Spoiler alert: Sentient machine learning does not exist. Robots cannot think for themselves. Not the Terminator, not HAL 9000 nor R2-D2 – all a mere construct of fiction, powered by some magical element that hasn’t been invented yet.

Because our definition of AI originated in science fiction, our expectations are radically misaligned.

In fact, not only can AI not think for itself, it's 100 percent reliant upon humans to survive and thrive.

You know how humans are the only mammals that are completely helpless when we pop out? We can't stand up, we can't feed ourselves, we are utterly defenseless and reliant on other humans to survive and thrive. If you think of AI as a human, it would be about three months old and – say it with me – totally reliant on humans to survive and thrive. It’s a symbiotic relationship.

Which is exactly why you can't just "set it and forget it" when it comes to artificial intelligence. All AI must be trained by a combination of large datasets, patterns of variants and, yes, humans (more on this later), in order to analyze the output of the data and course-correct as needed. So, yes, while AI is radically altering how work gets done and who does it, the technology’s larger impact will be in complementing and augmenting human capabilities, not replacing them.

What's more is, AI "prediction" is a misnomer, again propelled by fiction and further mucking up our understanding of true artificial intelligence in contrast with what we have now. It's not like you're going to a medium and asking them to predict the winners of next year’s World Cup: the AI must have an enormous amount of data and patterns within to make a recommendation that seems most likely based on that information. That’s all it is, a recommendation – same as Siri suggesting text as you type on your phone. She's not predicting what will come next, she's using historical data to make an educated guess based on the zillions of times X has been followed by Y.

In other words, all AI runs off a series of “if this, then that” scenarios made up of zeroes and ones in a massive, human-input flowchart. (Admittedly, these "flowcharts" are now so complex that no one human can read or potentially even understand all the steps, it feels like magic. But under the hood, it's still the same fundamentals.) And by the way, the quality of that data is essential because it is very much a "garbage in, garbage out" state we are in – another reason for the human to always be in the mix. (And by the way, of course Siri isn't even a "she," she's technically an "it," but even I succumb to the throes of fiction as truth when it comes to AI – that's how much the misdefinition of AI has become a part of our lexicon!)

Another way of understanding the importance of patterns, data and grounding AI expectations is to consider an example from the HBO show Silicon Valley – which is to say "not hot dog" is actually good as it gets.  For those not in the know, that's when Jian-Yang creates an app that's like Shazam for food, but he only has time to feed the app images of hot dogs. So when you hold the app up to a hot dog, it recognizes it as a hot dog. But if you hold the app up to anything else, it only recognizes it as "not hot dog." Why? Just think of all of the data that would have to be entered for the app to additionally recognize even pizza, for example: zillions of variants…round, triangular, square, deep-dish, thin crust, red, white, unlimited types of toppings, all chopped in unlimited different sizes – you get the point.

So the magic recipe is not only data, patterns and variants but must include a hefty sprinkle of humans, too. In a recent Harvard Business Review study researchers found that companies using AI collaboratively with humans see a return on investment of two to seven times that seen by companies using AI without human-in-the-loop components. “The biggest performance improvements come when humans and smart machines work together, enhancing each other’s strengths,” writes H. James Wilson and Paul R. Daugherty.

Through collaborative intelligence, humans and AI actively enhance each other’s complementary strengths: the leadership, teamwork, creativity, reference points, analysis and social skills — including emotion — of the former, and the speed, scalability and quantitative capabilities of the latter. The power of collaborative AI is the difference between great results and truly galactic results, driving more effective messaging, a tenfold increase in social media visibility, and 120 times the social media engagement than with AI without human collaboration, according to Lately.AI research across 3,500 companies.

Part of the essential role humans play in this collaboration is analysis. That means keeping a watchful eye on results and having both the skills to understand and assess the quality of those results as well as the skills to identify and oversee necessary changes. (Ironically, there's currently a massive data analysis skills shortage in the tech industry, which bodes  a continued surge in AI that lacks the human collaboration component, further amplifying the struggle to cut through the noise with any viable impact).  Plus, now, there's more noise than ever before, and without a human element to assist, that noise is worse than generic – it's meaningless, irrelevant, and essentially, garbage.

Exactly why "prompt engineering isn't the future," as another recent Harvard Business Review article gently put it, Posing instead that "problem formulation," meaning the ability to assess and clearly identify problems, which, by the way, relies on data analysis, is. The argument being that our proclivity to only present solutions, worried about the stigma of identifying problems, has essentially blinded us from calling a spade a spade. Hence the explosion of thousands of ChatGPT "wrappers" and all that generic noise. Because it doesn't matter how good a prompt expert you are, if we are all prompting from the same generic datasets and algorithms, unattended by competent human analysis (and one other key component, getting to that below), the results have no value other than "save time." And eventually, that thrill is gone. "Save time" is not enough. It has to resonate and, quite frankly, also make money. But because it's AI, our expectations are high. Not just make money. Make a boatload of money.

But it's not just human collaboration alone that garners galactic results. One challenge with most AI is its inability to deliver results that are genuinely relevant and unique to you and your audience. Buyer beware… "Customized" still mostly means narrowing down generic datasets, a.k.a., still a far cry from customized. What's missing is a continuous performance-learning loop, where results are always tied to your unique analytics. What you don't want is for results to be pulled out of thin air with no understanding of what's valuable to your audience – what will make them react, engage, and buy. When you have a performance-learning loop in place, you're able to not only deliver highly targeted results that are unique to both you and your brand as well as your audience, you have information that is up-to-date, in real time (not a year old as is the case with most LLM models) and not "generic." Which is a long way of saying, only meaningful input, powered by human collaboration and analytics-based results, can ring the register.

It should be noted too that, historically, humans are not only critical to the process but desperate to remain in the loop when it comes to tech advancements. Consider the example of Betty Crocker launching cake in a box in the ‘50s — all you had to do was add water. But the target audience for this innovative product, housewives, thought it was too weird, too lacking. They didn’t feel they’d made anything at all; they had no ownership in the process. When Betty Crocker pulled out the powdered eggs and ran with the slogan, “Just add an egg,” sales skyrocketed.

The symbiosis of AI needing humans and humans needing AI — or technology, or cake-in-a-box, etc. — is a movie we've all seen before. When we fight it, we stall. When we embrace it, we soar.