For most of my adult life, I have been maniacally focused on my work. I would answer emails instantly during the day, and even get up twice each night to ensure that all the emails were answered. Yes, I would spend time with my family members—but just so they didn’t complain, and not an hour more.
Then in September 2013, I was diagnosed with fourth-stage lymphoma. I faced the real possibility that my remaining time on Earth would be measured in months. As terrifying as that was, one of my strongest feelings was an instant, irretrievable, and painful regret. As Bronnie Ware’s book about regrets of people on their deathbeds all too accurately describes, I was wracked with remorse over not spending more time sharing love with the people I cared about most.
I am now in remission, so I can write this piece. I am spending much more time with my family. I moved closer to my mother. Whether on business or for pleasure, I travel with my wife. Formerly, when my grown kids came home, I would take two or three days off from work to see them. Now I take two or three weeks. I spend weekends traveling with my best friends. I took my company on a one-week vacation to Silicon Valley—their Mecca. I meet with young people who send me questions on Facebook. I have reached out to people I offended years ago and asked for their forgiveness and friendship.
This near-death experience has not only changed my life and priorities, but also altered my view of artificial intelligence—the field that captured my selfish attention for all those years. This personal reformation gave me an enlightened view of what AI should mean for humanity. Many of the recent discussions about AI have concluded that this scientific advance will likely take over the world, dominate humans, and end poorly for mankind. But my near-death experience has enabled me to envision an alternate ending to the AI story—one that makes the most of this amazing technology while empowering humans not just to survive, but to thrive.
My catharsis came at a point when we were losing perspective on AI. For much of my career, the great accomplishments of this scientific pursuit always seemed to be five years away. But recently they have been cascading one after another, most strikingly with AlphaGo’s victory in 2016. There is a feeling that HAL, the stubborn and deadly computer in 2001: A Space Odyssey, is looming at the gates, and a form of near-panic has set in. We are bombarded with dire predictions by a number of self-appointed futurists about “superintelligence,” “singularity,” “cyborgs,” and the unprovable claim that “we live in a video game.” These dystopian warnings are infectious, because they come from famous people—and perhaps because they are reinforced by the familiar plots of science fiction.
As someone who has worked on AI for 37 years, I assure you that there exists no engineering basis for these outlandish predictions. Science fiction is all fiction, and very little science, and it would be catastrophic for mankind to capitulate to these imaginative but irresponsible predictions.
What’s more, the real AI story is itself as fascinating as any novel—and indeed, it has its dark side. The excitement behind AI today is largely due to a 2010 invention called “deep learning,” which uses massive amounts of data to optimize decision engines with superhuman accuracy. Given a massive amount of data in a particular domain, deep learning can be used to optimize single objective functions, such as “win Go,” “minimize default rate,” or “maximize speech recognition accuracy.”
The results have been spectacular. Armed with deep learning and other machine-learning technologies, AI has proven capable of matching or surpassing some of the most impressive human feats of intelligence. It has vanquished human world champions in Go and poker, and is already superior than the average person in recognizing faces, videos, or words from speech. Critical mobile and internet applications, such as search ranking, e-commerce recommendation, and speech agents like Siri and Alexa, aren’t even imaginable without AI.
Naturally, businesses are using AI to automate tasks that humans used to perform. These include chatbots for customer service, loan officers for approving loans, and security guards for checking IDs. For example, my team invested in a company called Smart Finance, which built an app that uses an AI as a loan officer. Initially, this company lost money due to a high rate of bad loans—but the AI learning kicked in, and with enough data accumulated, the bad loan rate dropped dramatically. It can now make a loan decision in seconds, with higher accuracy than a loan officer who takes hours. And it is infinitely scalable: This company will underwrite about 30 million loans this year, more than any bank that I know of. All of this happened in under two years.
