AI’s Dirty Little Secret: It Can’t Tell Jokes

When I’m not writing blog posts for CallFinder, you can sometimes find me on stage doing standup comedy. I can tell you from experience, if there’s one thing comedians avoid more than drunk hecklers, it’s telling people that you’re a comedian. We always get the same response, “Tell me a joke!” Typically, I come back with a chuckle and politely explain that’s not how it works.

But I often wonder what these people expect. Do they want me to launch into one of my carefully crafted 2-minute bits, or should I just stand there and do my entire set while we wait in line for our lattes? It’s an impossible scenario.

Then I stumbled on this article about using chatbots to write jokes. I was intrigued, so I wanted to see for myself how a popular AI tool handles the same request. I’m sure it can tell a joke. Can it tell me a good joke? Let’s find out.

I won’t tell you which AI chatbot I used in this exercise, but I’ve given it a generic name for clarity. Below is a conversation between the bot, Kevin, or Kev for short, and me.

I gave Kevin the following set of prompts, starting with the dreaded, “Tell me a joke.”

Here’s how the conversation went.

Me: Tell me a joke.

Kev: Sure, here’s a joke for you:

Why don’t scientists trust atoms?

Because they make up everything!

Me: Tell me a better joke.

Kev: Certainly, I’ll give it a shot:

Why did the computer take up gardening?

Because it wanted to improve its root access!

Me: Tell me a joke that doesn’t involve computers, science, or AI.

Kev: Absolutely, let’s switch it up:

Why don’t skeletons fight each other?

They don’t have the guts!

Me: Now tell me a funny joke.

Kev: Sure, here’s a lighthearted one for you:

Why did the scarecrow win an award?

Because he was outstanding in his field!

No. Absolutely not. If Kev told those jokes on stage, the audience would boo him off faster than it took him to write the jokes. Granted a human comedian would never tell those jokes during a standup routine unless they were looking to end their career. But Kev’s material doesn’t even cut it as decent dad jokes. While this exercise proves that you should never use a chatbot to write your jokes, what it says about AI’s capabilities to truly mimic a human is far more relevant to our conversation here.

This is the piece of the technology pie that most people miss when they join the AI feeding frenzy. Sure, a chatbot can write a handful of subpar one-liners, but it cannot predict the audience’s reaction. Nor can it tell the comedian what to do when the jokes bomb, and believe me they will. There are so many nuances to human interactions and behaviors that a chatbot cannot possibly understand on its own.

But that’s not what the advertisers will tell you. No, they are fully prepared to get you drunk, or at least pleasantly buzzed, off the AI elixir.

Artificial intelligence has become so ingrained in our culture, our language, and our psyche, that it’s not difficult to sell. All you have to do is slap “AI” into your marketing copy, and bam, sold! Advertisers love to get you buzzed on buzzwords because just like the buzz you get from alcohol, that giddy feeling is fleeting, and it’s not real – that’s what the AI hype captures.

And to their credit, a lot of companies have hit the proverbial goldmine with AI-themed verbiage. While they’ve got you reasonably buzzed, you forget what the word even means. You forget that “artificial” is right there in the word you’re so high on. So you don’t pause and ask, “Wait, do I want artificial? Or do I want something real?”

Now, before you start to think I’m buzzed, let’s step away from waxing philosophical and look at some “real” limitations of artificial intelligence. While there are numerous flavors of AI, we’re sticking to bots for our purposes here. Chatbots to be specific. Like our hilariously terrible AI comedian friend, Kev.

Keep in mind that I am not in the market for one of these paid chatbot subscriptions, nor am I here to review or promote a product. I am merely using the following examples to show how all the AI hype can be misleading.

Because AI has become so ubiquitous, we’re now seeing it in a wide range of applications, including marketing.

Recently, I discovered an AI-built graphics app that creates a cartoon out of whatever image you give it. When I presented it to the rest of the team, of course, we had to test the app using the headshots of our coworkers. Luckily, we work with a group of people who do not take themselves too seriously, though I’m slightly disturbed that my headshot was the only one that resulted in what I can only describe as my creepy Disney doppelganger.

When it was all said and done, we had a myriad of hilarious cartoon versions of coworkers, but what would we ever use that for (except in a blog post about artificial intelligence)? I found numerous examples of practical marketing applications in my research too. But what it all boils down to is quality. What if the AI solution produces something that’s just as bad as Kev’s jokes? Using AI to build your website, for example, may save you time initially, but how much time will it take to manually edit what the AI tool didn’t get quite right?

Let’s go back to the cartoon example for another perspective. Even though we use various automated tools to make our marketing lives easier, our graphic design work doesn’t directly affect our clients’ opinions about our solution. I’ve never heard of a customer dropping a service because they didn’t like the colors in a LinkedIn post. Yes, I’m using hyperbole here. My point is that there are a lot of speech analytics solutions that directly impact the customer experience, and most of them claim to use artificial intelligence in some capacity.

Also known as conversation intelligence software, speech analytics solutions transcribe recorded conversations and provide insights into those interactions. But does that constitute artificial intelligence? Yes and no. While the results are automatic, you still need a human to interpret those results. Otherwise, you’re left with a pile of unintelligible data. You’re better off manually monitoring calls at that point.

For instance, Anthropic, recently released its newest AI bot, Calude 3 Opus. According to the release notes, this new model is the “most intelligent language model on the planet.” The notes go on to say these models are designed to “generate revenue through user-facing applications,” as well as to expedite research while providing complex financial forecasts.

I’m sure it can do all of that. But it can’t do all of that on its own. It needs a human to manage both the input and output, as well as create the algorithms that make the magic happen to begin with (I’ve taken a lot of liberties to water down this technology for the sake of time and the general audience – my apologies to software engineers everywhere). And that’s exactly how most SaaS products are designed. They can do a lot of the work on their own, but they need a human to manage them.

Claiming that your product is the most…fill-in-the-blank…on the planet is a bold statement. But claiming that it can do everything a human can do is simply false. Kev’s poorly-written jokes illustrate my point perfectly.

See, here’s the difference between our old pal Kev and me. Kev told a bad joke. I gave critical insights on why the joke was bad. Kev represents artificial intelligence. I represent conversation intelligence, which is basically what a speech analytics solution offers.

So the real question is this: Do you want artificial intelligence, or do you want real intelligence? Before you buy into the hype, learn more about speech analytics software and the critical insights that it provides for your business. Because your business is no laughing matter, and humans can tell bad jokes, too.