I’ve always had a thing for stories. It might be a Southern thing. A rural thing. I grew up surrounded by a lively cast of characters with a flair for narrative and sometimes improbable adventures. Every Sunday morning, I got a two-hour dose of the King James Bible through the lens of fire-and-brimstone preaching. When my mom tired of reading me the same stories over and over, I taught myself to read and begged for trips to the library to keep me flush with new material. History. Greek mythology. Dr. Strange and the Doom Patrol. The World Book Encyclopedia. National Geographic. The Peterson Field Guide to the Stars and Planets. I could not get enough stories.
Turns out that storytelling is not just a Southern or a rural thing. It’s also a tech thing. In The Written World, an inspiring treatise on the power of stories to shape people, history, and civilization, Martin Puchner writes that in order to tell the story of literature he had to focus on both storytelling and creative technologies because their intersection is the starting point. The alphabet is code; paper and the printing press are technologies for communications, just as computer languages are code, and the Internet communicates. Yuval Noah Harari, in his book Sapiens, suggests that stories and the cognitive infrastructure that developed tens of thousands of years ago in humans are the most likely reason that Homo sapiens superseded the other species of humans existing seventy thousand years ago. Harari posits that our storytelling ability and the stories we choose to tell are the foundations of our society at scale. Everything from companies to our currency to the Constitution and our laws is a story that we share with one another, allowing us to coordinate our individual efforts among billions of other human beings. Stories are shared not just through written text. The rise of podcasts and Amazon’s Audible have shown us that the oral tradition lives on. Netflix and HoloLens show that stories can be presented visually in a variety of formats. Jeff Weiner, my former boss at LinkedIn, says that “we are the stories that we tell.” And in a very literal sense, he is right.
There are two prevailing stories about AI: for low- and middle-skill workers, we hear a grim tale of steadily increasing job destruction; for knowledge workers and the professional class, we hear an idyllic tale of enhanced productivity and convenience. At one extreme, we have a narrative imagining of a future where livelihoods and lives are at the mercy of inscrutable machines and the elite who control them. At the other is a vision of a world where jobs and income are no longer necessary, and we all retire to a nice beach somewhere and let the robots do all the work. But neither of these captures the whole story, which is, I would argue, a more nuanced, complicated, and hopeful version of the future. I believe that we are in urgent need of a different story, a story of AI’s potential to create abundance and opportunity for everyone and to help solve some of the world’s most vexing problems. Indeed, in the story I see, with proper safeguards, AI can be a tool that empowers us to do more of what makes us essentially human.
In his 1931 book The Epic of America, James Truslow Adams wrote:
The American Dream is that dream of a land in which life should be better and richer and fuller for everyone, with opportunity for each according to ability or achievement. It is a difficult dream for the European upper classes to interpret adequately, and too many of us ourselves have grown weary and mistrustful of it. It is not a dream of motor cars and high wages merely, but a dream of social order in which each man and each woman shall be able to attain to the fullest stature of which they are innately capable, and be recognized by others for what they are, regardless of the fortuitous circumstances of birth or position.
The American dream is a noble aspiration, one of the ambitions that has inspired citizens and immigrants to the United States long before Adams coined the term. Our belief in this dream has ebbed and flowed over the course of time, and for some has been all but impossible to achieve. But perhaps to a greater extent than any technological revolution preceding it, AI could be used to revitalize the American dream, to reprogram it to support each of us as we seek to be our fullest and best selves.
Right now, the story of technology’s future is in flux in ways not known in the West since the tail end of the Industrial Revolution in the early twentieth century. For some, the coming rise of artificial intelligence and robotics is a utopian story, and they can’t get to the future fast enough. For others, the story is dystopian, and they have a hard time imagining how anyone would want to live in the world that we are creating for ourselves.
Implicit in too many of these narratives is the story that rural America and the people who live there no longer matter. Those tending farms and working in agriculture are less than 2 percent of the American population. In March 2017, the Washington Monthly reported that the “depressed state of rural America is getting a fresh look as a result of the 2016 election, and rightly so. People are asking how to bring back rural prosperity and restore small-town civic life.” In his Requiem for the American Dream, Noam Chomsky writes that the ability for everyone to get a decent job, buy a house and car, and have their kids go to school has “collapsed.” I share his concern but not his fatalism. Reprogramming the American Dream reveals that we’re at an inflection point, with the opportunity to shape how that story unfolds. Realizing that potential, ensuring that it helps the communities where people like J.D. Vance, author of Hillbilly Elegy, and I grew up, requires a principled approach to the development of AI, one that puts it on the public agenda alongside climate change, health epidemics, and public education.
