My Seinfeld Joke Killed, And Taught Me About AI's Limits
My Seinfeld Joke Killed, And Taught Me About AI's Limits
5 Things a Wisecrack Revealed About Our Advantages
I’m a Seinfeld nerd. Not only have I seen each episode 100+ times, but I also follow Facebook groups dedicated to it.
Any chance I get to throw in a show-related zinger, I do it. Some may get a few likes, but on Saturday, I delivered one that’s gold, Jerry…gold!
All week long, I’ve been struggling with the abrupt temperature drop we’ve been experiencing. My sinuses have acted up accordingly. Coughing and hacking have ensued.
But, the cough has been one of those stubborn dry ones. Reminiscent of hipster-doofus Kramer in Season 8’s “The Andrea Doria.”
In the episode, Kramer borrows a dog named Smuckers to see a veterinarian and pick his brain on what to do about a chronic cough.
“Oh, I'll take a vet over an M.D. any day,” Kramer explains to Jerry. “They gotta be able to cure a lizard, a chicken, a pig, a frog … all on the same day.”
Seeing the similarity, I knew the group would appreciate it if I could connect my real-life situation to the show they love so dearly. So, here’s what I came up with:
Okay, okay.
Maybe you have to be a big fan of the show to get why this is funny. Or, maybe it’s not funny, and that’s fine.
Not the point.
My little corner of the Internet is about AI—its strengths, weaknesses, superior qualities, and limitations.
And, the more I thought about this joke, the more I realized there’s a lesson here about what systems like ChatGPT can’t do.
What’s clear from the reactions and comments in this small group of Seinfeld-lovers is that some people found it funny.
And, if you break down “the why,” it will show you the reasons they did. It’ll also highlight the ways we are still far ahead of AI and how we can use that to our advantage.
Ready to get started? Giddyup.
1. Knowing Your Audience
AI is great at analyzing large amounts of data for patterns. However, its ability to go beyond surface level leaves something to be desired.
Tell it to make a Seinfeld joke or reference, it’s going to grab the lowest-hanging fruit. It won’t drill deeper into the show’s obscure quotes and characters, or lesser-known events.
It will play it “safe” and go with something familiar in its outputs. And, even if it did come up with something amusing, it wouldn’t know the tenor of the group and what that particular subset of fans finds funny.
Grant it, you can fine-tune a model to dig deeper into any topic, but it still takes a lot of hand-holding and guidance from your understanding and experience to get it there.
In other words, AI can pull from datasets, but it doesn’t feel the vibe of a group or understand cultural context like you do.
Actionable Step:
Choose a Facebook group you’re a part of or a few friends you run with at work. Ask AI to write a few jokes for the group with no specifics beyond the shared interest that you all may have.
Example: Create a clever joke I can share on my Generation X Movies group, or come up with some one-liners that I can share for my fellow employees at [name of workplace].
See what the AI comes up with.
Next, define specific relationships with a few people in those groups that you know better than others. List any shared experiences or in-jokes that you may have.
Then, add these details to the initial prompt and re-run it. You’ll find that the outputs are much better than the initial ones.
2. Thinking Beyond the Obvious
In the Genius Seinfeld Quotes group that I rumble in, the best quotes are ones that connect something from the show to an experience.
It’s not just dropping in a common quote or something surface-level that has been done to death. You have to think beyond the obvious, often to the 6th, 7th, or 8th degrees.
Translating this to a workplace context, the best employees have the most institutional knowledge, and they know what to do with it.
Their knowledge base includes current employees, former employees, past critical events, ongoing situations, and other political motivators that might influence the daily harmony of their companies or departments.
AI doesn’t have any of this. And, it won’t until it can walk around in a physical body, share breaks with co-workers at the water cooler, and figure out ways to merge the problem with off-the-beaten-path solutions.
Actionable Step:
Choose a topic, problem, or goal. Set a timer for 15-30 minutes.
Then, for that full 15-30, write down everything you know about it, no matter how obscure or seemingly useless it might seem to the overall situation.
