Last Thursday, Amazon launched Echo Look. The company’s latest addition, much like its digital cousins, is essentially a Wi-Fi speaker — a sort of Siri-in-a-tube — that listens to and obeys people’s commands. It can call Uber, order pizza from Domino’s, and answer trivia questions. But it can do more than just listen — the oval-shaped device can now see as well.
In addition to reading out news and weather forecasts, the device is equipped with a hands-free camera and LED flashlights. Echo Look can snap pictures when asked and can perform one additional neat trick: telling us what to wear. Try a few outfits on at home, snap a few photos, and the Style Check function will tell you which combination looks best, all based on “machine learning algorithms with advice from fashion specialists.”
None of these should come as a surprise. Amazon has a history of turning human decisions over to machines. To CEO Jeff Bezos, data and facts should reign supreme. During Amazon.com AMZN +0.5%’s earliest days, an editorial group composed of real writers sold products with witty taglines and decided what to promote. Over the years, a competing group used a machine algorithm called “Amabot” to generate recommendations based on customers’ previous purchases and their web searches. The two groups fought on, pitting personable, handcrafted messages against automatic suggestions. Before long, the commercial result was clear: humans just couldn’t compete in generating additional sales. The editorial group was soon dismantled.
So, if machine algorithms are already so good at predicting how we click, buy, and pay online, why not extend the decision-making power of smart machines to consumers directly, telling them what to do at home? Echo Look appears to precisely take this path forward. But the specific choice of the initial application — giving fashion advice — must seem downright peculiar to most outsiders.
From Canary to Nest, smart home cameras are all the rage these days, but most systems focus on home security, energy conservation, and other important utility applications. Amazon, despite being the world’s largest cloud computing provider (AWS) and possessing such an enormous share of the expertise in artificial intelligence and machine learning, has instead chosen to offer a cloud-connected fashion consultant, an almost trivial feature. Why would the company do that?
When Groundbreaking Innovations Start As Toys
In 1957, an upstart company named Tokyo Tsushin Kogyo (TTK) released a small transistor radio about the size of a cigarette pack. It came in a colorful box complete with a soft leather case, an antistatic cloth, and a pair of earphones. The CEO of the company, Akio Morita, went to New York City to canvass electronics store owners and convince them to stock up. His timing couldn’t have been better.
Emerging at that time was a distinctly American phenomenon called “rock and roll.” Adolescents were only too happy to snatch up an inexpensive portable device and listen to their rock music out of their parents’ earshot. Although the sound quality of the transistor radio was poor compared to vacuum-tube countertop models, Morita returned to Japan with a full book of orders. To appeal to the Western market, he christened his company after the Latin word sonus, which means “sound,” and Sony was born.
Technological progress often takes surprising turns. What is predictable, however, is that the initial iterations of a product are always inferior to the later offerings. For this reason, innovators who choose to target applications for which the customer demand is low, often succeed in the long run.
Disruption, it turns out, is as much a form of social evolution as it is of technological improvements. It was teenagers who openly embraced a bewildering range of gizmos as their new forms of entertainment and spurred manufacturers to improve product quality, which in turn ushered transistor electronics into the mainstream. Parents who would otherwise be technologically adverse saw their kids carrying boomboxes and playing on game consoles and suddenly felt compelled to learn how to use videotape recorders. Different segments of the market influenced one another. However, a large body of research has invariably demonstrated that successful disruptions usually begin at the low end.
One’s Awful Is Another’s Awesome
For Amazon, fashion advice is perhaps the lowest denominator for its mighty artificial intelligence, but it is certainly not mindless. Recognizing color, detecting facial expressions, and assessing body shapes are complex skills for a computer to acquire. To form information in real time based on the latest fashion trends, the hottest items on the market, and expert opinions requires the machine algorithm to handle images, numeric input, and unstructured data that include the natural human language found in magazines and on Wikipedia pages. The beauty of such a setup is that there will be few real consequences, even if something goes wrong — aside from wearing a bad outfit and being jeered at by colleagues. Meanwhile, the learning opportunities for Amazon are bountiful.
This is exactly what famed entrepreneur Eric Ries called the “lean startupmethodology.” He coined the term to describe organizations that maximize the opportunity to learn, gather market insights, and minimize resources spent to achieve market commercialization of a new technology.
The biggest danger for innovators is when they become fixated on building a perfect machine in the dark and the market has already moved on, leaving the entrepreneurs stuck with a product that no one wants. To resolve this dilemma, speed matters, particularly in the high-tech world. That is why Ries champions the idea of a “minimal viable product” over a perfect creation. It is often better to build a viable product with the smallest set of features in order for the company to complete, as quickly as possible, the cycle of “build-measure-learn.”
This outlook might stand in stark contrast to Apple AAPL -0.28%, which releases only the most elegant products to the marketplace. But elegance is also in the eye of the beholder. A decade ago, when the first iPhone was released, critics sneered at Steve Jobs because the device had no keyboard. Consumers may have liked its industrial design, but its back-end support, nonetheless, seemed rudimentary. Dropped calls were frequent, coverage was patchy, and the security safeguards were so primitive that most CIOs in big companies refused to switch from the existing BlackBerrys for their employees for fear of cyber insecurity.
For Apple, however, the enterprise market, though big and lucrative, was of secondary concern. It was the consumer market that mattered and delivered growth. Specifically, Apple was concerned about the early adopters who would embrace a general-purpose device that played music, sent emails, made calls, and browsed websites. It wasn’t until much later, only after nailing down all the basics—including developing a sizable app store with hundreds of thousands of third-party apps—that Apple would marshal its attack on BlackBerry. In its early days, the iPhone was an awesome product for adventurous consumers but very lousy for enterprise users. And that was okay. In fact, it was the preferred path for growth.
Viewed in this light, Amazon, despite being the biggest retailer in the world and one of the most valuable publicly traded companies, remains a lean startup at heart. When there were no electronic versions for books, Kindle was launched as a single-purpose device for serious book readers who didn’t bother with color print. And, when every startup needed to buy servers to host its website, Amazon launched cloud computing services, targeting the most commoditized functionalities that companies needed. Amazon is so determined to grow from the low end that CEO Bezos reportedly once said, “Your margin is my opportunity.”
Still, at the heart of all these frenzies is the explicit embrace of experimentation in spreading the corporate risks over rapid learning. “I’ve made billions of dollars of failures at Amazon.com. Literally,” Bezos once recounted. “What matters is [that] companies that don’t continue to experiment or embrace failure eventually get in the position where the only thing they can do is make a Hail Mary bet at the end of their corporate existence. I don’t believe in bet-the-company bets.” And there is no better way to experiment with a technology as disruptive as AI than by giving fashion recommendations, which are simple and easy to fulfill with no regulatory obligations.
Originally published on Forbes.
Outlast your competition and thrive in an ever-changing world
In Leap, Howard Yu, LEGO professor of strategy and innovation at IMD, explains how companies can prosper, not just survive. Leap identifies five fundamental principles that allow companies to stay successful in the face of such competition.