feels really bad to be someone who writes like chatgpt

Submitted by twovests in just_post

I stumbled upon old notes of mine from 2019. Wonderfully formatted markdown, with # Headers and ## Subheaders and - **Boldly formatted lists** introducing concepts and - **Sublists** replete with the same formatting. Even sections labelled, e.g. Principles behind the 12-tone scale:

It reeks of AI-generation, but this is a tenderly crafted document I'm really happy I produced. It's full of personal anecdata and math.

It's pretty crushing to see writing I generated get detected as "likely AI" by these detectors. It makes me kind of sick.

10

Comments

You must log in or register to comment.

emma wrote

all written communication is just a big game of secret hitler now.

oh you used bold text? that's sus, you couldn't possibly have done that on your own <-- an actual accusation i've seen lobbed at someone who made the mistake of showing off their project to people on reddit.

i absolutely don't trust people who copy and paste from chatgpt, so i can see where the distrust and suspicion is coming from, but that's not sufficient justification for throwing hurtful accusations around. especially when your evidence is 'they know markdown' and 'some ai crap told me you posted ai crap'.

7

twovests OP wrote

Perfectly articulated, yeah. I also am looking out for LLM people while also hyperaware that my posts look like I didn't even bother to prompt "respond in a casual, whimsical style befitting Reddit dot com"

5

flabberghaster wrote

I think the thing is, chat GPT is aping all of us. Its style is a mishmash of the most common styles. I don't change how I write to sound less AI, I just write like I always did.

In the immortal words of office space guy, "why should I change? [it]'s the one who sucks."

5

twovests OP wrote

But it's also a combination of scraped text, and a style guide trained into it. It just so happens that OpenAI and I both like to communicate in the form of a README.md roughly following dyslexia style guides and whatever the SEC has cookin' for their Terms of Services.

4

I_got_killed_one_time wrote

you still have hope as long as you dont use an em dash

4

twovests OP wrote

That's a great observation I_got_killed_one_time– so long as I am judicious in my choices in punctuation, I can

5

hollyhoppet wrote

honestly you don't have the sycophancy problem which imo is chatgpt's worst feature

4

twovests OP wrote

I can barely constrain how much I love and respect people on this site. I try to keep it toned down and be a little mean to compensate

4

alenky wrote

It would be more fair to compare your work to the prompts ai bros write. Chatgpt text may be slop but it’s still beyond the abilities of AI bros to even match the slop.

2