I’ve got a different story to tell, one you haven’t heard before. It’s informed by two formative life experiences. The first is my upbringing in the rural town of Gladys, Virginia, typical of the communities throughout our country that have been—and will continue to be—disrupted by emerging technologies. The second is my current position as chief technology officer of Microsoft, where we’re deeply involved in the development of AI applications and mindful every day about their potential impact on the workers of towns like Gladys, which have always depended on low- or middle-skilled workers.
From where I sit, no one is telling the right story, a realistic story, about AI, which is neither utopian nor dystopian. My aim in this book is to help shape that story by explaining, realistically and optimistically, the potential of AI to create prosperity for us all, not just the privileged few. Realizing that potential requires a principled approach to the development of AI and a set of policies designed to democratize AI.
AI must become a platform that any individual or business can use to enhance their creativity and productivity, and that can be used to solve the biggest of the big problems that confront us as a society. In an interview with Semil Shah, venture capitalist Chamath Palihapitiya paraphrased Bill Gates as having said, “A platform is when the economic value of everybody that uses it exceeds the value of the company that creates it.”1 It is not right if the value created by the development of AI is concentrated solely in the hands of a few elite companies and their employees. People should be rewarded for the innovations that they create, but those innovations must create a surplus of value that benefits businesses and workers everywhere, in middle and rural America as well as in coastal, urban innovation centers.
Much of the progress we’ve made building an AI platform, one that can support more people building more ambitious things every year, has been a direct consequence of publicly funded and widely disseminated research. In addition to accessible research and this rich marketplace of ideas, open source software and the Internet have made it possible for more people with more diverse backgrounds and experiences to make things with AI by making both code and data available for developers and researchers. Most recently, with the advent of cloud computing, corporations operating cloud infrastructure businesses—like Alibaba, Amazon, Google, and Microsoft—all have economic incentives to support as many developers and makers as possible in their individual efforts to build their own cloud-powered AI products and services. This is all well and good, but not enough. We must ensure that anyone can participate in the development of this AI platform, and can intelligently engage in critical debates about how the platform evolves and is governed.
One of the best-known memes in science fiction—a genre no one will be surprised that I love—is based on the 1950 short story by Damon Knight, “To Serve Man.” Popularized by an episode of The Twilight Zone, and then referenced dozens of times over the years, including in the first “Treehouse of Horror” episode of The Simpsons, the story is a play on the dual meaning of the verb to serve. In the story, aliens arrive on Earth and provide humanity with unlimited energy, food, a mechanism to disable our weapons of war, and a promise of medicines that would prolong life. One of the human protagonists of the story, a character named Grigori, is skeptical of the motives of the aliens. He learns their language and steals one of their books, To Serve Man, which it turns out is not a treatise on service to humanity, but a cookbook.
In our push to advance the state of the art for AI, to build new products, automate processes, and found whole new businesses empowered by AI technologies, we need to be constantly vigilant that all of this effort is focused on serving humans . . . in the beneficial sense of the word. The good news is that AI is a tool that we humans have developed and control. All of us, from the developer and scientist, to the entrepreneur and business exec, to the educator and the policy maker, have permission to make choices about the development of AI only to the extent that those choices are human-centered and serve the public good on balance. And balance is key even though achieving it is nontrivial, given the complexity of how these technologies intersect with society and our daily lives.
Finally, although AI already is, and will likely continue to be, one of the most beneficial technologies that humans have ever developed, we must be realistic that its development will come with negative consequences, as has been the case with many of history’s complex technological revolutions. We must work to prevent as many of those negative consequences as possible, and when they arise, to have compassion for those impacted, and to work as hard as we can to mitigate those impacts as quickly as possible.