This is clearly threatening news for loan officers. The core functions of other jobs—such as tellers, tele-salespeople, paralegals, reporters, stock traders, research analysts, and radiologists—will gradually be replaced by AI software. And as robotics evolve, including semi-autonomous and autonomous hardware, AI will perform the labor of factory workers, construction workers, drivers, delivery people, and many others.
The AI revolution is on the scale of the Industrial Revolution—probably larger and definitely faster. But while robots may take over jobs, believe me when I tell you there is no danger that they will take over. These AIs run “narrow” applications that master a single domain each time, but remain strictly under human control. The necessary ingredient of dystopia is “General AI”—AI that by itself learns common sense reasoning, creativity, and planning, and that has self-awareness, feelings, and desires. This is the stuff of the singularity that the Cassandras predict. But General AI isn’t here. There are simply no known engineering algorithms for it. And I don’t expect to see them any time soon. The “singularity” hypothesis extrapolates exponential growth from the recent boom, but ignores the fact that continued exponential growth requires scientific breakthroughs that are unlikely to be solved for a hundred years, if ever.
So based on these engineering realities, instead of discussing this fictional super-intelligence, we should focus on the very real “narrow” AI applications and extensions. These will proliferate quickly, leading to massive value creation and an Age of Plenty, because AI will produce fortunes, make strides to eradicate poverty and hunger, and give all of us more spare time and freedom to do what we love. But it will also usher in an Age of Confusion. As an Oxford study postulates, AI will replace half of human jobs, and many people will become depressed as they lose their jobs and the purpose that comes with gainful employment.
It is imperative that we focus on the certainty of these serious issues, rather than talking about dystopia, singularity, or super-intelligence. Perhaps the most vexing question is: How do we create enough jobs to place these displaced workers? The answer to this question will determine whether the alternate ending to the AI story will be happy or tragic.
One suggested solution is to try to move people to jobs that are a step or two ahead of what machines can do. The idea would be to transition people to jobs that require higher dexterity (e.g., retrain an assembly line worker to be a plumber), hidden talent (e.g., encourage an accountant to pursue her dream of becoming a comedian), or new skills (e.g., train a cooling expert for a giant AI data center). Of course we should try this, but these numbers would be infinitesimal compared to the number of jobs displaced. And it is only the rarest accountant who can kill it at the Comedy Cellar.
There are other optimists who try to “hand-wave” the problem away by saying that new jobs have been created with every technological revolution, so we should “have faith.” These modern Panglosses often cite the Industrial Revolution, the office revolution (typewriters, calculators, mimeograph machines, etc.), and the computer revolution as examples. As a well-known 2013 Oxford study by Carl Frey and Michael Osborne has shown, each of the previous revolutions created some jobs (such as assembly line workers) even as they destroyed others (trained hand-craftsmen). But in the upcoming AI revolution, when AI replaces humans for a “task” it often does so completely, without creating new jobs or tasks. So, we cannot expect AI to solve our employment problem. We must solve it for ourselves.
The answer I propose would never have come to me when I was myself somewhat of an automaton, living to work rather than the other way around. It was only my cancer diagnosis, and the sudden realization of what my own stupidity had made me miss, that led me to my suggestion. Our coexistence with artificial intelligence hinges on combining what is humanly unattainable—the hugely scaled narrow AI intelligence that will only get better at any given domain—with what we humans can uniquely offer to one another. And that is love. What makes us human is that we can love.
We are far from understanding the human “heart,” let alone replicating it. But we do know that humans are uniquely able to love and be loved. The moment when we see our newborn babies; the feeling of love at first sight; the warm feeling from friends who listen to us empathetically; the feeling of self-actualization when we help someone in need. Loving and being loved are what makes our lives worthwhile.
Love is what will always differentiate us from AI. Narrow AI has no self awareness, emotions, or a “heart.” Narrow AI has no sense of beauty, fun, or humor. It doesn’t even have feelings or self-consciousness. Can you imagine the ecstasy that comes from beating a world champion? AlphaGo bested the globe’s best player, but took no pleasure in the game, felt no happiness from winning, and had no desire to hug a loved one after its victory.
