The world economy has had sluggish growth for years. Cities are congested like never before. Man-made fake news pollutes citizens’ thinking. And our political systems seem to crack under pressure. Are people no longer able to solve the most complex problems themselves? We may find out soon enough, because the moral machine is about to become reality.
This year, a quantum computer will be among the heads of state rubbing shoulders in the Swiss town of Davos. A group of neuroscientists and technology pioneers will present the “moral machine,” one able to digest petabytes of information and carry out trillions of calculations per second; it is designed to autonomously lead complex systems—from businesses to entire cities.
Some of us questioning our capacity to carry the responsibility that comes with democratic rule these days might sympathize with this idea. To those we must apologize: decision-makers will still show up in human form to Davos in 2017. The “moral machine” won’t be real, but it will be the centerpiece of a design exhibition where artists and scientists explore the power that algorithms may soon wield over our lives. It asks a provocative question: What if we were led by machines that know us better than we do?
Visions of computers outperforming and outbehaving (or outmaneuvering) us have always triggered human imagination. But this time is different. First, we seem to be on the cusp of exponential change—the “knee of the curve”—in the shift from narrow artificial intelligence (think an algorithm beating a chess master) to general artificial intelligence (algorithms performing any intellectual task a human can). Secondly, there is a growing feeling that our traditional governance systems neither are responsive nor responsible, unable to cope with the political, economic, and societal challenges of our time. Both shifts could trigger great innovations—and great mistakes.
Moral machines on the march
One of the most profound visions of a “moral machine” was conceived by Isaac Asimov in the 1955 science-fiction essay “Franchise.” It envisions a time when elections have become obsolete as a mysterious computer called “Multivac” takes in all the information there is about economic and political conditions to pick the best candidate. Because the machine cannot carry out value judgments, it hooks up to the brain of a “representative voter” who is asked about the mood in the country. His responses along with the computer’s analysis settle the election. You don’t need to vote—your data does.
Asimov’s story is a work of fiction but was inspired by a real breakthrough: the “Univac.” In the 1952 presidential elections between the Republican Eisenhower and the Democrat Stevenson, Univac was the first computer ever to forecast election results. It was successful in predicting a landslide for Eisenhower, but ironically, as all surveys had expected a tighter race, the engineers did not trust their results and held them back until traditional methods confirmed Univac’s prediction.
Sixty years later, machines still haven’t replaced voters, but they have turned into vital instruments in the hands of analysts. The MIT Media Lab’s Electome applies machine learning to millions of social-media feeds to predict voter concerns. It was used, among others, to prepare this year’s presidential debates. But algorithms are used to sway voters directly, too. So-called social bots—social-media profiles controlled by algorithms—can create the impression of a large followership, making them effective campaigning tools. Bots were reportedly used in the US elections and could also affect upcoming European elections.
As algorithms are getting better and better at understanding our preferences, when will we allow our smartphone to vote? And, if they already are the better campaigners, how far are we from them becoming candidates? After all, thousands of people have thronged to pop concerts of figures like Hatsune Miku—a hologram. And a World Economic Forum survey suggests that machines might soon join corporate boards.
And intelligent machines don’t stop here. Earlier this year, CrimeRadar was launched, an app using machine learning to predict crime rates in Rio de Janeiro. It sounds like a sci-fi thriller, but is not speculative at all. CrimeRadar draws on five years of crime data and 14 million crimes to predict how likely it is to be robbed or shot in a particular street, date, and time. Similar tools are already being used (or soon will be) by police forces in Atlanta, New York, Philadelphia, Seattle, and dozens of other US cities. The promise is to remove human bias from crime investigations; the risk is to encourage racial profiling and singling out people as criminal without justification.
Combine such breakthroughs in the unification of big data and machine intelligence with new sensor technologies—and artificial general intelligence will rapidly enter our physical world, too. The US Department of Transportation is studying how cars communicating with each other could make evacuation procedures more efficient. The promise of this technology is to avoid dense crowding and traffic disturbances; a more extreme vision could involve someone striking a button in order to have cars’ computers take over control of vehicles, shuttling even the most reluctant citizen out of a crisis zone.
If modernity was all about transferring power from cosmic to human systems, the coming age could be about transferring it from human to computer systems. And, thanks to new findings in the domains of evolutionary biology and neuroscience, tomorrow’s “Multivac” may not even need to tap the brain of a “representative voter.” Morality, scientists like EO Wilson or Jonathan Haidt assert, is predominantly biological. It would mark the transition from “general intelligence” to what Oxford professor Nick Bostrom calls “super-intelligence:” “an intellect that is much smarter than the best human brains in practically every field, including social skills.”
It took us a while to trust our machines. We are about to trust them more than each other. Enter the new age of Dataism.
Dataism—or Russian roulette for republics
Silicon Valley has never been short of bold visions. However, even the ones leading the AI revolution are realizing its potentially creepy side effects. To address such concerns, some of the most powerful tech companies, including Google, Microsoft, and Facebook, just reassured the public by promising to bring more openness to the development of AI. And president Barack Obama called upon developers to design algorithms in ways that reflect broader values than the ones of their mostly white and male developers.
No doubt, a future where major political decisions are made by machines is disconcerting, but isn’t the human way equally worrying? The US just elected its president with an 18th-century voting system—and the winning candidate’s campaign claims were identified by fact-checkers as false 78% of the time. The UK decided its exit from the EU with a single vote on a rainy Sunday afternoon. A day later the “Leave” campaign gave up its central promises and Google reported that “what is the EU” became the top search term.
