GenAI has not yet tasted pineapples

Hugh Schulze
10 min read

You could reasonably argue that more than enough words have already been spilled in the name of generative artificial intelligence (GenAI) by humans and bots. Although the title of this post has a slightly Luddite tone to it, this is no critique of GenAI. But neither is it another breathless promotion of it. Rather, the following is more an exploration of the opportunities—and limitations—of GenAI for marketers and agencies.

In spite of some serious questions regarding "hallucinations"and "poor data quality" of some GenAI programs, businesses are diving headlong into its application. A study by TechNalysisResearch found that 88% of businesses are (as of this writing) currently using GenAI. How it is being used varies from off-the-shelf productivity assistants like Microsoft 365 Copilot and Google Workspace to applications being built on a company's proprietary database.

For marketing organizations, GenAI is producing content for blog posts, social media, and catalog copy. A/B testing, chatbots, and analyses of customer sentiment and preferences are already well-established applications. In email campaigns, GenAI is personalizing content based on user preferences and behaviors. Many of these applications establish feedback loops that connect those insights and recognized patterns back into human experience.

When we talk of "human experience," we're talking about creativity that doesn't rely upon computational power or pattern recognition, but upon living, breathing human beings bringing their own insights and past experience to bear. (No search or query in the world can help me experience the taste of pineapples. It may describe it or propose analogies but language cannot replace human experience.)

A recent survey of 200 marketing professionals conducted by the ad platform, Basis Technologies, found that a surprising 11.5% of C-Suite and VP-level agency executives have already replaced employees with GenAI or are in the process of doing so.

While these numbers are sobering for those of us in marketing, our contention is that human insights and expertise remain essential for the effective application of GenAI solutions.


Human experience and AI input

Our Senior Design Lead, Rob Kerr, created two layouts: one using Midjourney, the other, his own illustration. Without knowing which was which, I chose Rob's collage. (Sure, he could have included "collage" in his query, but I'm pretty sure he still would have won.)

While the internet can't seem to definitively agree on the first use of the phrase "Garbage In/Garbage Out," I think we can agree that the concept itself—of optimizing, decluttering, or de-trashing input—is as important when applied to GenAI as it is to any area of computer science.

Recently, Bloomberg pronounced the "Prompt Engineer" (aka "AIWhisperer") AI's "HottestJob." Among the many job descriptions that have popped up for this role in the past year, this one speaks to bringing human experience to bear: "a professional who specializes in developing refining and optimizing AI-generated text prompts to ensure they are accurate, engaging and relevant for various applications."

Here, at the threshold of engagement with GenAI, we can see how essential human experience will be for marketers. The better the query or prompt, the better the outcome. But any query or prompt is also deterministic. How do we ensure our query for "pineapples" has the guardrails to ensure our results are solely relevant to the fruit? What parameters define the "best" recipes for "pineapple upside-down cake"?


Human experience and AI refinement

In the year-plus since c|change began working with the beta-test version of the GenAI program, Midjourney, our writers and designers have learned a great deal about how to create more effective prompts. Part of the process is learning how to refine searches to generate images that resonate with our clients and their customers.

For example, as Adam, one of our designers puts it, "Midjourney has a strange idea of what 'futuristic' is—and it's like that with many ideas you'd think would get a better result."

Maggie, another Midjourney expert, has refined ways of working with Midjourney tools such as parameters to define more than just aspect ratios and tiling effects. She's ventured into working with "chaos" which changes how varied the results will be. As the parameters define it, "Higher values produce more unusual and unexpected generations."

"My search and refinement tends to look like this," Adam explains, "One: most important subject; Two: details to help define the overall mood and tone. For example 'bright' or 'dreary,' or I might add the time of day to help with lighting—'night,' 'twilight,' or 'morning'; Three: then I add render words like 'ultraphotorealistic,' 'cinematic,' 'noir,' and then Four: finally I keep randomizing the output until it gives me something close to what feels right. For example, if it keeps including balloons, I need to go back and specify 'no balloons' in the prompt and start over."

Humbly, Adam will tell you that the exponential development of Midjourney (version 5.2 was released in June 2023) has improved the quality of the images as much as the refinement of his searches. But, at the core of the compelling, high-quality images he and Maggie create are their years of experiences as digital designers and the unquantifiable variable of their exceptional curiosity.


Human experience and creative decision-making

With the proliferation of APIs, the application of neuro-linguistic programming, and access to internet databases, we might have finally achieved giving a near-infinite number of monkeys a near-infinite number of typewriters. It shouldn't take long until they write Shakespeare's Hamlet. But who is going to search the other texts to discover his long-lost, never-seen-before drama? And what about all the non-Shakespearean plays that will be created in the process? Who/what will find those? Another GenAI program? With what (human) criteria?

There are reports that entertainment executives are at work on AI-generated scripts (as well as investigating deep fake, AI-generated actors, and other applications). But in the not-too-long ago, pre-AI days, executives hired film school students to sift through "slush piles" of unsolicited scripts. Even then, the question of overlooked or rejected masterpieces made them questionable gatekeepers. As legendary screenwriter William Goldman (Princess Bride, Butch Cassidy and the Sundance Kid) famously pronounced, "Nobody knows anything... Not one person in the entire motion picture field knows for a certainty what's going to work." Of course, this has not kept the studios from using AI to assess scripts for certain "formulas."

For marketers, as for entertainment executives, the question becomes: are we ready for what GenAI is going to deliver? How can we find the best results when you're drinking from the firehose?

If all marketers wanted to do was deliver "content," Gen AI's work would largely be done. But our goal in communicating is connecting to living, breathing human beings and demands we not just get the message "right" or deliver information in a computational sense. We have to connect on an emotional level of some kind—depending on what we need to communicate (an essential question all its own)—we must intrigue, tantalize, engage, excite, surprise our target audience: human beings we're asking to pay attention, remember, and act.


Conclusion: Deepening the experience

After completing the first draft of this article, I asked ChatGPT, based on a summary of the topic, to provide 10 headlines for the article. IMHO, the top three were:

  1. From Creativity to Compassion: Unraveling the Limitations of AI's Human Replication
  2. Soulful Insights: How AI Lags in Understanding the Depth of Human Consciousness
  3. Beyond Algorithms: Why AI's Decision-making Can't Match Human Judgment

All 10 included the "Headline: Subhead" pattern and in a novel way, introduced words and concepts I had not used such as "soul" or "consciousness" or "the beauty of art and expression." The third one is, to my thinking, a misinterpretation of what I've been trying to communicate.

But more significant for this exploration, there were no pineapples. And yes, we had no bananas either. No fruit of any kind. No metaphor. No music beyond the informational.

Now, some readers may prefer one of the three headlines above as a title for this article. In my many decades as a creative director, I've had clients who would prefer them as well. (Certainly, our social media expert would prefer them for search engine results—but that's for another day!)

But my goal was to do more than explore GenAI. I wanted to connect to that place in your cerebral cortex that understands the experience of pineapple, that may have even evoked the taste of pineapple on the tip of your tongue. Because for marketers today, the need to break through the noise to truly connect with people is going to require more than delivering "content," more than "information." It will require human beings moving other human beings.