‘Tis the season to receive a wave of applications for the Amgen undergrad scholars programme. Ever since ChatGPT, most of the ones I receive are clones of each other, drawing from text in my academic institutional webpages. Clones. Each applicant of course doesn’t know that the other applicants are sending near-identical letters; they may even feel they’ve given it their most, making a smooth, beautiful letter.

Makes the choice all that much easier for me: if you write an original letter, even if in half-broken English, you’ll have a better chance. Nobody likes slop. And more: if you use AI instead of using your brain, frankly, there’s no need for you to join my lab, or any lab. Don’t think of us as a future line in your CV. Join in to grow as a scientist instead.

#academia

@albertcardona Very tangential, but I attended a wedding recently where one of the speakers withdrew his speech, because the LLM had given him the same speech the sister of the bride had made.

@rustedivan

Why does everyone expect LLMs to yield original answers? Just because the prompts and answers happen in one's computer, in an environment associated with privacy? While not deterministic, LLMs are quite close to it. A stochastic parrot indeed.