Oh, it did not feel AI but Pangram does say AI, high confidence. Good catch.
(I'd trust an ML algorithm that only has to classify in two boxes, thus easy to evaluate, over my own gut feeling).
(I'd trust an ML algorithm that only has to classify in two boxes, thus easy to evaluate, over my own gut feeling).
---
Send an e-mail + pick up the phone. Mismatched analog. First one is outgoing, second is incoming. Yes, I know you can pick up "to make a call". In which case, then we have the classic LLM pattern of eliding details that matter to make it sound "smooth" (like blended cardboard smooth, glug glug).
someone senior enough - Way too vague. Most people would add details here that are unique to the actual human being that they spoke with. But Claude can't do that, so we get this.
a hard question - Same thing as the previous one. Which question? Why is there literally no detail or actual flavor here?
- Real - Claude looooves punchy, empty emphasis words like this.
... , the kind you ... - Tacked on extra phrase that should be a separate sentence. Claude does this all the time. We aren't writing a shitting noir novel, Claude.
-a decision go sideways in a boardroom - vague shadow of an anecdote, of just the right length. The "a" in front of "boardroom" is a blinking red light for me, can't fully explain why.
Not what the textbook says. - Too short, too vague, with the rhythmic punchy feel. (Damn... I've been infected with the triplet pattern myself).
textbook, 5 studies, 30 years of experience - Pattern of 3...
- that a search engine couldn't - Oh man... this type of phrasing is so triggering for me, heh. I've been busting my ass trying to sanitize Claude's output so it haven't doesn't these overly cute, inverted constructions.
I often leave out details and specifics or make things purposely vague because I am writing about real people.
Thanks for the the correction-- often when people say they weren't using AI they are very clearly lying, so it's hard to get useful corrections on this point so I appreciate your candor.
1) vibe red teaming (having Codex RE the data flow of various extensions, comparing the attestation with the ToS, seeing if I think it's a potential FTC violation) which is just for fun and I do nothing with,
2) feeding tons of documentation into NotebookLM and asking if everything in a short document is attributable (often when it says it's not, it is, but you have to kind of read between the lines). I think this is a good use of AI if rigorously fact-checked.
3) feeding 20-30 job descriptions + cover letters, asking for outlines for new cover letters. For this one I think it may be faster to go back to creating templates because I have to entirely rewrite things anyway. (It does add lies! And conflates things! V bad for a cover letter!) But I think use of machines to write letters being read by machines is pretty justifiable.
4) I've been experimenting with Claude editing. I think what bugs me about people assuming everything is LLM-written is that it's rather dismissive and difficult to tell the difference between "a human spent X hours on this and made some stylistic tweets" vs "a machine wrote this in response to a prompt."
I do find myself immediately scrolling away too, though. A friend had a LI post for a course he's teaching and I could tell it was AI because it's not how he writes. That said, the post performed phenomenally well, which is probably more important for selling a course than my opinion is.
I'm also not confident in my ability to distinguish between 1) something an AI write from scratch, 2) something he may have iterated with an AI, feeding it many different versions of a post, going back and forth, etc. I think both would show up as AI in the extensions I've seen, but I also think they're very different.
My thought process on it is complicated. I think I'm more disappointed than anything. It made me sad because his human writing is so unique, and I'd hate to see it all outsourced. But I also don't have the same "you didn't even bother to write this so why should I read it" reaction so many do, given that I don't know how much time he spent on it or the reasoning.
Mostly I worry that if the whole of human knowledge becomes AI slop, and AI is then being fed that slop, that the writing would get even worse.
Re: fact-checking, I am far too paranoid/obsessive/whatever about accuracy to not review changes! Even NotebookLM is weirdly drawn to the first document you feed it, so if you have a second contradictory document, it seems to want to tell you why it's wrong. To be fair, I know people like this, too. The first version of a story they hear is what they will forever believe, even if every other version contradicts it.
Increasingly common these days... I've caught myself more than a few times.
I also recently reread a blog post I wrote a decade ago, and it had the "It's not X, it's Y" pattern over and over again.