Microsoft CTO Kevin Scott and I share a few things in common. We both grew up in small American towns in the ’70s and ’80s—he in Virginia, me in Nebraska. We both now live and work in the Bay Area. We both make fairly frequent trips back to rural America to see family and friends.
And we’ve both watched as two extremely important trends have taken shape in the first part of the 21st century. The tech industry’s wealth, influence, and relevance to daily life have steadily increased, and will likely accelerate with the further application of automation, robotics, and AI. Big West Coast tech companies such as Facebook and Uber have celebrated IPOs on the floor of the NASDAQ, minting millionaires in the process.
Meanwhile, rural America struggled through a painfully slow recovery from the last recession, exacerbated by the continued exporting of jobs to cheap labor in China and Mexico, and by the destruction of jobs by automation. Largely ignored by the media, the symptoms of that distress began to show, first in the Tea Party movement, then in Occupy, then in the 2016 victory of Donald Trump, the politician most skilled at weaponizing rural America’s growing anger over a “rigged” system.
In the new book Reprogramming the American Dream, Scott and coauthor Greg Shaw confront those two big issues and offer a far more optimistic alternative to the “robots are coming to take our jobs” dystopia much of America associates with the advent of AI. In Scott’s view, AI and robotics can be democratized to inspire and empower new businesses in rural America. Properly applied, it might give traditional businesses like farming and rural healthcare new tools to survive. Far from being a job killer, AI might become the superpower small businesses in small places need to survive and thrive.
Midway through his book, Scott drops the brilliant William Gibson line, “The future is here, it just isn’t equally distributed yet.” His point is that if AI is the future, there’s still time to get the distribution right. Little America may get its seat at the table. A lot may hinge on whether or not that happens.
“In order for us to have a more equitable and inclusive future, we have to have people from all over the country, from all geographies, participating in the process of creating our future,” Scott recently tells me. “I really don’t think that it’s a great thing to say that all novel, innovative things have to be created within these coastal urban innovation centers, and then the folks who live everywhere else are consumers of those things.”
Scott’s own love of computing started while he was growing up in rural Virginia. He became fascinated with video games and soon decided he wanted to write his own. After graduate school, he landed at Google in 2003 and began writing machine learning algorithms. He then left to lead engineering at mobile-advertising startup AdMob—which itself was then acquired by Google.
In 2011, Scott joined LinkedIn, which was acquired by Microsoft in 2016. Shortly after that, Microsoft CEO Satya Nadella chose him to be CTO.
Nadella says that Scott combines computer-science chops with a sensitivity to tech’s real-world implications. “Kevin is a guy who knows how to really go deep in new trends and new technology, like deep learning or these multimodal large-scale models,” says Nadella. “But at the same time he’s grounded in how it’s having a broad impact, so that’s what really drew me to him.”
Home to Virginia
Scott tells me one of the things that put the book in motion was a trip back to his hometown of Gladys, Virginia, a tiny burg not far from rural Appalachia. During his stay, he visited the businesses of a few old friends and saw signs that the wider distribution of advanced technologies was already starting to happen.
“I had been living in Silicon Valley for a while . . . working on machine learning and AI, and a bunch of frontier technological things,” he explains. “The thing that I was really surprised by, when I got home, and was looking with a critical eye, was how many businesses were already using reasonably advanced technology to run their operations.”
In the opening chapter of his book, Scott describes a visit to a friend’s precision plastics machining company whose competitive edge is the use of intelligent machines that cut out intricate plastic parts automatically, based on designs that are sometimes sent from clients (such as Disney) directly to the machine via the internet. The machine does a lot by itself, and needs a single skilled person to oversee it.
Another set of friends, after the collapse of the tobacco industry, switched to a new kind of farming—growing grass sod. Their company invested in an expensive machine that cuts the strips of sod precisely and automatically. Again the machine requires that one skilled person tell it what to do.
In both of these instances, humans weren’t replaced by a machine but rather assumed a role working alongside one. In many cases, Scott argues, the machine can take over the redundant and mundane parts of the job that people don’t want to do anyway. That can free them to do more human things, like pitching new clients, listening to existing ones, and using creativity to solve problems.
Still, it seemed to me that while the machines weren’t replacing people, they were certainly being used to reduce the number of humans needed to do jobs. At the plastics machining company, the machine intelligence increased the efficiency and accuracy of punching out parts, but it also turned a manufacturing job that used to require a team of workers into one that required one machine and one skilled human.
