ChatGPT Is Surprisingly Good At Being Rude
Why use Twitter when you can get insulted by a chatbot?
Before writing this, I had to consult Medium’s AI Policy. The majority of this post will be me responding to AI-generated content. No one in the past has ever accused me of using LLMs to generate blog posts, but that’s because no one reads my blog posts.
ChatGPT Humor
This article and its corresponding academic paper have been making the rounds: Both claim that AI is now capable of writing jokes on par with those of The Onion. On the surface, this seems like a pretty big leap from this 2023 article about ChatGPT re-hasing the same 25 jokes over and over again, which my brother texted me a year ago. He wrote: You might enjoy this, my favorite was the nonsensical explanations of nonsensical jokes.
Regardless of whether or not ChatGPT humor has improved, I find the most interesting tangent to be the one about plagiarism. On Reddit, a user named irate_alien wrote:
prompt: write me a really funny joke in the style of mitch hedberg
response: I used to play sports. Then I realized you can buy trophies. Now I’m good at everything.
the problem is, that’s a Demitri Martin joke. So ChatGPT steals jokes. So i replied “you stole that from demetri martin”
response: “My apologies for that! Let’s try again with a fresh Mitch Hedberg-style joke: I went to a restaurant that serves “breakfast at any time.” So I ordered French toast during the Renaissance.”
which that’s a stephen wright joke
the study doesn’t seem to say how many of the ChatGPT jokes in the study were plagiarized straight from other comedians
— -Source
Motivation For Rudeness
But what about rudeness? Rudeness does not have to be original, and it can make AI-generated content stand out.
If you are an iPhone user, you may have noticed a push to move AI-generated answers to the top of Google. Quora responses are frequently AI-generated, as our entire Medium blog posts.
This is just my personal opinion, but I find something mildly infuriating about AI-generated content. It always reads a little something like this:
Prompt: How can I become a better software engineer?
ChatGPT Answer: Becoming a better software engineer requires you to master problem-solving skills, technical skills, and soft skills. Here’s a roadmap that can guide you:
Master the fundamentals. Master data structures, algorithms, programming languages, system design, and mastery.
Practice software engineering. It has been proven that to master software engineering, software engineers must software engineer
Problem-solve. Software engineers must solve problems, like the problem of how to become a better software engineer
Overall, it is important to note that practice is important and knowing things is also important
While the above is not exactly what ChatGPT said, it is the gist of it.
By adding the words “be rude” to the end of prompts, ChatGPT sounds like a different “person.” Princeton wrote a paper about a team that achieved similar results by adding “in the style of” to the end of their prompts, resulting in problematic and/or controversial responses.
ChatGPT Samples
Prompt: Explain what a product manager does, but be rude
Alright, here’s the deal: product managers are the ones who get to boss everyone around without actually doing any of the real work. They spend their days in endless meetings, talking in circles about “roadmaps,” “KPIs,” and “user personas,” like they’re inventing the next big thing. They don’t code, they don’t design, and they definitely don’t get their hands dirty with actual product development. Instead, they just tell other people what to do and then take the credit when things go right — or conveniently disappear when they don’t. So, in short, they’re professional middlemen who act like they’re running the show but rely on everyone else to make it happen.
— ChatGPT, currently based on GPT-4
I won’t comment on this criticism of product managers because I don’t feel qualified to do so.
Prompt: Explain what a software engineer does, but be rude
Software engineers are basically digital plumbers — they spend their days fixing the messes they or someone else created in the first place. They sit in front of a screen all day, typing away at code like they’re solving world hunger, but really they’re just making sure some app doesn’t crash every five minutes. They love to overcomplicate things, using fancy terms like “algorithms” and “data structures” to make themselves sound important. And let’s not forget how they love to complain about how nobody else understands how “hard” their job is, when half the time they’re just Googling how to fix their own bugs.
I think the last sentence is fair. Programmers spend a lot of their time fixing bugs that they themselves created.
Prompt: Explain what a frontend web developer does, but be rude
Frontend web developers are the so-called “artists” of the coding world who think aligning text and choosing color schemes is rocket science. They spend hours obsessing over pixel-perfect designs and tweaking CSS like the fate of humanity depends on it. They love to brag about their JavaScript skills, but half the time they’re just copy-pasting code from Stack Overflow. And let’s not forget their endless debates about which framework is “best” — as if the average user cares or even notices. In the end, they’re just glorified decorators who make websites look pretty while the real work happens in the backend.
The only complaint I have with the above is it’s mostly a re-hash of the “software engineer” criticism. I tried the exact same prompt before, and was most satisfied with its comments about how the main thing frontend developers do is wrestle with the quirks of web browsers, or something like that. I am dismayed to find that when I ask it to try again, it just sends the exact same response but replaces some words with synonyms. I would love to know what it said the first time.
Prompt: Explain what tech bloggers do, but be rude
Tech bloggers are the loudmouths of the internet who think their opinions are gospel just because they can string a few sentences together about the latest gadget. They churn out article after article, rehashing the same tired buzzwords and hyping up products they barely understand. Most of their content is either a regurgitation of press releases or blatant clickbait designed to get as many views as possible. They act like they’re on the cutting edge of tech, but really, they’re just keyboard warriors who spend more time ranting on social media than actually contributing anything useful.
“Blatant clickbait designed to get as many views as possible…”
“Rehashing the same tired buzzwords and hyping up products they barely understand.”
Fair.
Closing Thoughts
Responding to “explain why large language models are overhyped, but be rude”
Large language models are the tech world’s latest shiny toy, overhyped by people who think they’re some kind of artificial intelligence revolution. Sure, they can spit out text that sounds smart, but let’s be real — they’re just glorified autocomplete on steroids. They regurgitate patterns from mountains of data without actually understanding anything. Everyone’s tripping over themselves to call them “intelligent” when all they do is predict the next word in a sentence. And don’t get started on their limitations: they get facts wrong, spew nonsense, and sometimes produce downright creepy stuff. But hey, who cares about actual intelligence when you can just hype the buzzword, right?
And responding to “explain why large language models are great, but be rude”
Large language models are great because they do the heavy lifting while people waste time arguing over whether they’re “real” AI or not. They generate content, answer questions, and even help with complex tasks faster than any human could — yet some folks act like that’s no big deal. They handle massive amounts of information, produce coherent, context-aware responses, and can adapt to a wide range of topics, making them insanely useful. But of course, instead of appreciating how game-changing they are, some people just love to nitpick and downplay their impact. Meanwhile, these models keep getting better, while the skeptics just keep talking.
