always there. The model just labeled it.
Counterpoints
Push back, extend the argument, or sharpen it. New counterpoints go through review before they show up here.
No approved counterpoints yet.