Scott argues that I was looking at it the wrong way.
“Without the machines, there would be zero jobs there, and so the only way to make the business competitive, and actually the way that they started this company, was because they were able to use the best tools that they had, in their hands, to create a brand-new enterprise,” he says.
Small, rural businesses must compete with larger ones that are already well down the road in using technology (or cheap overseas labor) to streamline processes. But Scott says that as technologies mature, and as the platforms that deliver them mature, the playing field may be leveled. Smaller concerns will get access to AI tools, get more experience applying the tools, and grow more competitive. If a small business can use technology to operate in a lean and agile way, its chances of surviving long enough to grow go way up.
“The exciting thing to me is how much easier it is now, with the power of all of this technology, to start an agile manufacturing concern in 2020 than it was in years past,” he says. “Because of the sophisticated ways these machines use automation and robotics and sensing, this allows them to have a cost/productivity curve that looks fairly similar to Moore’s Law.”
Automation has already reshaped the manufacturing landscape in the U.S. “There are not the manufacturing concerns that have very, very large factories full of people, doing very repetitive work,” says Scott. “But they are places where you have lots of machines and you have skilled labor who have to figure out how to use these machines, these tools, in the best way possible to build the things that the company is building.”
In a broader sense, in developed countries at least, the way humans add value to the creation of goods and services has shifted, and will continue to shift, as robotics and AI find their way into more processes in more businesses. Fewer workers add value by punching out the same hole in a piece of metal for eight hours, and more are asked to perform tasks that require creative problem-solving, machine teaching, and soft skills.
Building a support system
On a larger scale, the adoption of high-tech tools by small, rural businesses may not happen as organically as in the cases of Scott’s friends’ businesses in Virginia. The process may need active support from government and the investment community. Each has a key role in helping to make sure that the benefits of AI are more evenly distributed.
Directing capital and investment into rural and middle America is a very good idea right now.”
One of the things that makes it a good idea is the “opportunity zones” section of the 2017 Tax Cuts and Jobs Act, which offers investors tax breaks when they invest in businesses in distressed communities. Scott cautions, however, that it’s important to incentivize investors to put significant amounts of their money in for the long term, and not just make smaller, short-term investments for quick profits. The government could go much further in those incentives and could offer small businesses themselves incentives for investing in AI.
Scott says in his book there will be a temptation for companies to harvest the efficiencies created by AI for short-term gains rather than reinvesting the money they save. As an example, a company may decide to fire all its human customer service agents and replace them with software agents. This would save the company from paying those salaries, but it would throw away human talent that could be used elsewhere in the company, and it would harm the community by increasing the number of its unemployed.
The Virginia plastics machining business Scott visited was indeed able to operate with fewer humans because of its investment in intelligent machines. This efficiency allowed it to remain viable, and it also created new capital to spend on new machines and skilled people to run them. This benefited the existing employees and the owners but also benefited the community by creating more jobs.
The question of universal basic income
Those two ways of using AI could result in two very different futures. The short-term profit-driven approach, if universalized, points out to a grim future. There are smart people who believe that AI will eventually take away many millions of low-skilled jobs and not create new ones. Large numbers of people would not only be unemployed but unemployable. It’s a scenario that many believe casts universal basic income as an absolute necessity.
Scott, however, doesn’t buy the idea that AI will render millions of people useless. “I most assuredly, violently disagree with this notion that you’re going to have people that are unemployable,” he tells me. “I think to imagine a future where you’ve got people who can’t find a purpose that allows them to connect to society in some meaningful way is just a horrible vision.”
Still, he’s willing to consider the possibility of UBI solving societal problems. He argues that the need for it is more likely to be caused by the country’s inability to bring basic subsistence costs under control. Healthcare, for instance, is more expensive in the U.S. than in any other developed country in the world. UBI, then, might be necessary to make sure people can at least feel sure they can afford the most basic things, especially during hard times.
“You should not have to worry, for instance, when you’ve got a pandemic happening, that the very basic things about your life are insecure—your food, clothing, shelter, your education, your physical security,” Scott says.
Broadband in the boondocks
The lifeline that will distribute AI tools to the country’s rural areas is broadband. And that lifeline, right now, is in bad shape. A Microsoft study last year found that 162 million Americans were not using internet service fast enough to be called “broadband,” using the FCC’s definition of the term (at least 25 megabits per second download speed).
