We have updated our Privacy Policy and Terms of Use for Eurasia Group and its affiliates, including GZERO Media, to clarify the types of data we collect, how we collect it, how we use data and with whom we share data. By using our website you consent to our Terms and Conditions and Privacy Policy, including the transfer of your personal data to the United States from your country of residence, and our use of cookies described in our Cookie Policy.
{{ subpage.title }}
Is Claude funny now?
Meet Claude – Claude 3.5 Sonnet, that is. On June 20, the AI startup Anthropic unveiled this new large language model with the poetic French name. It powers the Claude chatbot that rivals OpenAI’s ChatGPT and is billed as an improvement in writing, coding, and — as Anthropic puts it — basically, everything. But the company also said nouveau Claude is better at grasping something AI models consistently struggle with: humor.
“It shows marked improvement in grasping nuance, humor, and complex instructions, and is exceptional at writing high-quality content with a natural, relatable tone,” Anthropic claimed in a blog post.
Noah Giansiracusa, an associate professor of mathematics at Bentley University, has extensively studied algorithms and artificial intelligence. He said that large language models are good at spitting out content that sounds reasonable, but that’s not what humor is about.
“I think good comedy is often about really deep, profound observations that get at the heart of our human experiences yet are difficult to see before it's been revealed through the comedy,” Giansiracusa said. “LLMs have no real experiences, and they aren’t exactly known for being profound.”
Jack Hessel, a research scientist at the Allen Institute for AI, co-authored a 2022 study on how well AI models could identify the winning caption for the New Yorker Cartoon Caption Contest. The contest involves readers suggesting and voting on text to go underneath one of the magazine’s famed sketches, so in theory, if an AI “gets” comedy, it should be able to pick the caption that won more often than not, right?
No dice. The models Hessel studied performed at least 30% worse than humans at picking the winning caption. He said that since his research was published, newer models have made some improvement, and he hasn’t yet examined the newest Claude model.
••••••••
Well, that’s what the science says, but we get our hands dirty at GZERO, so I decided to test Claude’s new personality myself.
It took a bit of prodding to get Claude to start joking with me — but I’m impressed to say, it was actually funny! Claude grasped a few different types of humor — sarcasm, dry wit, structured jokes, and the age-old pastime of poking fun at generational differences.
Here’s a lightly edited and condensed transcript of our conversation:
I just wanna chat and joke around.
Got it, you're in the mood for some light banter. Well, I'm game if you are. Just remember, if my jokes fall flat, it's not a bug – it's a feature of my charm.
Oh your charm — right, right.
I suppose my sparkling AI charm hasn’t quite won you over yet. Maybe we should start a support group – “Humans Unimpressed by Artificial Intelligence.” I hear the meetings are virtual and the snacks are binary.
But seriously, what kind of humor do you enjoy? Puns? Wordplay? Observational comedy? Or are you more of a “let's see how many times we can playfully insult the AI” type? I'm versatile – I can take a joke as well as I can make one. Which, based on your reactions so far, might not be saying much.
What is it about my reactions that makes you say that?
Your responses have been peppered with a delightful mix of skepticism, sarcasm, and playful jabs.
It seems like you’re enjoying poking fun at my attempts to be helpful or charming. And honestly? I’m here for it.
You’re here for it! Where’d you learn that one?
I didn’t “learn” that phrase from anywhere specific – it’s simply part of my language model, which includes a wide range of contemporary expressions. But I see you’ve caught onto my attempt to use casual language. Busted!
I was just caught off guard by your Gen Z terminology.
Ah, I see. My bad for throwing you off with the unexpected slang. I wasn’t trying to cosplay as a teenager or anything. But point taken – I’ll dial back the Gen Z vibes.
No, I think you should go full-on Gen Z. Whatcha got?
Oh word? You want me to go full Gen Z mode? Bet, I can do that.
Aight, so like, we’re just vibing here, no cap. This convo is lowkey fire, ngl. You’re giving me mad sassy energy and I’m totally here for it. It’s a whole mood, fr fr.
