The word television comes from Greek ‘tele’, which means ‘far’, and Latin ‘visio’, sight. It was originally conceived as a device that allows us to see from afar. But soon, it might allow us to be seen from afar. As George Orwell envisioned in Nineteen Eighty-Four, the television will watch us while we are watching it. After we’ve finished watching Tarantino’s entire filmography, we may have forgotten most of it. But Netflix, or Amazon, or whoever owns the TV algorithm, will know our personality type, and how to press our emotional buttons. Such data could enable Netflix and Amazon to choose movies for us with uncanny precision, but it could also enable them to make for us the most important decisions in life – such as what to study, where to work, and who to marry.
Of course Amazon won’t be correct all the time. That’s impossible. Algorithms will repeatedly make mistakes due to insufficient data, faulty programming, muddled goal definitions and the chaotic nature of life.10 But Amazon won’t have to be perfect. It will just need to be better on average than us humans. And that is not so difficult, because most people don’t know themselves very well, and most people often make terrible mistakes in the most important decisions of their lives. Even more than algorithms, humans suffer from insufficient data, from faulty programming (genetic and cultural), from muddled definitions, and from the chaos of life.
You may well list the many problems that beset algorithms, and conclude that people will never trust them. But this is a bit like cataloguing all the drawbacks of democracy and concluding that no sane person would ever choose to support such a system. Winston Churchill famously said that democracy is the worst political system in the world, except for all the others. Rightly or wrongly, people might reach the same conclusions about Big Data algorithms: they have lots of hitches, but we have no better alternative.
As scientists gain a deeper understanding of the way humans make decisions, the temptation to rely on algorithms is likely to increase. Hacking human decision-making will not only make Big Data algorithms more reliable, it will simultaneously make human feelings less reliable. As governments and corporations succeed in hacking the human operating system, we will be exposed to a barrage of precision-guided manipulation, advertisement and propaganda. It might become so easy to manipulate our opinions and emotions that we will be forced to rely on algorithms in the same way that a pilot suffering an attack of vertigo must ignore what his own senses are telling him and put all his trust in the machinery.
In some countries and in some situations, people might not be given any choice, and they will be forced to obey the decisions of Big Data algorithms. Yet even in allegedly free societies, algorithms might gain authority because we will learn from experience to trust them on more and more issues, and will gradually lose our ability to make decisions for ourselves. Just think of the way that within a mere two decades, billions of people have come to entrust the Google search algorithm with one of the most important tasks of all: searching for relevant and trustworthy information. We no longer search for information. Instead, we google. And as we increasingly rely on Google for answers, so our ability to search for information by ourselves diminishes. Already today, ‘truth’ is defined by the top results of the Google search.11
This has also been happening with physical abilities, such as navigating space. People ask Google to guide them around. When they reach an intersection, their gut feeling might tell them ‘turn left’, but Google Maps says ‘turn right’. At first they listen to their gut feeling, turn left, get stuck in a traffic jam, and miss an important meeting. Next time they listen to Google, turn right, and make it on time. They learn from experience to trust Google. Within a year or two, they blindly rely on whatever Google Maps tells them, and if the smartphone fails, they are completely clueless. In March 2012 three Japanese tourists in Australia decided to take a day trip to a small offshore island, and drove their car straight into the Pacific Ocean. The driver, twenty-one-year-old Yuzu Nuda, later said that she just followed the instructions of the GPS and ‘it told us we could drive down there. It kept saying it would navigate us to a road. We got stuck.’12 In several similar incidents people drove into a lake, or fell off a demolished bridge, by apparently following GPS instructions.13 The ability to navigate is like a muscle – use it or lose it.14 The same is true for the ability to choose spouses or professions.
Every year millions of youngsters need to decide what to study at university. This is a very important and very difficult decision. You are under pressure from your parents, your friends and your teachers, who have different interests and opinions. You also have your own fears and fantasies to deal with. Your judgement is clouded and manipulated by Hollywood blockbusters, trashy novels, and sophisticated advertising campaigns. It is particularly difficult to make a wise decision because you do not really know what it takes to succeed in different professions, and you don’t necessarily have a realistic image of your own strengths and weaknesses. What does it take to succeed as a lawyer? How do I perform under pressure? Am I a good team-worker?
One student might start law school because she has an inaccurate image of her own skills, and an even more distorted view of what being a lawyer actually involves (you don’t get to give dramatic speeches and shout ‘Objection, Your Honour!’ all day). Meanwhile her friend decides to fulfil a childhood dream and study professional ballet dancing, even though she doesn’t have the necessary bone structure or discipline. Years later, both deeply regret their choices. In the future we could rely on Google to make such decisions for us. Google could tell me that I would be wasting my time in law school or in ballet school – but that I might make an excellent (and very happy) psychologist or plumber.15
Once AI makes better decisions than us about careers and perhaps even relationships, our concept of humanity and of life will have to change. Humans are used to thinking about life as a drama of decision-making. Liberal democracy and free-market capitalism see the individual as an autonomous agent constantly making choices about the world. Works of art – be they Shakespeare plays, Jane Austen novels, or tacky Hollywood comedies – usually revolve around the hero having to make some particularly crucial decision. To be or not to be? To listen to my wife and kill King Duncan, or listen to my conscience and spare him? To marry Mr Collins or Mr Darcy? Christian and Muslim theology similarly focus on the drama of decision-making, arguing that everlasting salvation or damnation depends on making the right choice.
What will happen to this view of life as we increasingly rely on AI to make decisions for us? At present we trust Netflix to recommend movies, and Google Maps to choose whether to turn right or left. But once we begin to count on AI to decide what to study, where to work, and who to marry, human life will cease to be a drama of decision-making. Democratic elections and free markets will make little sense. So would most religions and works of art. Imagine Anna Karenina taking out her smartphone and asking the Facebook algorithm whether she should stay married to Karenin or elope with the dashing Count Vronsky. Or imagine your favourite Shakespeare play with all the crucial decisions taken by the Google algorithm. Hamlet and Macbeth will have much more comfortable lives, but what kind of life will it be exactly? Do we have models for making sense of such a life?
As authority shifts from humans to algorithms, we may no longer see the world as the playground of autonomous individuals struggling to make the right choices. Instead, we might perceive the entire universe as a flow of data, see organisms as little more than biochemical algorithms, and believe that humanity’s cosmic vocation is to create an all-encompassing data-processing system – and then merge into it. Already today we are becoming tiny chips inside a giant data-processing system that nobody really understands. Every day I absorb countless data bits through emails, tweets and articles; process the data; and transmit back new bits through more emails, tweets and articles. I don’t really know where I fit into the great scheme of things, and how my bits of data connect with the bits produced by billions of other humans and computers. I don’t
have time to find out, because I am too busy answering all these emails.
The philosophical car
People might object that algorithms could never make important decisions for us, because important decisions usually involve an ethical dimension, and algorithms don’t understand ethics. Yet there is no reason to assume that algorithms won’t be able to outperform the average human even in ethics. Already today, as devices like smartphones and autonomous vehicles undertake decisions that used to be a human monopoly, they start to grapple with the same kind of ethical problems that have bedevilled humans for millennia.
For example, suppose two kids chasing a ball jump right in front of a self-driving car. Based on its lightning calculations, the algorithm driving the car concludes that the only way to avoid hitting the two kids is to swerve into the opposite lane, and risk colliding with an oncoming truck. The algorithm calculates that in such a case there is a 70 per cent chance that the owner of the car – who is fast asleep in the back seat – would be killed. What should the algorithm do?16
Philosophers have been arguing about such ‘trolley problems’ for millennia (they are called ‘trolley problems’ because the textbook examples in modern philosophical debates refer to a runaway trolley car racing down a railway track, rather than to a self-driving car).17 Up till now, these arguments have had embarrassingly little impact on actual behaviour, because in times of crisis humans all too often forget about their philosophical views and follow their emotions and gut instincts instead.
One of the nastiest experiments in the history of the social sciences was conducted in December 1970 on a group of students at the Princeton Theological Seminary, who were training to become ministers in the Presbyterian Church. Each student was asked to hurry to a distant lecture hall, and there give a talk on the Good Samaritan parable, which tells how a Jew travelling from Jerusalem to Jericho was robbed and beaten by criminals, who then left him to die by the side of the road. After some time a priest and a Levite passed nearby, but both ignored the man. In contrast, a Samaritan – a member of a sect much despised by the Jews – stopped when he saw the victim, took care of him, and saved his life. The moral of the parable is that people’s merit should be judged by their actual behaviour, rather than by their religious affiliaton.
The eager young seminarians rushed to the lecture hall, contemplating on the way how best to explain the moral of the Good Samaritan parable. But the experimenters planted in their path a shabbily dressed person, who was sitting slumped in a doorway with his head down and his eyes closed. As each unsuspecting seminarian was hurrying past, the ‘victim’ coughed and groaned pitifully. Most seminarians did not even stop to enquire what was wrong with the man, let alone offer any help. The emotional stress created by the need to hurry to the lecture hall trumped their moral obligation to help strangers in distress.18
Human emotions trump philosophical theories in countless other situations. This makes the ethical and philosophical history of the world a rather depressing tale of wonderful ideals and less than ideal behaviour. How many Christians actually turn the other cheek, how many Buddhists actually rise above egoistic obsessions, and how many Jews actually love their neighbours as themselves? That’s just the way natural selection has shaped Homo sapiens. Like all mammals, Homo sapiens uses emotions to quickly make life and death decisions. We have inherited our anger, our fear and our lust from millions of ancestors, all of whom passed the most rigorous quality control tests of natural selection.
