Hacker News new | past | comments | ask | show | jobs | submit login

Why care whether something is AI slop or human slop? It’s not worth reading in either case.

The arguments presented here look suspiciously like the arguments scribe classes of old used against the masses learning to read and write.

Seems like we’ve gotten to the point where sloppy writers are worse than LLMs and assume that all “meticulous” writers are LLMs. The only convincing “tell” I have ever heard tell of: characters such as smart quotes, but even those can just be a result of writing in a non-standard or “fancy” editor first. I’ve even seen people say that em dashes are indicative, I guess those people neither care about good writing nor know that em dashes are as easy as option + shift + hyphen on a Mac.




Because AI slop can be generated in massive quantities that dwarf anything prior in history. Sifting through these bales of hay is honestly exhausting.


> Why care whether something is AI slop or human slop? It’s not worth reading in either case.

The problem is human slop is typically easy to detect by grammar and other clues.

AI slop is often confidently incorrect.


> human slop is typically easy to detect by grammar and other clues

I'm not sure this is true. There have been a lot of times where I see a very well made video or article about some interesting topic but when I go to the comments I end up finding some corrections or realizing that the entire premise of the content was poorly researched yet well put together.


Most of the human-made slop you'll see is going to be, at least on its surface, high quality and well produced since that's what goes viral and gets shared. To that end, I agree with you.

It is worth noting though that the other 99% of human-made slop doesn't make it to you since it never gets popular, hence why hard to filter human-made slop can seem over-represented just through survivorship bias.


Hence the word "typically".

I am noticing it in work emails a lot now - poor performers cutting and pasting chatgpt content with no validation.


Nit: em dashes also appear when you type two dashes in a row in word or mostly any other MS product. That’s a terrible heuristic.


I am a proof of that.

I write in word to correct my text when I use my pc. Additionally, it's a better editor than the one supplied for commenting...


The new Azerty has a bunch of typographically correct punctuation too :

https://norme-azerty.fr/en/


> I’ve even seen people say that em dashes are indicative, I guess those people neither care about good writing nor know that em dashes are as easy as option + shift + hyphen on a Mac.

They are virtually indistinguishable from regular dashes (unless you're specifically looking for them), and contribute nothing of significant value to the text itself. They were only ever a marker of "this is either professionally edited, or written by a pedant".


> Why care whether something is AI slop or human slop? It’s not worth reading in either case.

That's not always true, and this is one of the fundamental points of human communication that all the people pushing for AI as a comms tool miss.

The act of human communication is highly dependent on the social relationships between humans. My neighbor might be incapable of producing any writing that isn't slop, but it's still worth reading and interpreting because it might convey some important beliefs that alter my relationship with my neighbor.

The problem is, if my neighbor doesn't write anything other than a one sentence prompt, doesn't critically examine the output before giving to me, it violates one of the basic purposes of human to human communication—it is effectively disingenuous communication. It flies in the face of those key assumptions of rational conversation outlined by Habermas.


I'm pretty sure that anyone saying "emdashes are a tell" doesn't mean the literal character, but also the double-hyphen or even single hyphen people often use in its place.


Human slop is still written by humans, implying there was effort and labour. Not sure how to put it in words properly, just "the vibes" are different.


<Compose>--. on Linux.


That's an en-dash, not a em-dash. – vs —


Oops. Then it must be <Compose>---


Hyper + Shift + -




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: