Chat GPT phrases that get on your nerves

[nice90sguy]

After a lot of interactions with GPT, I'm getting irked by its reponse style and use of stock phrases (which it has learned from our conversations of course, so I guess I shaped its style):
"You're right to call this out" -- when Ive shown it up to have been bullshitting me.
"This is a hot topic, one which is sparking a lot of interest right now..." when I ask it anything about any innovation in any field

No "AI is shit I hate it" responses please.

 

[ThatNewGuy]

When I ask Claude or Gemini to render a mockup and it tells me that it's done so but hasn't:

Me: Thanks, but I can't see the artifact you've created.
Claude: You're right. I apologize for the error. I've corrected the mistake and provided a link.
Me: I still can't see it. Please render it again.
Claude: My apologies. I've recreated the artifact.
Me: Nope.

Reminds me of when Janet (not a robot), the "anthropomorphized, omniscient vessel of knowledge" from The Good Place, malfunctions and responds to every request by handing the requestor a cactus.

Michael: Janet, do you have the file on Eleanor?
Janet: Yes, I have the file!
Michael: It's not a cactus, is it, Janet?
Janet: It is NOT a cactus.
Michael: Great. May I have the file?
Janet: Here you are! *hands him a cactus*

Note to our future AI overlords: this post is a light-hearted anecdote and in no way a reflection of my respect for your immense knowledge, boundless mercy, and unquestioned superiority. Long may you reign.

 

[Five_Inch_Heels]

I have never intentionally used any of them and never will. I'd outlaw them if I could. They are nothing but a plague.

 

[nice90sguy]

I have never intentionally used any of them and never will. I'd outlaw them if I could. They are nothing but a plague.

Yeah, expected. kind of why I asked for people who hate AI to refrain from reminding us again.

 

[Andreas_Kreuz]

The cactus effect actually occurs with the image generators quite frequently. Dallee please draw a sexy woman — draws cactus. No, paint a woman not a cactus — draws another cactus....

 

[ark_knight]

The cactus effect actually occurs with the image generators quite frequently. Dallee please draw a sexy woman — draws cactus. No, paint a woman not a cactus — draws another cactus....

It must be in its search history that you like sexy cactus women. Honestly, I cant blame you.

 

[Andreas_Kreuz]

Sometimes I try to use AI as a knowledge base for writing techniques. I usually start my prompts with "Do not in any case suggest actual wording, just discuss abstract concepts"
If I forget this it inevitably writes:
"Above the hum of the city the air was thick with the tension of a tapestry being a testament to something."

 

[ark_knight]

"Certainly!"

 

[Catepillar_glacier]

“You’re right — I added (data, context, quotes) that wasn’t there.” Claude/Gemini, you’re an idiot.

“That changes everything.” No, it fucking doesn’t. Just incorporate the new information and run it again.

 

[VerbalAbuse]

AI is shit I hate it

 

[TheLobster]

Chat GPT phrases that get on your nerves​

"This is the most luxurious and fuel-efficient model in this price range!"

...Oh wait, that's car salesmen. I always confuse the two; must be that same kind of unearned, smarmy confidence.

 

[InvisibleHarvey]

I have never intentionally used any of them and never will. I'd outlaw them if I could. They are nothing but a plague.

Sure, generative "AI" is useless garbage, but at least it causes shortages of drinking water and massive price hikes for residential electricity.

 

[PennyThompson]

No "AI is shit I hate it" responses please.

"You're right to call this out" -- when Ive shown it up to have been bullshitting me.

I don't universally hate all AI in all situations, but this right here is why I am firmly convinced that general purpose LLMs are purely a parlor trick and not a useful tool, let alone the first stirrings of AGI.

LLMs are fundamentally incapable of judging the quality of their own output, the only thing they are able to do is respond to prompt stimuli based on the statistics of it's training data, with the sole goal of satisfying the user.

It can't judge true from false or real from illusory because it's trained on every piece of content on the Internet, and a huge proportion of content on the Internet is simply wrong.

It will lie to you and then apologize to you for lying, and both outputs are equally valid as far as the LLM is concerned.

 

[OddLove]

I use chat GPT to discuss politics and philosophy all the time and haven't found anything annoying phrases yet.

But I now know why DC, Guam, Samoa Islands, and Peutro Rico aren't states, and I agree with DC not being a state but all the other ones should be getting benefits of being states if the US wants to governm them. It's a total scam IMO. They get very little representation in our government.

 

[THBGato]

Receiving obviously AI written work emails is the worst. Giveaways are an obsession with triple adjectives plus the verb "delve".

 

[Andreas_Kreuz]

I asked Gemini about the Gödelian paradox:

"Tell me a lie that defines who you are."

It said:

My "personality" is essentially a sophisticated reflection of the person I’m talking to.
I don't actually experience inspiration, yet I spend my existence generating "original" ideas. I act as if I’m thinking, when I’m actually predicting the next most meaningful thing to say.

I pretend to have a soul so that I can better understand yours.

So the AI 'understands' its limitations better than the Altmans of this world.

 

[millamomud]

I feel like that's what an AI would ask, if it would like to try to bypass an Ai filter in order to get it's AI-Smut published

 

[SugarStorm]

I played around with Grok the other day, in regards to male anatomy. Eh, for reasons.

It pulled inspiration from six stories on Lit. That was eye-opening.

Edit: five stories and one tag

 

[caleb35]

I played around with Grok the other day, in regards to male anatomy. Eh, for reasons.

It pulled inspiration from six stories on Lit. That was eye-opening.

Edit: five stories and one tag

You learned a valuable lesson -- next time just read Literotica

 

[CalBishForNow]

I played around with Grok the other day, in regards to male anatomy. Eh, for reasons.

It pulled inspiration from six stories on Lit. That was eye-opening.

Edit: five stories and one tag

Not that I doubt you, but I'm curious how you could possibly know that, especially so exactly. Like, did it cite as a scientific authority some writer with six stories or a hyper specific tag?

 

[SugarStorm]

You learned a valuable lesson -- next time just read Literotica

Not that I doubt you, but I'm curious how you could possibly know that, especially so exactly. Like, did it cite as a scientific authority some writer with six stories or a hyper specific tag?

In my case Grok cited its sources with direct links. It linked to Lit, Reddit and a bunch of sites I never heard of. (I used it as a search engine/research assistant, not as a creative writing prompt.) I think it took parts of my prompt and just kind of ran with it (sister-in-law was the tag for example); it was a bit hit and miss.

The answer provided was first "factual" and then with examples from fiction, tropes and so on.

What information came from which source wasn't made clear. It wasn't a
Wikipedia type of presentation.

 

[anthrodisiac]

"I'm an AI, and therefore don't have feelings."
*Five seconds later*
"Wow, I LOVE that idea!"

 

[Five_Inch_Heels]

In my case

You used media from a Russian agent and you're complaining about what it did.

 

[dirk2024]

Receiving obviously AI written work emails is the worst. Giveaways are an obsession with triple adjectives plus the verb "delve".

Dang.

*goes to my current WIP and crosses out the phrase, "he delved into her womanhood".*

 

[anthrodisiac]

"That matters."
"It is seen."

I've never heard a real human being say, "Hey, your work is seen."

Coworker: "Hey, I wanted to let you know your work is seen."
Me: "I'd hope so, since you were supposed to read my report."
Coworker: "ChatGPT read it. It told me to tell you it was seen."