Unfortunately, what was good for survival and reproduction in the African savannah a million years ago does not necessarily make for responsible behaviour on twenty-first-century motorways. Distracted, angry and anxious human drivers kill more than a million people in traffic accidents every year. We can send all our philosophers, prophets and priests to preach ethics to these drivers – but on the road, mammalian emotions and savannah instincts will still take over. Consequently, seminarians in a rush will ignore people in distress, and drivers in a crisis will run over hapless pedestrians.
This disjunction between the seminary and the road is one of the biggest practical problems in ethics. Immanuel Kant, John Stuart Mill and John Rawls can sit in some cosy university hall and discuss theoretical problems in ethics for days – but would their conclusions actually be implemented by stressed-out drivers caught in a split-second emergency? Perhaps Michael Schumacher – the Formula One champion who is sometimes hailed as the best driver in history – had the ability to think about philosophy while racing a car; but most of us aren’t Schumacher.
Computer algorithms, however, have not been shaped by natural selection, and they have neither emotions nor gut instincts. Hence in moments of crisis they could follow ethical guidelines much better than humans – provided we find a way to code ethics in precise numbers and statistics. If we teach Kant, Mill and Rawls to write code, they can carefully program the self-driving car in their cosy laboratory, and be certain that the car will follow their commandments on the highway. In effect, every car will be driven by Michael Schumacher and Immanuel Kant rolled into one.
Thus if you program a self-driving car to stop and help strangers in distress, it will do so come hell or high water (unless, of course, you insert an exception clause for infernal or high-water scenarios). Similarly, if your self-driving car is programmed to swerve to the opposite lane in order to save the two kids in its path, you can bet your life this is exactly what it will do. Which means that when designing their self-driving car, Toyota or Tesla will be transforming a theoretical problem in the philosophy of ethics into a practical problem of engineering.
Granted, the philosophical algorithms will never be perfect. Mistakes will still happen, resulting in injuries, deaths and extremely complicated lawsuits. (For the first time in history, you might be able to sue a philosopher for the unfortunate results of his or her theories, because for the first time in history you could prove a direct causal link between philosophical ideas and real-life events.) However, in order to take over from human drivers, the algorithms won’t have to be perfect. They will just have to be better than the humans. Given that human drivers kill more than a million people each year, that isn’t such a tall order. When all is said and done, would you rather the car next to you was driven by a drunk teenager, or by the Schumacher–Kant team?19
The same logic is true not just of driving, but of many other situations. Take for example job applications. In the twenty-first century, the decision whether to hire somebody for a job will increasingly be made by algorithms. We cannot rely on the machine to set the relevant ethical standards – humans will still need to do that. But once we decide on an ethical standard in the job market – that it is wrong to discriminate against black people or against women, for example – we can rely on machines to implement and maintain this standard better than humans.20
A human manager may know and even agree that it is unethical to discriminate against black people and women, but then, when a black woman applies for a job, the manager subconsciously discriminates against her, and decides not to hire her. If we allow a computer to evaluate job applications, and program the computer to completely ignore race and gender, we can be certain that the computer will indeed ignore these factors, because computers don’t have a subconscious. Of course, it won’t be easy to write code for evaluating job applications, and there is always a danger that the engineers will somehow program their own subconscious biases into the software.21 Yet once we discover such mistakes, it would probably be far easier to debug the software than to rid humans of their racist and misogynist biases.
We saw that the rise of artificial intelligence might push most humans out of the job market – including drivers and traffic police (when rowdy humans are replaced by obedient algorithms, traffic police will be redundant). However, there might be some new openings for philosophers, because their skills – hitherto devoid of much market value – will suddenly be in very high demand. So if you want to study something that will guarant
ee a good job in the future, maybe philosophy is not such a bad gamble.
Of course, philosophers seldom agree on the right course of action. Few ‘trolley problems’ have been solved to the satisfaction of all philosophers, and consequentialist thinkers such as John Stuart Mill (who judge actions by consequences) hold quite different opinions to deontologists such as Immanuel Kant (who judge actions by absolute rules). Would Tesla have to actually take a stance on such knotty matters in order to produce a car?
Well, maybe Tesla will just leave it to the market. Tesla will produce two models of the self-driving car: the Tesla Altruist and the Tesla Egoist. In an emergency, the Altruist sacrifices its owner to the greater good, whereas the Egoist does everything in its power to save its owner, even if it means killing the two kids. Customers will then be able to buy the car that best fits their favourite philosophical view. If more people buy the Tesla Egoist, you won’t be able to blame Tesla for that. After all, the customer is always right.
This is not a joke. In a pioneering 2015 study people were presented with a hypothetical scenario of a self-driving car about to run over several pedestrians. Most said that in such a case the car should save the pedestrians even at the price of killing its owner. When they were then asked whether they personally would buy a car programmed to sacrifice its owner for the greater good, most said no. For themselves, they would prefer the Tesla Egoist.22