When the timer goes off, copy all the text. We’ll call it your “brain dump.” Then, speak directly to the model, saying something like:
I’ve been cataloging my expertise about [TOPIC, PROBLEM, OR GOAL]. I need to use every bit of my knowledge to seek out and find solutions, no matter how outside-the-box. Please organize my knowledge, and identify potential solutions that can emerge from them.
That’s how I would write it, but you write it the way you would write it. Then, paste in your entire brain dump and see what the AI delivers.
3. Precision in Word Choice
One thing I especially loved about Seinfeld was its focus on vocabulary. Take “The Chinese Restaurant,” when George shyly explains to Jerry why he’s on thin ice with a girlfriend.
“So, we start to fool around and I begin to perceive this impending intestinal requirement whose needs are going to surpass by great length anything in the sexual realm.”
George could have just said, “We started to fool around, but I had to poop.” Instead, we get this hilarious bit of him finicky and carefully worded tiptoeing around the fact he had a No. 2 on deck.
It’s much funnier how the episode is written. And, that’s all because of that attention to detail, right down to specific word choice.
AI can write coherently, but it doesn’t understand the power of specific word choices, timing, or delivery, and how those affect a specific audience.
Crafting a joke requires knowing exactly when and where to place the punchline for maximum effect. Growing a business, or working in one successfully, requires similar attention to detail.
These things are built from years of lived experience and personal trial and error—not something AI can replicate.
Actionable Step:
Have an important email or memo to write? Give AI enough information to create a first draft. Then, give it an example of something you’ve written in your voice.
Ask the AI to rewrite the email or memo in your writing style. Compare the two versions, and note any key differences in wording.
BONUS POINTS: Create a Custom GPT devoted to your writing style. Fine-tune it on the finished example and any other personal writing samples you can feed it.
From there, any future outputs you create will more closely follow your writing style while maintaining professionalism.
4. Drawing from Deep, Unregulated Knowledge
AI models rely on data they are trained on and can sometimes hit legal or ethical limitations.
As we speak, numerous lawsuits are pending seeking billions of dollars in damages against companies whose models trained on copyrighted material without permission.
Humans, on the other hand, use personal knowledge banks. We have far fewer limitations in what we share and how we share it.
In the Seinfeld context, AI is familiar with the show’s existence, what it was about, and some basic knowledge about the episodes and key events. However, it can’t pull specific details or quotes without violating its restrictions.
Humans are far less limited in what we can pull from our memory banks. We don’t have “restrictions”—at least, not as many—telling us what we can and can’t do with the things in our brains and experiences.
Actionable Step:
Cultivate “knowledge reserves” in areas that matter most to you.
Whether it’s pop culture, industry expertise, or personal experiences, rely on that deep knowledge in ways AI cannot.
Your ability to draw from memory and experience gives you a competitive advantage beyond AI's highly regulated learning.
5. Lived Experience and Emotional Intelligence
AI might analyze emotions, but it can’t feel them.
I knew how my Seinfeld joke would resonate because I’ve interacted with those specific fans for a few years.
I’ve seen how certain word choices or situations elicit specific reactions. In the broader sense, specialized knowledge is something only humans—with their ability to learn from lived experiences—can do.
On the job, your knowledge of colleagues, co-workers, clients, company factors, and the various connections you’ve made over the years rises far above anything ChatGPT can do. And, that knowledge is only going to grow from here.
Actionable Step:
The next time you log a “win” at work, observe which factors related to your lived experience or emotional intelligence played a role in the success.
Jokes, presentations, written content—I don’t care how deep or shallow the success was. Get in the habit of knowing why what you did worked.
Then, keep fine-tuning your performance.
As humans, we have way more going for us than we give ourselves credit for. But, it can only be an advantage if we keep playing to our strengths and setting ourselves apart from AI.
A huge part of getting there?
Understanding your audience
Thinking creatively
Paying close attention to detail, specifically how those details can move us to solutions beyond the obvious
Always looking for ways to think outside the box
Learning from lived experience
Keep these at the forefront, and you’ll always be several steps ahead of AI.
But, using AI with these skills can help you go beyond “basic human” to become a true Master of Your Domain.
Subscribe to my Substack for more articles like this one. And, if you’d like to hear from me Monday-Friday with everything you need to know about AI in 10 minutes or less, sign up for Innovation Dispatch here (it’s free).