Hacker News new | ask | show | jobs
by drc500free 14 days ago
It's easy to focus in on particular linguistic tics, which will probably get smoothed away in future training. The underlying issue is that the LLM is trying to ape meaningful writing - which takes the reader from A to a surprising Z - without generally basing it on a meaningful insight.

Most of the common tells stem from that desire to signal the gap between what it's writing about now and how you previously thought things worked. And they really fall flat when it's writing about some milquetoast truism with all the edges sanded off.

Rather than have "Z" be self-evidently interesting, the LLM need to tell us that it's not "A". Except no one thought anything was "A" in the first place, and the "Z" is barely a "B" let alone a "Z".

Or things are "quietly X," implying that there is some secret knowledge that other people have.

Barring that, the LLM will signpost arguments, telling you how interesting things are. "This is the critical part..." before launching into another banal non-observation. Why have only "Section Header" when you can have "Section Header (another idea I had) - omg I just have so much to say about this". Maybe a list of weak ideas, presented in multiple ways. Bulleted with emojis. As a rule of three. As choppy sentences in a paragraph.

All of these follow the same pattern - trying to follow the FORM of having something interesting to say that differs from consensus understanding, without having the intelligence or boldness to actually midwife a new concept into the world.

3 comments

Yes, this is the kind of thing I've been hoping more people would notice.

I think another tell with a long lifespan is that it will try to write about a thing it experienced, but having no real knowledge of the qualia, it'll end up only referring to it (not uncommonly using "that") and leaving all the substance as an exercise for the reader

It strikes me as an Oder of operations problem.

If the system prompt is “flesh out template y with thought x” the form drives the generation, it feels compelled to use the whole template.

Of the system prompt is “refine thought x and then format it with the appropriate parts of template y” it becomes a simple transformation and formatting.

A lot of current gen llm pain appears to be being in the early days of understanding the nuance of system prompts.

Totally agree, I think it can get there.

If you present the idea as "an idea" rather than "my idea", it's pretty good at challenging, solidifying, and refining it. It still leans on the same forms in the writing part, but I think it can be tuned to not insist upon the idea's brilliance so breathlessly.

AI output reads like homepage marketing content (e.g. the text that fades in when you scroll down an apple product page) expanded to fill some context window size (the paragraphs tend to be about the same size).

What you get is vacuous, choppy, wordy, and hyperbolic. I have found that adding secondary passes for tone and style improve the readability dramatically.

Marketing content without a reader persona or a writer persona, or rather, without the backing of a real consciousness with agency and life experience. I wonder if this might be part of why there are waves of different tells (in addition to the model updates): People pick up on an uncanniness in the writing but aren't fully aware of it or its nature, so they seek indicators and find what's there to find at the time