On a recent Monday morning, Elon Musk busied himself on Twitter by predicting how World War III would start.
Inspired by news that Vladimir Putin had told Russian students the country that leads in artificial intelligence will rule the world, the Tesla and SpaceX CEO declared the global race to dominate AI might turn into real war—and that the first strike could well be launched by an algorithm rather than a flesh-and-blood leader. Chastised by one of his followers for the gloomy prognostication, he apologized and then confessed, “I was depressing myself too. :-( ”
Musk is a techno-provocateur with few equals. However, plenty of people share his take on AI. Even sunnier forecasts about the future of AI, detailing how self-driving cars might radically reduce highway carnage, are typically too long-range to offer much of a sense of comfort.
Meanwhile, as everyone muses about where AI might take us, the technology has arrived. First given its name by scientists at a seminal conference held at Dartmouth College in 1956 (they had predicted that programmers would be able to simulate the workings of the human brain in just a few years), AI now has a pervasive and obvious impact, particularly when it comes to the branch known as machine learning and, in especially advanced form, as deep learning. AI is how Google Photos knows that two snapshots taken 50 years apart are both of your great-uncle. It’s how Facebook weeds spam out of your feed. It’s even how the iPhone ekes as much life as possible out of a battery charge.
Increasingly, smartphones, smart-home gizmos, and other devices are morphing into front ends for AI-infused services, such as Apple’s Siri, Amazon’s Alexa, and Google’s Assistant. “If you unpack what’s behind Alexa, behind the Echo, it’s not just a speaker,” says Swami Sivasubramanian, VP of Amazon AI. “It’s actually an intelligent, cloud-enabled digital assistant, using deep-learning-driven speech recognition and natural language understanding.”
As AI begins to touch every aspect of their businesses, the tech giants are jousting to recruit superstars in the field, pilfering brainpower from academia (New York University’s Yann LeCun is now at Facebook) and each other (Jia Li left Snapchat to co-run Google Cloud’s machine-learning group). “Because the technology is so powerful, there’s a large demand for talent that understands how to apply it,” says Scott Penberthy, director of applied AI for Google Cloud. Research firm Paysa released a study in April that showed Amazon investing $228 million in new AI positions, followed by Google ($130 million) and Microsoft ($75 million).
Only a handful of companies can compete at this level. Amazon, Apple, Facebook, Google, and Microsoft “have all these PhDs, and they have all this PhD tech,” says Ali Ghodsi, the CEO of Databricks, a startup that works with businesses to add AI to their processes. “But the rest of the Fortune 2000, they don’t,” leading to what Ghodsi states is an emerging “1% problem,” in which only the largest players have the wherewithal to take full advantage of this new technology.
The saving grace is that businesses of all sizes can avail themselves of some of these AI innovations. In fact, Amazon, Microsoft, and Google are counting on it. Their cloud-computing platforms—Amazon Web Services, Azure, and Google Cloud, respectively—include enterprise AI offerings such as image recognition, natural language processing, and language translation. All three companies see AI as the key to driving future growth of their cloud platforms; at present, Amazon Web Services is a $16 billion business that’s increasing 42% year over year, though that pace has slowed as Microsoft and Google begin to catch up. Then there’s IBM, which calls its flavor of artificial intelligence “cognitive computing” and has effectively branded it as “Watson” to sell as a service. While Facebook and Apple don’t offer their own platforms, they publish academic papers on their research—and, in Facebook’s case, it open-sources some of the technologies it’s created.
As a business tool, AI is still in its infancy. Recent studies by both the McKinsey Global Institute and MIT/Boston Consulting Group reported that only about 20% of companies have implemented the technology in a meaningful way. But unlike past technological inflection points—such as the emergence of e-commerce in the 1990s—AI doesn’t naturally favor nimble startups. Because AI craves data of the sort that can take years to accumulate, “there’s actually an advantage to incumbency, because the more knowledge you have to train your AI, the more valuable it is,” argues David Kenny, IBM’s senior VP for Watson and IBM Cloud.
What AI does share with past tech trends, however, is a tendency to be overhyped in such a way that can obscure its real capabilities. In September, for example, an investigation by medical news site Stat determined that IBM’s Watson for Oncology service, for medical institutions, failed to live up to a promotional campaign that suggested it was a cancer-fighting breakthrough. “We’re probably at the peak of hype and expectations: ‘AI is going to do everything for us, it’s going to take over the world, if you don’t touch AI you’re going to be left behind,’ ” says data scientist Steven Finlay, the author of Artificial Intelligence and Machine Learning for Business.
Still, AI is no mere fad. The dollars flowing into R&D are huge—at the rate of more than $30 billion a year—and the ultimate impact on productivity and enhanced consumer demand are projected to be in the trillions. It’s no surprise that AI has become the focal point in the war among tech’s biggest players, and that it is affecting how many enterprises, in tech and beyond, view their futures.
To understand where AI is today—and where it is heading—businesses must acknowledge both the excitement and the uncertainty. Here are seven practical lessons from the front lines, in industries from tech to retail, craft brewing to real estate. AI is officially everywhere, in ways we all need to appreciate.
