Nuance, Context, and AI Kryptonite

Beyond killing the high school essay and putting C-minus creatives out of work, we’re still sorting out how AI will ripple through our lives.

Much of the speculation about AI questions if there’s any moat around our humanity. Is there anything humans do so well that we can never be replaced by machines? Our livelihoods, and for the more dramatically inclined, our lives, may hang in the balance.

Humans have already lost the battle of raw processing power. Our wonderful and mysterious brains are no match for the brute-force computing power of machines. Given that we’re massively outgunned on what scientists call “pace of operation,” we need a uniquely human ability to leverage in our battle to stay relevant.

Self-driving cars offer a hint at a possible human differentiator. Putting aside a healthy number of crashes and “minor contact events,” AI-powered autonomous cars have logged millions of safe, real-world miles. But despite their overall success navigating a wide variety of complex driving scenarios, self-driving cars often still struggle with the simple four-way stop.

The rules for a four-way stop are straightforward, but things rarely go the way they taught us in driver’s ed. Most stops are an exercise in erratic eye contact, frantic hand-waving, gentle encouragement, and mumbled expletives. Things usually sort themselves out but only after a messy, random, very human set of interactions. Could AI sense that the guy in the truck across from you is angry and impatient, or the kid in the SUV to your left is a jittery new driver, or the driver to your right may or may not even know there’s a stop sign? Even the most thick-headed of us humans can usually read subtle, non-verbal cues, assess the larger context, and determine how to respond. AI knows objects, rules, and patterns.

Nuance and context could be the AI kryptonite we’re looking for. I recently watched a replay of a presentation by renowned relationship expert Esther Perel from this year’s SXSW. She tells a story about discovering someone had built an “AI Esther” based on her publicly available work. Real Esther was nervous and curious, worried about the clinical and ethical implications of such a service and concerned about being replaced by a synthetic, super-therapist doppelganger.

AI Esther provided basic, sound guidance, but Real Esther described it as shallow, a cartoon version of herself, not completely worthless, but lacking the depth and dimension of Real Esther. Playing excerpts from her real-world interactions with clients quickly exposed the complex underpinnings that create and shape client experiences. Patients described traumas, childhood abuse, rational and irrational fears, complicated life circumstances, economic challenges, health problems, and a litany of other factors affecting their lives. Understanding the underlying context is critical to establish the empathy and trust required in Esther’s patient relationships. Failure to consider that full context wouldn’t just be ineffective. It would be irresponsible. Addressing nuance and context isn’t optional. It’s essential.

Where else can we see the power of nuance and context in action? Comedy. ChatGPT can spit out a solid, groan-inducing Dad Joke or, with the right prompts, render a mediocre comedy series script, but good luck extracting anything that resonates like genuinely great comedy. Dave Chappelle is controversial to some, a comedy master to most. His ability to access the subtleties and intensity of culture, race, and religion is unmatched. Dave’s fearless about tackling highly-charged themes and isn’t afraid to pay the PR price if his material doesn’t land well. But most often, his audacity and intimate understanding of culture create intelligent, insightful comedy gold.

Imagine what might pop out the other side of an AI prompt to create a Dave Chappelle-style stand-up comedy set focused on current racial and societal trends. Would you stand up in front of an audience and deliver the output of that prompt? Uncanny content isn’t the only issue with an AI Dave. Real Dave’s delivery is as important as the content itself. No AI could replicate Dave’s distinctive voice, tone, cadence, and nonchalant hit on his vape pen before the deadpan delivery of a punchline. It seems unlikely any AI, now or in the future, could ever possess Chappelle-level originality, intuition, and finesse.

Cormac McCarthy, author of books like Blood Meridian, No Country for Old Men, and The Road, was also famously known for not being a fan of punctuation. In an interview, he once told Oprah Winfrey, "If you write properly, you shouldn't need to punctuate.” Revered by the most seasoned and successful authors, McCarthy’s next-level skill allowed him to create compelling stories while thumbing his nose at the traditional rules of writing. Properly punctuated Cormac McCarthy wouldn’t be Cormac McCarthy. Could anything McCarthy-esque be created by AI?

I advise CMOs on brand and strategy. A well-crafted AI prompt for creating a marketing strategy returns surprisingly credible recommendations. AI is effective if you need a jump start on the basics. But marketing strategies like driving a car, mental health, comedy, and writing can be messy.

A realistic marketing strategy prompt might be “Create a marketing strategy for an organization where the CEO doesn’t believe in marketing, the marketing budget is half what it should be, the previous CMO was barely competent, the current marketing team is demoralized, and our competitors are at least three years ahead of us.” An off-the-rack AI marketing strategy recommendation of “Begin by understanding your target audience” isn’t going to be very helpful.

Understanding multi-layered context, implied messages, subtlety, and subtext are human superpowers. We still have time before AI can match the human ability to provide deep empathy and emotional support, contextually complex business advice, or the perfectly delivered comedic truth bomb.