The problem with ai is the problem with ai.
If you’re expecting a diatribe about the evils of artificial intelligence and how it’s going to destroy humanity, you’re going to be disappointed by what this human has written here.
AI has actually shown itself to be rather helpful as a supercharged search engine, an explainer of jargon-laden medical and legal reports, and a retoucher of photos for those of us who don’t have time for Photoshop.
What it isn’t very good at, though, is coming up with original ideas. And why should it?
The underlying principle of the large language models fundamental to artificial intelligence is to absorb everything that’s ever been said, written, photographed, or recorded, and then look for predictable patterns.
So, when you ask AI to write a press release about the new air fryer your company just released, it looks at all the press releases it’s ever seen, all the appliance specs it’s ingested, and all the marketing hyperbole it’s endured and spits out a very passable replica of that which has come before.
The problem is, that which has come before is inherently familiar to your audience, and thus, inherently boring and ignorable. What makes marketing communications interesting and intrusive is when it throws you a curveball — something you haven’t seen before.
Great communications also has strong emotional underpinnings, but it’s a lot to ask AI to pluck our heartstrings. After all, an LLM has never been fired from a job, had a first kiss, or sunk a hole-in-one. So if there’s an achilles heel to AI-generated communications, it’s that the end-product often lacks both novelty and emotionality.
By contrast, think about communications and marketing success stories. Why was the ice bucket challenge a viral social media sensation for ALS fundraising? Because it was weird and different. What makes the Dr. Rick commercials for Progressive stand out in the humdrum world of marketing insurance? It’s the same. Or, rather, it’s the difference. How do you explain the success of Apple’s retail stores? There’s never been anything like them.
There’s undeniable power in newness. And so if you bring AI into the mix, a tool that excels in recombining elements of oldness, you’re unlikely to till new soil.
The other day, a friend sent an audio file that sounded very much like a podcast about the considerable achievements of the nonprofit where she works. If you had any doubts about the good your donations were doing, the info conveyed made a strong case.
It was outstanding. But if there was one flaw in the file, perhaps it was because this slick bit of audio sounded a little too synthetic, despite the pregnant pauses and asides of the smooth-as-silk voices of the two presenters. That’s because it wasn’t real. It wasn’t human.
This quasi-cast was created by Google’s Gemini AI in a product they offer called Notebook LM. Give it some printed words to gobble up, and it will give you back a very credible-sounding podcast.
Because it lacked the fallability, humanity, and rough edges of a human-made project, it came off as too slick. You might say it was slippery, making it less effective in gaining a toe-hold in the listener’s mind.
So, while AI is good at so many things – writing computer code, for instance, a study in predictability – it’s not so great at being strange, twisted, imperfect, nutty, spontaneous, joyous, distracted, moody, and all the other sweetly human traits that make us stand up and take notice.
Take this little essay. Could it have been written by AI? Sure. But we bet you wouldn’t have stuck with it all the way to the end.
