AI, AI, AI, you’re just not funny!

Marketers & Writers - How to Create Content that Stands Out from the AI Competition

By Paul Brunato and Bob Finlayson

We’ve all seen plenty of AI-generated content – both text and images – and it can be pretty impressive. But just how far can AI go in replacing human creators??

Many people claim to be able to recognize content that is AI-generated, suggesting it is bland, template-based, and/or lacking depth, originality, creativity, etc. They contend that humans possess an intangible creative capability that AI lacks. Others believe AI will soon overtake human creativity.?

This debate has obvious implications not only for writers, artists, and other “creatives,” but for all marketing and communications.

We don’t know what the future holds for AI, but let’s do a simple, fast, subjective experiment to test one type of creativity with two of the most popular Large Language Models (LLMs): GPT-4 from Open AI and Gemini from Google.

We already know these products can whip out some impressive short stories and blog posts, so we decided to challenge them with a more subtle example of human creativity – humor.

THE TEST?

Prompt: “Tell me five one-liner jokes about algae.”

Gemini Response:

• I went on a date with a piece of algae, but it was a little too clingy.
• My friend tried to open a business selling algae smoothies. It didn't take off - no one wanted to be "seen with green."
• What did the stressed-out algae say to its therapist? "I feel like I'm constantly under a microscope."
• A fish walks into a bar and orders a drink. The bartender says, "Sorry, we don't serve algae here."
• Why did the algae win an award? Because it was outstanding in its field.

Chat GPT 4 Response:

Sure, let's dive into the world of algae humor with these ten one-liner jokes:

• Algae always know how to have a good time; they're always up for a little photosynthesizing!
• Why don't algae struggle with loneliness? Because they're part of a huge "sea-weed" social network!
• What did the algae say to the chloroplast? "Don't leave me, you're the light of my life!"
• Why did the algae go to therapy? It had too many submerged issues.
• What's algae's favorite type of music? Algae-Rhythms!

I hope these bring a splash of laughter to your day!

Um, okay… No. These are horrible.

Human Humor

It is tricky to identify exactly what is missing, but these jokes remind us of a very young child’s attempts at “adult” humor.

For contrast, here are five algae jokes presumably written by humans, from upjoke.com:

• I made a video about the symbiotic relationships between fungus and algae. Don't forget to lichen subscribe!
• Why had the two algae never had sex? Because they had a planktonic relationship.
• I saw a rainbow flag covered in seaweed. AlgaeBTQ.
• What kind of math do fish like? ALGAE-bra.
• What did the Hawaiian mathematician say when he was at the aquarium? That’s algae brah.

Okay, that’s what we’re talking about… It’s a stupid subject, but some of these are actually funny!

Why can’t the AI do this?

I don’t want to speak for them, so let’s ask one.

Prompt: “Tell me why LLMs are not good at humor.”

Chat GPT 4 Response:

Language models like me can struggle with humor for several reasons, and understanding these can help you see both the strengths and limitations of these systems:

1. Understanding context: Humor often relies on understanding nuanced social contexts or shared human experiences. While language models can process and generate text based on patterns they've learned from their training data, they might not fully grasp the deeper social or cultural contexts that make something humorous.
2. Subtlety and sarcasm: Humor often involves subtlety, wordplay, and sarcasm. Capturing the nuanced tone or the intention behind a sarcastic remark can be challenging for language models, as these require an understanding of not just the literal meaning of words but also the intention and context in which they're used.
3. Timing and delivery: Much of humor comes from timing and delivery, which are inherently tied to human interaction and can be difficult to replicate in text form by a language model. The effectiveness of a joke or humorous comment can depend significantly on how and when it's delivered, which is a nuance that may not be fully captured by a text-based model.
4. Cultural and personal sensitivity: Humor is highly subjective and culturally dependent. What is considered funny in one culture or to one person might be offensive or incomprehensible to another. Language models can find it challenging to navigate these sensitivities without inadvertently crossing lines or misunderstanding the nuances of different cultures and personal preferences.
5. Creativity and originality: While language models can generate content based on patterns seen in their training data, creating truly original humor that feels fresh and surprising can be difficult. The essence of much humor lies in seeing the world in unique ways and making unexpected connections, a process that is inherently creative and human.
6. Emotional intelligence: Successful humor often requires a high level of emotional intelligence, including empathy, social awareness, and the ability to read the room. These are areas where current language models, which do not have emotions or consciousness, can fall short.

Despite these limitations, language models are continually improving and can often generate humorous content, especially in well-understood contexts or when using patterns of humor seen in their training data.

However, the nuanced, personal, and context-dependent nature of humor means that there will always be challenges in this area.

And there you have it.?

One way to stand out from the AI competition as a writer/artist/creator is to focus on your ability to create “nuanced, personal, and context-dependent” content. That should help keep you ahead of the competition… For now.

Need help creating content that stands out from your AI competition? Let us know. Message us for a free consultation Bob Finlayson or Paul Brunato .?