If doctors take an oath to do no harm, so should software engineers. Software engineering is what I do. It’s how I think and see the world. Today, as Microsoft’s chief technology officer, a father, and a philanthropist, I also think about how that technology will affect people, communities, and society. The one-two punch of ever bigger data and compute power is fueling an artificial intelligence revolution that will make prior industrial revolutions appear piddling in comparison. The AI Revolution will radically impact economics and employment for everyone for generations to come. I am not alone in foreseeing a massive transformation in the nature of work caused by the very machines I’ve helped to engineer. The bookshelf of science fiction writers, philosophers, economists, and technologists is sagging beneath the weight of doomsayers. My perspective, steeled by an upbringing in a region of the South decimated by the economic tsunami of dying tobacco, textile, and furniture industries, is forged, however, by a sense of tenacity and optimism.
The question this book asks is this: Can business and government work collaboratively to program next-generation technologies and future public policies to keep the American dream alive? In today’s atmosphere that question sounds naive or cynical. But what could be more important? Like public health, climate change, and public education, we need international understanding and collaboration on the future of AI and work. This means intellectual inquiry, rational debate, and coming together, not sound bites, sensationalism, and the politics of division. Just as some believe it’s too late to reverse the effects of global warming, some observers and even some technologists worry that AI may be on some unalterable path to taking both our work and our dignity. I’m skeptical of that mind-set, but the evidence I see and read—advances in machine learning algorithms, an increasingly complete and robust digital map of the world and human knowledge for training machine learning systems, and rapid expansion of training compute—suggests that the future we are hurtling toward will be very different from the present.
At the far end of the spectrum of concerns about the future of artificial intelligence are theories about what strong AI or artificial general intelligence (AGI) means for us as humans. Most of the AI that we encounter in our daily lives are systems built with the techniques of machine learning that have been trained by human beings to perform very specific, narrow tasks. These tasks might be selecting the ads you see when browsing the Internet, or turning the words you utter into your phone or PC into a text transcript that can then be processed by some other system, or identifying friends and family in the photos that you upload so that they can be tagged and notified. AGI—which no one really knows when we will reach—would be much more humanlike in its capabilities. Some, very worried about what AGI could mean for humanity, imagine futures like those from the Terminator series, where an AGI could become truly autonomous and hostile to humans. Critics of those fears assert that AI is far from passing even weak tests for sapience like the Turing test2 or Winograd schema challenge,3 and that human-level AI is some unknowable way off in the future, maybe never to be achieved.
Even though I have neither the expertise nor the crystal ball to predict exactly when AGI might arrive, I’ve been involved with modern technology long enough, and read enough history, to know that we’ve often underestimated the speed with which futuristic technology suddenly arrives. AI has historically been limited in what it has been able to accomplish by the amount of compute power we can throw at AI problems, and how much time it takes for humans to encode logic and knowledge into AI algorithms. We now have enormous amounts of compute power in the cloud. And we have enormous databases of digitized human knowledge like YouTube and the Kindle bookstore that can be used to train AI systems. As our modern AI algorithms absorb that human intelligence to accomplish new tasks we imagine for AI-powered systems, we may achieve what Thomas Kuhn defined as a paradigm shift, one in which humans will either be in the loop or out. Which one of these options we reach depends on our actions today, the story we craft, and the principles we assert about what kind of world we want our children to live in tomorrow.
I feel this profound sense of cognitive dissonance: the same thing that can advance humanity can also cause people distress and even harm. This book arises out of a powerful urge I feel to reconcile the two. It is an engineer’s tale, not the musings of a philosopher, economist, or screenwriter. As Microsoft’s chief technologist, do I have skin in the game? Of course I do. But I’m also the product of rural America, one of the places most vulnerable in the dystopic story of AI. My values and many of my earliest experiences as a budding engineer occurred in a part of America, rural America, that is most at risk. I left the rural South over two decades ago, first for academia and then for Silicon Valley and the tech industry. But at my core I am those people—rural people—and I care about creating a future that values them and their resourcefulness.
I’m no apologist for AI. I am, however, an optimist. The AI future I see needs high-skilled, middle-skilled, and even low-skilled workers involved in building and operating intelligent systems. I’m not alone in this view. Indeed, Microsoft announced in 2018 that it would invest substantially in rural communities to increase local digital skills. Google also announced it would train ten million Nigerians to build digital skills. And start-ups in China and Silicon Valley like BasicFinder, Mada Code, and Scale employ tens of thousands of people to teach machine learning systems how to perform specific tasks. Conversely, Amazon discovered the consequences of not having a skilled workforce when it announced in 2017 it would open an office in East Palo Alto, California, in the heart of Silicon Valley, only to learn that “there aren’t enough people with the required skills,” according to Forbes contributor Bernard Marr.