1. Pick Your Battles
The leader of a major fashion business recently decided that AI needed to be a tool in his company’s arsenal. But he wasn’t sure what that meant. The enterprise had worked with Google, IBM, and Microsoft in the past. Should it align with one of them on AI? And to do what, exactly?
Even those in the AI field warn against becoming smitten with the technology just because it’s trendy. “Sometimes I talk to customers who are like, ‘Hey, we want to use AI,’ without really thinking about why, or what it can do for them,” says Marco Casalaina, VP of product for Salesforce’s AI, which is branded as Einstein.
“Don’t come in with very high expectations because you think a problem is easy and AI can solve it,” advises Microsoft corporate Vice President Gurdeep Singh Pall, who points out that the sort of tasks that humans regard as mundane–such as folding laundry–can look like insurmountable challenges for software. “At the same time, be ready to be surprised about how well AI can perform,” he adds.
A good way to start is by identifying business problems that AI might help with—and, instead of trying to tackle all of them at once, choosing a manageable pilot project. “If you think you’re going to solve this in one go, it’s never going to happen,” says Deep Varma, VP of engineering at real estate information provider Trulia, who advises AI newbies to “pick very specific pain points.”
The key is to not be seduced by AI’s potential but rather to focus on your own goals. Processes that human beings (read: employees) regard as drudgery are often the best place to begin. For instance, by using natural language processing services from Microsoft’s Azure, travel technology company Sabre is experimenting with a Facebook Messenger bot that can field straightforward questions about existing reservations. Its customers in the travel industry, says director of Sabre Studios Chad Callaghan, “see a future where agents are focused on highly complex itineraries where you really want that person-to-person interaction, and the bot is able to support more routine sorts of requests.”
2. Make Your Big Data Meaningful
Around the turn of the last decade, a bit of technological jargon gained currency: “big data.” Its buzziness reflected a new understanding that there was value in collecting, organizing, and analyzing vast amounts of information about every aspect of a business, from manufacturing procedures to customer interactions. Yet it was far easier to hoard big data than to figure out what to do with it. Many companies “kept collecting data for years and years and years, and it sat on servers and collected dust,” says Mark Johnson, CEO of geographic AI startup Descartes Labs. Enter artificial intelligence, which can identify patterns on a scale that would flummox a mere mortal.
“Data is the food that feeds AI,” says Salesforce’s Casalaina. The more it consumes, the smarter it gets. At Google’s I/O developer conference in May, Google AI chief John Giannandrea explained the concept to me by using an example involving his 4-year-old daughter. She had spotted a giant-wheeled, 19th-century “penny-farthing” bicycle and—once he’d told her what it was—she was immediately capable of identifying any other penny-farthing she might encounter. With computers, “we’d have to show them 100,000 penny-farthings and tell them it’s a bike. But once they’d seen 100,000, they’d probably be better at identifying them than humans are.”
Even companies with plenty of data to mine often need to clean up messy databases (Trulia’s Varma winces as he recalls a company that had stored a default time stamp of 00:00:00 on Thursday, January 1, 1970, for every record), merge disparate repositories, and generally make information algorithm-friendly. “The first thing to do is to take the data out of the databases, make it freely available and accessible,” recommends Jean-François Faudi, senior innovation manager at Airbus Defence and Space. For Airbus, that involved moving its satellite imagery to Google Cloud. Now the company can use machine learning to distinguish between snow and clouds—a feat that, it turns out, computers are more adept at accomplishing than humans.
3. Put Your Knowledge To Work
Companies that already care about data have a head start when it comes to AI, no matter what their category. Craft brewing, for instance, would not make anybody’s list of the industries most obviously poised to benefit from the technology. But Brian Faivre, the brewmaster at Bend, Oregon’s Deschutes Brewery—the eighth-largest craft brewery in the United States—happens to have a computer science degree. (“I home brewed throughout college but didn’t know that you could have a real job in the craft-beer industry,” he explains.)
Faivre has long been intrigued by how data science could be applied to beer making, and the brewery has been logging stats about its production process for years. Making suds is all about controlling fermentation, which breweries do by adjusting temperature. They know when it’s time to do so by extracting liquid samples from tanks and measuring their density—a cumbersome, inexact procedure. But working with a data infrastructure firm called OSIsoft, Deschutes fed data about past production into Microsoft’s Cortana Intelligence Suite, part of the Azure platform. That has allowed Deschutes to begin predicting the optimum time to raise the temperature, eliminating the need for the density-measuring step and shaving a couple of days off the 12-day fermentation cycle. The result: The company can produce more beer without compromising quality.
Ultimately, craft brewing is not about ruthlessly efficient mass production, and Deschutes is a long way from using AI to eliminate the human element of its business: Beer makers are free to tinker with the algorithmically generated recommendations, and they do. “What we’ve always stressed is that our brewers are in control,” says Faivre. But the company’s use of AI to increase production is critical to its future: The additional sales are helping to fund the construction of a new brewery in Roanoke, Virginia, which will give Deschutes a national footprint for the first time.
4. Piggyback If You Can
The fact that tech giants are turning their own in-house AI into on-demand services is a boon for organizations that are tight on resources. Chris Adzima, for example, a senior information-systems analyst for the sheriff’s office in Washington County, Oregon, became intrigued last year by a new Amazon Web Services offering called Rekognition, which includes the ability to recognize faces. The county’s trove of hundreds of thousands of booking photos taken at the time of arrests has become so overwhelming that even filtering a search by age, gender, or race often doesn’t meaningfully narrow things down. That limits its usefulness when police officers need to identify a person of interest, such as a shoplifter caught on camera. “I am not a data scientist, nor do I have any idea how facial recognition or artificial intelligence works,” Adzima cheerfully admits. Within a couple of months, however, he was able to fashion a system that uses Rekognition to match newly taken photos with ones from the archive. So far, it’s helped identify 20 suspects.
It was also an extraordinary bargain. The initial setup cost the sheriff’s office only around $400; the monthly bill from Amazon Web Services is about $6. “With every dollar I spend, I’m accountable to the taxpayers,” says Adzima. “We’re spending such small amounts of money and we’re getting a huge return on investment.”
5. Build If You Must
Facial recognition is a type of AI that’s applicable in various scenarios, making Amazon’s version immediately useful in many fields. In some instances, however, companies need to employ AI that’s been carefully tweaked for a particular purpose.
“We don’t tend to ask our radiologist for art advice, or our lawyer for stock-picking advice. You go to experts for different things,” says IBM’s Kenny. That’s why IBM tailors Watson for specific industries, from education to supply-chain management. His point reflects a basic truth about AI: The more ambitious you get, the less likely a plain-vanilla algorithm will suffice.
Real estate hub Trulia hoped to use AI to rummage through its collection of millions of photos of homes for sale and rent and distinguish among kitchens, bedrooms, and bathrooms—and even notice when a kitchen features such price-boosting extras as granite countertops. That’s not the sort of intelligence that’s available as a commodity.
“Trulia needs to innovate,” says Varma, the company’s data-science guru. To do so, he concluded, “we need to own the computer vision internally.” As a division of Zillow, the leading digital broker valued at $5.5 billion, the company could reasonably aspire to treat AI as a strategic imperative and invest appropriately. Even though it’s an Amazon Web Services customer, Trulia chose to hire its own machine-learning experts and develop its own proprietary models.
Sometimes the individualization can be minimal. The car-buying site Edmunds, which offers prospective buyers resources such as specs, prices, and reviews, has integrated AI into numerous aspects of its business, from forecasting revenue to securing its website. Much like Trulia, it wanted to use the technology to help it sort through hundreds of thousands of photos, in this instance, to identify the types of exterior and interior shots it has of specific makes and models. “We got 90% there using Google off the shelf, and then we were able to just tweak it at the end to be better about understanding vehicle images versus all the images that Google is looking at everywhere else,” says VP of product innovation Greg Shaffer.
6. Get Everybody Involved—And Keep Them Involved
Whether a company seeks lots of help or takes on more of the heavy lifting itself, AI’s worth is deeply tied to the specifics of individual business challenges. Which means that it can only be effective if stakeholders are as engaged and committed as IT staffers are.
“Organizations have the tendency to sit back, just like [after] they purchased technology in the past, and expect tech solutions to do all the hard work for them,” says Sjoerd Gehring, global VP of talent acquisition at Johnson & Johnson. “That’s the one thing that really doesn’t work with AI.” Though Gehring’s job focuses on people rather than technology, he championed J&J’s effort, in collaboration with Google Cloud and recruiting software provider Jibe, to incorporate AI into the way it finds everyone from medical researchers to truck drivers. The company says that appropriate applicants are up by 41% since it began using a search engine powered by Google’s machine-learning algorithms to match a million job candidates a year with the 25,000 positions it fills.
After that, “it’s a continuous process of refinement and training to get your implementation better and better and better,” says Meg Sutton, director of retail client experience at H&R Block. The tax behemoth began integrating advice from IBM’s Watson into its routine this year and found that this input—based on 74,000 pages of U.S. tax code and delivered on a second screen used by its preparers—increased client satisfaction. Now the company is working on version 2.0 for next year’s tax season.
7. Don’t Expect Too Much, Too Soon
The final lesson comes back to a simple, human one: patience. There is much to be gained from utilizing AI, but there’s also much yet to discover. The technology’s ultimate cultural impact—despite all the prognostication, from Elon Musk to Mark Zuckerberg—is impossible to know.
Eventually, as with previous epoch-shifting technologies, “there are people who will ride the wave and be successful, and other people who will go against the wave and then be swept away,” says Yunkai Zhou, who spent years building machine-learning technologies into Google’s ad platform before cofounding startup Leap.ai. Thanks to the bold new experiments of companies both big and small, we are all learning where the current might lead.