“Not having the right tools to build and sustain a business in the digital era is what it feels like for those outside the urban innovation areas,” writes Scott in his book. “If you live outside of a bustling city, that helpless feeling of not being able to participate in the digital economy can be crippling. You lack access to speedy broadband as well as the digital tools and skills that are practically taken for granted in urban areas.”
Our community is, in a sense, the whole world.”
“I’ll call it the market failure,” Nadella tells me. “It’s the classic example of, ‘Hey look, the markets by themselves are just not going to address this.’ But at the same time, there are very capable telecommunications companies in this country. They will probably need some encouragement and incentive.”
The government could provide that encouragement and incentive. Yet the FCC has failed even to develop an accurate map of where broadband, and broadband competition, exists and where it doesn’t. Congress recently passed a law directing the FCC to develop a better method of gathering and mapping its broadband availability data. And with the current president and Senate leadership, the government is unlikely to do anything drastic to compel big ISPs such as Comcast, AT&T, and Verizon to extend their networks in rural America.
Now that many people are being forced to work from home because of the pandemic, the shortcomings of America’s broadband networks, especially in remote areas, are being laid bare.
“You need it even to deliver the core services,” Nadella tells me. “Think about the COVID-19 crisis. In order to be able to triage patients who are in these remote areas, you need to deliver telemedicine to be efficacious, and that needs broadband connectivity. And so it’s an existential need.”
What do tech companies owe?
All of this leads to the broader question of Big Tech’s responsibility to distribute the goodness of AI and robotics to everyone. Companies such as Amazon, Apple, Google, Facebook, and Microsoft have grown to play central roles in our daily lives. And they’ve achieved their success with the help of American investment capital, talented people educated in American universities, a well-monied market for their products, and a largely hands-off approach from U.S. regulators. Is it enough for these companies to sell tech to consumers and businesses, pay their taxes (well, some of them), and maybe plant a couple of token data centers out in the sticks? Or should they be doing more to evenly distribute the massive wealth they create, even in the flyover states?
“I think we owe a ton to society,” emphasizes Scott. “It’s not just the urban innovation centers in which many tech companies are headquartered. Because we have been successful at building a bunch of technologies that are now deeply integrated into everyday life, our community is, in a sense, the whole world.”
Microsoft is on firmer ground than some other tech companies to say that with sincerity. It sells business software, but it also operates the Azure cloud platform that distributes enabling technologies, including AI, to other companies. And with the arrival of Satya Nadella as CEO, the company in recent years has become less aggressive about dominating markets, choking competitors, and wringing maximum profits out of customers. Nadella’s Microsoft is more about compatibility, cooperation, and enablement than the Microsoft of Gates and Ballmer.
“One of the things that Satya says that I really like is that Microsoft is successful when our customers are successful,” Scott says. “We build things that they use to then go run their businesses and to create the things that help them serve their own customers.”
Like its cloud-computing rivals Amazon and Google, Microsoft has been driving hard to package various kinds of AI—natural language bots, robotics, and computer vision—into modules that customers can easily pull from the cloud and plug into their businesses. But Nadella tells me there has to be a social contract between tech companies and the markets they serve, not just a business contract.
“The social contract of any corporation or enterprise is to create profitable solutions to overcome the challenges of people and planet,” he says. “As a first unit of scale, at the core of your business model, are you creating a surplus around you? That, I think, is the question.”
Scott believes that AI, in its various forms, has matured and become more accessible and easier to manage for small companies. “When I think about AI and machine learning, I know for a fact that it has become so much easier to build things with machine learning now than it was 15, 16 years ago when I did my first machine learning work,” says Scott. “That was a very daunting undertaking in 2004, requiring a lot of mathematical sophistication and writing a bunch of tools from scratch. Now a high school student can do a similar set of things in a weekend using open-source tools and the public cloud.”
Scott’s version of our AI future is a refreshingly optimistic one, and it’s coming from someone who knows the technology very, very well. His book sketches a way of applying AI that could distribute its benefits more evenly than the industrial revolutions of the past. I hope it starts a discussion that might result in the country taking thoughtful, early steps toward a more inclusive and equitable future. With something as potentially transformative as AI, the stakes couldn’t be higher.