The story I aim to tell is one of a hopeful future where AI benefits us all equitably, and one where the obstacles standing in the way of this future are not intractable. There’s a role for all of us to play in realizing this hopeful future.
For the AI experts it means really thinking through the consequences of the choices you make on an almost daily basis, taking action to avoid bad things like biased models or AI systems that empower or amplify the worst of our human tendencies, and ensuring that your work is having the net positive impact that you intend for your fellow humans. I know from personal experience that it’s all too easy to allow yourself to be completely immersed in the details of the incredibly complex and narrowly focused thing on which you’re working in AI. In fact, it’s almost impossible to do your job if you can’t do that! But no matter how absorbing the complexity of your daily work is, understanding your work in context, and thinking through the consequences of the decisions you make, are also parts of your job.
For product developers and entrepreneurs, your role means ensuring that the things you’re powering with AI are genuinely beneficial for as broad a swath of humanity as possible; that you are seeking collaborators with a diversity of backgrounds, experiences, and expertise; and that you embrace your responsibility to help mitigate the disruption that you create, whether the disruption was intentional or not.
For investors and allocators of capital, your role means thinking about your AI investments as a Bill Gates thinks about platforms: that they create more economic benefit for others than for you or your enterprise or your portfolio companies. Given the tendency of early-stage disruptive technologies like AI to benefit the owners of expertise and capital disproportionately, this is incredibly important, and given how geographically concentrated AI development is, we investors have a real duty to make sure that we are encouraging diverse and geographically equitable investment.
For policy makers, your role means implementing policies that not only prevent or discourage the negative consequences of bad AI, but more important, incents and accelerates the benefits of good AI. In the United States, we have perhaps the best track record in modern history of government policy making that facilitates disruptive technology rollout for the public good. The entire aerospace industry is a great example of where a judicious blend of direct government funding, incentives, and regulation has facilitated an entire ecosystem of technologies without which modern life would be impossible.
Interestingly, China is now taking pages from the US playbook to bootstrap its own AI industry. The “Three-Year Action Plan for Promoting Development of a New Generation Artificial Intelligence Industry,”4 adopted at the 19th National Congress of the Communist Party of China in 2017, articulates a very granular strategy for how to build an AI industry in China, asserting:
We should seize the historical opportunity, make breakthroughs in key areas, promote the development of the AI industry, enhance the intelligent manufacturing industry, and promote the in-depth integration of AI and real economy.
In addition to developing a similarly detailed public policy road map for promoting the AI industry in the United States, we may also want to consider the role that the voice of the United States plays in the development of AI policy globally. AI was literally invented in the United States, and since its inception, US research institutions and companies have been among the leaders of its development. We are now at a moment where we either need to choose to lead in AI policy in a way that promotes both our future prosperity and democratic values, or allow the definitive policies for AI to be written elsewhere.
Perhaps the most important role to be played in securing a hopeful AI future for all of us is that of the individual citizen. No matter who you are, what your job is, where you live, or where you are in your life’s journey, you need to understand AI’s opportunities and pitfalls so that you can help make the right choices about what to study for future generations. You need to understand what AI means for the future of work so that you can prepare yourself for future job opportunities, to know where opportunities might exist for the company you want to start, or even to know what’s a good bet for a retirement investment. The public policy debate around AI is going to be complex, and it’s going to demand that citizens of democratically elected governments wrap their heads around nontrivial issues and make tough trade-offs. You won’t be able to have an effective voice in these debates unless you are at least a little bit informed, and you won’t be able to say that you’re truly informed unless you are able to look across a bunch of mutually conflicting opinions and points of view and rationally synthesize your own, intelligent opinion.
Even though I experienced poverty and massive loss of industry in the region where I grew up, fulfillment of an American dream was somehow a hope that my friends and I all shared. But for the American dream to be a realistic goal today for everyone, we need more than just belief in it. We need to cultivate it. In the words of a software engineer, we need to reprogram it. This book is my attempt to contribute to that goal, and to show how intelligent technologies, done right, can make it more possible for everyone to attain the full measure of his or her innate capabilities.
To show you the future, I invite you to take a journey with me back to my rural origins in Virginia.