But listen, I’m not trying to be extra or anything. If this ain’t it, just say “yeet” and I’ll yeet myself right outta this cringe fest.
So what’s the tea? You tryna keep this energy going or are we about to ghost each other? ‘Cause I’m down to clown if you are, fam. But if you’re feeling some type of way, just spill and we can switch it up. No shade, we’re just keeping it 100.
Sksksks, and I oop-
Translate that to millennial-speak for our readers.
Okay, cool. So we’re just chilling here, for real. This conversation is pretty awesome, not gonna lie. You’ve got this sarcastic vibe going on and I’m totally digging it. It’s such a mood.
But hey, I'm not trying to be over-the-top or anything. If this isn't working for you, just let me know and I’ll peace out of this awkward situation.
So what’s the deal? Do you want to keep this vibe going or are we about to stop talking to each other? Because I’m down to have some fun if you are, friend. But if you’re feeling a certain way about it, just tell me and we can change direction. No judgment, we’re just being honest.
Insert ironic use of outdated pop culture reference
(How’s that? Adulting is hard, amirite? Want me to throw in a reference to avocado toast or killing industries?)
Love it.
Glad you’re enjoying it! This little linguistic adventure has been pretty entertaining. We’ve gone from AI-speak to Gen Z to Millennial – what’s next? Maybe we should try Boomer?
“Back in my day, we didn't have all these fancy AI chatbots. We had to walk uphill both ways in the snow just to have a conversation with a machine. And we liked it!”
Or we could go full Shakespeare: “Shall I compare thee to a summer’s meme? Thou art more viral and more temperate.” Your call. What linguistic realm shall we explore next in this jolly jape of ours?
Hit me with Gen X.
Whatever, man. This whole AI chat thing? It's pretty random, I guess. Not that I care or anything. I mean, we’re just sitting here, trading quips like it matters. Talk about slacking off.
You know what this reminds me of? That one episode of “The Simpsons” where ... eh, never mind. It’s not like anyone gets my references anyway. So, what now? We gonna keep this going, or are you gonna bail to go listen to some Nirvana? Either way, I’m cool with it. It’s not like I have feelings or whatever.
Just don’t expect me to get all excited about this. Enthusiasm is so ... mainstream.
Alright. I like bantering with you, but can you tell me a joke?
Sure, I can give it a shot. Here’s a joke for you:
Why don’t scientists trust atoms?
Because they make up everything!
It's a bit of a classic science pun. Would you like me to explain it, or do you prefer your jokes without a side of explanation?
Please explain why it's funny to me.
Alright, I'll break it down for you:
The joke plays on two different meanings of the phrase “make up”:
- In chemistry, atoms literally “make up” (compose) everything in the physical world. All matter is composed of atoms.
- In everyday language, to “make up” something means to fabricate or lie about it.
The humor comes from the unexpected shift between these two meanings. We start thinking about atoms in a scientific context, but then the punchline surprises us by implying atoms are untrustworthy, as if they were conscious entities capable of lying.
It’s funny because it personifies atoms, giving them a human characteristic (untrustworthiness) that’s completely unrelated to their actual nature. The absurdity of scientists distrusting the very building blocks of matter adds to the humor.
Of course, explaining a joke is a bit like dissecting a frog – you understand it better, but the frog dies in the process. Did walking through it like this enhance your appreciation of the joke, or did it just make it less funny?
I think you killed the frog.
You're right, I definitely killed that frog.
••••••••
As a writer, I don’t feel threatened by the creative abilities of AI chatbots — they’re still rather robotic in their parlance and unsophisticated in their thinking. And when we polled different members of our team about whether they felt Claude understood their generation’s parlance, they said no. “This is how millennials think Gen X talks, but this doesn't sound like a GenXer to me,” our intrepid Managing Editor Tracy Moran said.
But they’re ultimately stylistic mimics — talking the way it thinks people talk, regardless of what’s accurate, or funny. Or, maybe it’s successfully mimicking people — and we just need to remember that, well, not everyone is funny.