“Russian roulette for republics” is how Kenneth Rogoff described the process that led to Brexit. To the Harvard economist, the idea that any decision reached by majority rule is democratic is a “perversion of the term.” “Trump Won Because Voters Are Ignorant,” wrote Jason Brennan, a professor at Georgetown University. (He also acknowledged that if Clinton had won, it would have been thanks to the ignorant, too.)Brennan compares the US election with a final exam for 210 million students where people are told they won’t get their own personal grade but everyone would receive the average grade. As the cost of acquiring and weighing political information greatly exceeds the benefits, we should not be surprised that people show up uninformed at the ballot box.
Social media made matters worse. “This election has shown us how the same platforms that put a world of facts and information at our fingertips can just as easily be used to undermine basic truths,” the usually technology-enthusiastic Wired Magazine concluded the day before the election. Algorithms show us what we like, not what is “right.” As a result, they increase ideological segregation rather than creating a digital agora. Influencers no longer waste their time with facts, but rigorously A/B test behavioral responses to alternative signals. Rather than seeking truth, the age of data is creating its own.
The French philosopher and cultural theorist Paul Virilio famously said “when you invent the plane, you also invent the plane crash.” It is well possible that the 2016 US elections will be remembered as social media’s worst “crash” so far. No doubt, there are construction flaws that need fixing. Only hours after the Trump victory became clear, a group of vice presidents and executives of Facebook reportedly gatheredto analyze what role their company might have played in the election’s outcome.
Yet, making social media the sole culprit for social polarization is just as absurd as blaming a missing wall between Mexico and the US for the woes of America’s middle class. Digital echo chambers don’t create, but accentuate, the social and economic rifts that opened up in market economies. SOAS professor Guy Standing estimates that the deliberate dismantling of social safety nets along with a prolonged period of slow growth made a quarter of the population in advanced economies part of the precariat: a marginalized and potentially angry group, capable of veering to the extremes of the political spectrum.
This is most visible in the United States where stagnating wages hit the middle class hard. In the 1960s, General Motors was not only America’s biggest employer: It was the best-paying, too. Today, the country’s biggest employers are retailers and fast-food chains; almost all are thriving on low pay and low prices. In Europe, according to Bertelsmann’s Social Justice Index, more than 100 million people are threatened by poverty despite having full-time jobs. All the while, the wealthy became wealthier: Since 1988, almost half of global growth went to the world’s richest 5%. Growing inequality in a low-growth context is not the only factor, but the all-infusing backdrop to other national pain points—such as immigration—which played an important role in the US and UK vote.
The flip side of stagnating wages in the West is the growth story of China and other major emerging economies in Asia, Latin America, and Africa. Today, more people die from eating too much rather than too little, from old age rather than epidemics, and suicide rather than war, crime, and terrorism combined. On a global scale, the world is more peaceful, healthy, and equal than ever before. In 1988, a lower US middle class citizen earned more than six times what a well-off Chinese would earn. Today, both earn about the same.
Such facts, of course, won’t help an American politician to promote open markets. On the contrary, the global rebalancing of income and wealth that followed the period of economic expansion ever since the end of the Cold War now adds to the lingering fear of systems decline. A core tenet of Donald Trump’s nationalist appeal—“Make America great again”—is foreign policy: Trump argues that America’s pursuit of global leadership squandered the nation’s resources and benefited America less than its rivals.
We are at an inflection point where parts of the population feel that their leadership neither is responsive to their fears nor responsible in its interventions. While some put their hope into petabytes, others call in the populists. The former are concerned about the biased and shifting mind of the electorate, the latter are furious about the biases of liberal ideologies. In the long run, both might be headed for a remarkably similar fate: a gradual descent into an unvarnished form of authoritarianism.
Preventing complexity from tipping into chaos
At one point in Asimov’s tale, a girl asks her father how it was like before Multivac entered the scene to lead the country. “You see, Linda, till about forty years ago, everybody always voted,” the father said. “How did all the people know who to vote for? Did Multivac tell them?” his daughter asked, perplexed. Her father’s eyebrows hunched down and he looked severe: “They just used their own judgment, girl.”
For complex societies to work, we must consider the calculations of computers and rely on the integrity of experts. But we must not abdicate the work of critical thinking, or our moral and ethical judgment, to charismatic leaders or impressive technologies. Instead, we must resolve the biggest challenge societies are facing: People of affluence and people of modest means, people in cities and the countryside, people young and old, are living increasingly separate lives. Democracy does not require perfect equality, but if we don’t share a common life, its two core principle are at risk: responsiveness and responsibility.
Responsiveness is the precondition for open discourse. Technocratic dataists imply there is one smart solution to every question, hence there is no need for debate. Populist leaders imply there is one popular will, hence there is no need for debate either. One is about facts that don’t need interpretation; the other about interpretations free from facts. Both forget that democracy cannot exist without dialogue and respect for diversity.
Responsibility means caring not only for one’s own grade, but the grade of the class. The moral machine is already here—it is us. Recent votes in the United States and Great Britain have not only been decided by uninformed voters, but also by citizens who decided to forego their responsibility to engage in the political process. According to the latest estimates, almost 100 million people in the US decided not to vote. Disappointment with candidates in both camps played a role, but the reasons for poor voter turnout are more complex, reaching from administrative hurdles to the decay of traditional parties to the political establishment being perceived as too removed from everyday concerns.
Responsiveness and responsibility are the price of our freedom, and freedom is the condition for that those in power—be they men, women, or machines—exercise power in our interest. The digital revolution will fundamentally transform political participation; we are just seeing the beginnings. If these changes will be for the better or worse depends on our commitment to renewing and reaffirming these core principles. After the elections is before the elections. 2017 will be a year of destiny for the world’s democracies.