Despite what science fiction movies may portray, I can tell you responsibly that AI programs cannot love. Scarlett Johansson may have been able to convince you otherwise—because she is an actress who drew on her knowledge of love.
Imagine a situation in which you informed a smart machine that you were going to pull its plug, and then changed your mind and gave it a reprieve. The machine would not change its outlook on life or vow to spend more time with its fellow machines. It would not grow, as I did, or serve others more generously.
Love is what is missing from machines.
That’s why we must pair up with them, to leaven their powers with what only we humans can provide. Your future AI diagnostic tool may well be 10 times more accurate than human doctors, but patients will not want a cold pronouncement from the tool: “You have fourth stage lymphoma and a 70 percent likelihood of dying within five years.” That in itself would be harmful. Patients would benefit, in health and heart, from a “doctor of love” who will spend as much time as the patient needs, always be available to discuss their case, and who will even visit the patients at home. This doctor might encourage us by sharing stories such as, “Kai-Fu had the same lymphoma, and he survived, so you can too.” This kind of “doctor of love” would not only make us feel better and give us greater confidence, but would also trigger a placebo effect that would increase our likelihood of recuperation. Meanwhile, the AI tool would watch the Q&A between the “doctor of love” and the patient carefully, and then optimize the treatment. If scaled across the world, the number of “doctors of love” would greatly outnumber today’s doctors.
The same idea could apply to lawyers, teachers, accountants, and wedding planners. In innumerable instances, excellent AI tools may emerge, but the “human-to-human” interface is critical to ensuring we feel listened to and cared for when we encounter important life events. We should encourage more people to go into service careers, choosing the ones into which they can pour their hearts and souls, spreading their love and experiences—whether as a passionate tour guide, an attentive concierge, a funny bartender, an infectious hair dresser, or an innovative sushi chef.
We should also work hard to invent new service jobs that deliver joy and love. Imagine a nutritional chef who comes to your home to cook only with fresh, organic, local ingredients. Or perhaps the “season changer” who changes and redecorates your closets seasonally, with flowers and aromas that make changing clothes a fun experience. Or perhaps an “elderly companion” who takes your aging parents to see a "doctor of love" when you cannot.
There will also be a big demand for social workers who answer the hotlines for displaced workers, dealing with their depression and anxiety. Volunteering service jobs today may turn into real jobs of the future—that of assisting at a blood bank, teaching at an orphanage, mentoring at Scouts organizations, or being a sponsor at AA or the Veterans Recruitment Appointment. Each of these jobs will deliver love and empathy—and there will be so many that we can replace many, if not all, of that 50 percent loss that comes from automation. Most importantly, the people filling these new jobs will fill our planet with love and joy.
So, this is the alternate ending to the narrative of AI dystopia. An ending in which AI performs the bulk of repetitive jobs, but the gap is filled by opportunities that require our humanity.
Can I guarantee that scientists in the future will never make the breakthroughs that will lead to the kind of general-intelligence computer capabilities that might truly threaten us? Not absolutely. But I think that the real danger is not that such a scenario will happen, but that we won’t embrace the option to double down on humanity while also using AI to improve our lives. This decision is ultimately up to us: Whatever we choose may become a self-fulfilling prophecy. If we choose a world in which we are fully replaceable by machines, whether it happens or not, we are surrendering our humanity and our pursuit for meaning. If everyone capitulates, our humanity will come to an end.
Such a capitulation is not only premature and unproven, but also irresponsible to our legacy, our ancestors, and our maker. On the other hand, if we choose to pursue our humanity, and even if the improbable happen and machines truly replace us, we can then capitulate knowing that we did the responsible thing, and that we had fun doing it. We will have no regrets over how we lived.
I do not think the day will ever come—unless we foolishly make it happen ourselves. Let us choose to let machines be machines, and let humans be humans. Let us choose to use our machines, and love one another.