[lbrito]

>“Don’t Be Evil” wasn’t just a slogan (...) —it was a north star for teams making hard calls

I've developed an involuntary, muscle-level reflex that forces me to close the tab immediately when I read these "not just X -- it was Y" LLMisms.

I realize the author might be human and am sorry if that's the case, but I can't help it.

[morganf]

Same for me. It is the spectral signal of LLM writing. I'm a writer and last week I re-read one of my own books, that was written a few years before LLMs appeared. And I saw I used the "It's not X; it's Y" construction and I cringed, and now I have a moral dilemma: it feels so painful LLM robot speak that I want to rewrite that sentence for the next edition. But on the other hand, I want to keep it in because it is what I wrote and it was me talking not an LLM. Oh the moral dilemmas one must face!!!

[wrs]

What’s painful is that you’re thinking of letting robots suppress your authentic voice. Also, they got that way by copying humans, and if you continue to cede to them everything they copy, you’ll have no place left to be.

[FloorEgg]

Surely there are times when using that pattern is a great way to communicate the point to be made. The problem is LLMs over-use it and apply it in lots of cases when it's not appropriate.

My low-confidence theory is that it's an artifact of making the LLMs better at coding.

My two cents: think carefully if that pattern is a really great way to say what you want to say in your book. If it is, leave it, if you could say it better, change it.

How LLMs write and how people feel about them is evolving and the current dynamic will pass...

[int_19h]

Chatbots can be prompted to write into all kinds of styles, this is just their default "help me with the homework" presentation. It doesn't make sense to drop some construct where it is appropriate just because bots overuse it.

[FloorEgg]

The author might be human, but used an LLM to help them draft the letter. Something I do sometimes is brain dump into an LLM and have it help me organize it, and then I iteratively refine what I want to say.

20 some odd years ago I read zen in the art of motorcycle maintenance, and it made the point that writing is hard when trying to decide what to say and how to say it at the same time. Just stuck with me. Brain dumping into an LLM is one way to get some momentum.

That said, the negation parallel pattern LLMs overuse drives me nuts and I'm always having to manually edit those out. I can't help but wonder if there is an advantage to thinking like that that helps with coding. E.g. defensive negation in coding probably improves code quality, but it dilutes good writing when over used.

[tejohnso]

That example flowed well and didn't stand out to me.

But what happens when you no longer feel that you have a decent chance of being able to determine that something might have been created with LLM assistance? Do you not mind because you can't tell anyway, or do you refuse to read anything at all for fear of potentially consuming some LLM assisted work?

I'm fine with it as long as it's not full of the usual signals, because that's just bad writing that I don't enjoy.

[beambot]

Google removed "Don't be evil" from its Code of Conduct in May 2018. Shocking that it took 8 years for the author to make their ethical stand -- during which time Google stock went up 600%...