Phones have AI. Computers have AI. Cars have AI. Televisions have AI. Refrigerators have AI. Search engines have AI. Cameras, headphones, toothbrushes and household appliances are increasingly being advertised as “AI-powered.” Sometimes that means something genuinely useful. Sometimes it means something that was already possible a few years ago but has acquired a fashionable new name. And sometimes it is difficult to tell the difference. For consumers, that may be the more important question of the AI boom. We have spent the past few years asking what artificial intelligence can do. Perhaps it is time to start asking something else: Do we actually need it to do that?
AI has become a selling point
Every major technological shift eventually becomes a marketing language. “Digital” once appeared on almost everything. Then came “smart.” Televisions became smart TVs, watches became smartwatches and ordinary household devices acquired apps, Wi-Fi connections and touchscreens. Now the magic letters are AI.
There is a legitimate reason for this. Modern AI can recognize speech and images, summarize information, generate text, identify patterns and increasingly understand what a user is trying to accomplish. Those capabilities can make familiar products considerably more useful. But the AI label has also become commercially valuable. A product described as “AI-powered” sounds newer and more sophisticated than one described simply as automatic. That makes it increasingly important to distinguish between AI that changes what a product can do and AI that mostly changes how the product is marketed.
Start with the problem, not the AI
There is a surprisingly simple way to evaluate many AI features: What problem does this solve for me? Suppose your phone can remove an unwanted stranger from a vacation photograph. That solves an obvious problem. If your hearing aids can distinguish speech from background noise more effectively, the benefit can be meaningful. If software can summarize a 40-page document before you read it, there is a clear saving of time. But imagine a refrigerator that uses artificial intelligence to recognize some of the food inside it. Interesting. Now ask the second question: What happens because it recognizes the food? Does it help reduce waste? Does it reliably tell you what you are running out of? Does it make grocery shopping easier? Does it save money? Or does it mainly give you another screen, another account and another feature to configure? The technology may be impressive even when the benefit is not. Those are two different things.
The best AI may be the AI you barely notice
Some of the most useful applications of artificial intelligence don’t announce themselves every five minutes. Your phone may improve a photograph automatically. A car may detect an object in the road. Software may filter spam before you ever see it. A device may recognize your voice more accurately in a noisy room. In cases like these, AI isn’t asking you to admire it. It is quietly making something work better. That may ultimately be the most successful form of consumer AI. We don’t normally ask which algorithms make an elevator stop precisely at the floor or how the computer in a modern automobile adjusts the engine hundreds of times during a drive. We care that the machine works. Artificial intelligence may eventually become equally invisible. The paradox is that the more useful AI becomes, the less we may need to think about the fact that it is AI.
But intelligence can also make simple things complicated
There is another side to the story. Adding software and connectivity to an ordinary object can create capabilities that never existed before. It can also create dependencies that never existed before. A conventional appliance has a relatively simple relationship with its owner: you buy it, plug it in and use it. A connected intelligent appliance may depend on an account, an app, a cloud service, software updates and an internet connection. It may collect information. Its manufacturer may change its software after you buy it. Features can improve, disappear or stop being supported. The object hasn’t merely become smarter.
The relationship between you and the object has changed.
That doesn’t necessarily make the product worse. Remote diagnostics might identify a problem before a machine fails. Software updates can add capabilities and fix defects. Intelligent energy management could save electricity. But every additional layer should provide enough value to justify the additional complexity.
AI also wants your data
Intelligence usually requires information. For an AI system to personalize recommendations, it has to learn something about you. To anticipate what you want, it needs signals about what you have done before. To recognize your voice, photographs, habits or preferences, some information has to be processed somewhere. That doesn’t automatically make the technology invasive. Some processing can happen directly on a device, and companies can design systems to minimize the information they collect. But consumers should develop a new reflex. When a product promises to “learn about you,” ask: What exactly does it need to learn? And perhaps more importantly: Who else gets to know it? Convenience has always had a price. With AI, part of that price may increasingly be information.
Don’t buy the adjective
There is an understandable temptation during a technological revolution to worry about being left behind. If one laptop has AI and another doesn’t, the first one sounds more future-proof. If a television promises an AI processor, perhaps the ordinary model suddenly seems obsolete. If an appliance can “learn,” maybe the one that cannot feels old-fashioned before it even leaves the store. But “AI” by itself tells you remarkably little. It doesn’t tell you whether the feature works well. It doesn’t tell you whether you will use it. And it certainly doesn’t tell you whether it is worth paying more for. A better buying question is not: Does this product have AI? It is: What can this product do for me that the other one cannot? If there isn’t a convincing answer, the letters AI shouldn’t change the decision.
We have seen this movie before
Technology has a habit of arriving with more enthusiasm than discrimination.
The early internet produced websites for things that did not particularly need websites. The smartphone era gave us apps for tasks that sometimes required fewer steps without an app. The Internet of Things connected objects whose owners occasionally wondered why they needed to be connected at all. Eventually the novelty faded. The useful applications survived. The unnecessary ones became curiosities, disappeared or were quietly absorbed into other products. AI will probably undergo a similar sorting process.
We are currently living through the exuberant stage, when companies are experimenting and nobody wants to appear to have missed the next computing revolution. That experimentation is necessary. Some apparently ridiculous ideas will turn out to be surprisingly useful. Others will simply be ridiculous. The consumer doesn’t have to know which is which in advance. There is a much easier rule.
Make AI earn its place
Artificial intelligence is an extraordinary technology. Precisely because it is extraordinary, we shouldn’t lower our standards for it. If AI saves you time, helps you understand something difficult, makes a product safer, removes tedious work, expands what you can create or gives you an ability you didn’t have before, it has earned its place. If it adds complexity without meaningful benefit, demands information it doesn’t really need or increases the price primarily because the letters AI look good on the box, skepticism is entirely reasonable.
The future will almost certainly contain much more artificial intelligence than the present. But that doesn’t mean every object in our lives needs to announce that it is intelligent. Sometimes the smartest technology may simply be the technology that does something useful—and then gets out of the way.
