The rich and powerful have hijacked progress throughout history, says Daron Acemoğlu. They did so back in the Middle Ages and also now in the age of artificial intelligence. In an interview, the MIT economist dives into the question of whether Silicon Valley is plunging humanity into destitution.
DER SPIEGEL: Mr. Acemoğlu, your book reads like an effort to rewrite the history of progress. In it, you take quite a long look back: What are we supposed to learn from the Neolithic in the age of artificial intelligence (AI)?
Acemoğlu: The discussion about AI is in the grips of naïve techno-optimism: AI is going to transform everything, and if there will be rough edges, that’s going to be worked out. When you raise concerns about this narrative, one of the most powerful arguments that people reply with is: Are you claiming that this time is going to be different?
DER SPIEGEL: But it is true that humankind has indeed benefited a lot from new technologies.
Acemoğlu: That is the reason we have to go so far back in history. The argument that you just gave is wrong. In the past, we’ve always had struggles over the uses of innovation and who benefits from them. Very often, control was in the hands of a narrow elite. Innovation often did not benefit the broad swaths of the population.
DER SPIEGEL: Has the standard of living not risen steadily?
Acemoğlu: Today, we are so much more prosperous than the people in earlier ages, that’s true. But there is a tendency to think that the path between must have been a straightforward and inevitable process. We all tend to gloss over the difficulties on the way.
DER SPIEGEL: What do you mean exactly?
Acemoğlu: Take medieval windmills, a very transformative technology. It changed the organization of textile manufacturing, but especially agriculture. But you didn’t see much improvement in the conditions of the peasants. The windmills were controlled by landowners and churches. This narrow elite collected the gains. They decided who could use the windmills. They killed off competition. At the same time, during this period, impressive cathedrals and churches were built all over the continent. Up to 20 percent of the econmic power went into their construction.
DER SPIEGEL: Many people enjoy these buildings today.
Acemoğlu: I find them very impressive as well. But they were amazingly costly, especially for societies that were, at most times, on the brink of subsistence and famines. The cathedrals were built because a small elite wanted to show off their wealth. Their monuments are the equivalents of the Egyptian pyramids. Today, they give us biased impressions of that era. The suffering of the peasants remains invisible.
DER SPIEGEL: What does that have to do with AI?
Acemoğlu: Every other week, tech leaders say: "There is nothing to worry about. AI is going to solve all of our problems." They say the only problems we have to worry about are attempts by governments to regulate AI. The United States has become so naively optimistic about this technology. Unfortunately, many American journalists joined this bandwagon, they became mesmerized with the tech industry. They started believing all sorts of propaganda from the companies.
DER SPIEGEL: Are you saying that AI is doomed to follow the same path of earlier technologies?
Acemoğlu: I am certainly not saying that AI is not a promising technology. I am also optimistic that with the right decisions being made, AI could be amazingly useful to knowledge workers. But again, there is nothing automatic about that. We have to pay much more attention to the direction of technology. We have to think more about the institutional structure in which digital technologies are embedded, or we will fail to create shared prosperity once again.
DER SPIEGEL: Where is your pessimism coming from? Average incomes in the West have never been higher than they are today.
Acemoğlu: In the United States, for 40 years, we have had declines in the real earnings of workers without a college degree. The decline amounts to around a half a percentage point per year. This is an enormous amount. If nothing changes, AI is going to double down on that. AI might still become very useful for well-off citizens, knowledge workers and highly skilled employees. But it is not going to be good for most people on its current path.
DER SPIEGEL: Isn’t that the natural course of technological evolution?
Acemoğlu: When Industrialization took off, wages in the first factories were extremely low for most workers, and living standards deteriorated in Britain. A bit later, when the U.S. started to industrialize – with the same set of technologies – things worked out quite differently. In North America, the kind of skilled work needed to maintain and monitor complex machinery was much scarcer than in Europe. That is when, in the U.S., a different approach to mechanical engineering emerged, with the heavy use of standardized components, and a more modular structure. These machines could also be operated by unskilled workers. This, then, started to increase the productivity of low-skilled workers a lot, as well as their wages. So, it is possible to place people at the center of the technological process rather than at the sidelines as is currently the case.
DER SPIEGEL: You blame this on a vision that prevails in Silicon Valley and that in fact is suppressing other ideas of progress. What do you mean by vision?
Acemoğlu: A vision is an interpretation of how we should push technology and what is the right direction. It is based on a set of beliefs about the world. An example of a vision that has been equally successful and disastrous is the idea of shareholder value, which has been taught in many management schools for the past several decades.
DER SPIEGEL: It states that managers should care about little more than increasing the corporate value of their business.
Acemoğlu: Companies that follow this approach cut wages and, in return, raise returns to shareholders. The advance of shareholder value has been a fundamental ideological shift in the way companies are run. In this logic, employees are seen as mere cost drivers.
DER SPIEGEL: This is not a problem that is exclusive to today’s tech industry.
Acemoğlu: The tech industry combines it with our current obsession with autonomous machine intelligence, meaning that what we should aspire for is to have machines that are as human-like as possible. This vision is rooted in the work and thoughts of Alan Turing, the brilliant British mathematician who first articulated the Turing Test. That's the benchmark that all AI engineers want to pass.
DER SPIEGEL: The test is, roughly speaking, about whether a computer succeeds in deceiving a human to think that it is speaking with another person and not a computer.
Acemoğlu: Exactly. Turing himself called this test an "Imitation Game." This vision of technology is being fueled by Hollywood science fiction. It has now become an overarching theme for all of the tech industry. That’s a problem.
DER SPIEGEL: Has humankind lost its way?
Acemoğlu: At the beginning of the digital age, there were powerful alternative conceptions. The hacker scene and many of the first PC pioneers dreamed of a decentralized technology that would empower employees. At the center of their thinking was the priority to maximize the usefulness of technology for people and societies. This approach became known as "machine usefulness." But ultimately, corporations like IBM, Microsoft and Oracle became the main conduits.
DER SPIEGEL: What could have been different?
Acemoğlu: These companies pushed the technology towards mass application. But this led to (technology) becoming much less useful for the workers. It became a tool under the control of the employers. Companies used it so they could automate a lot of simple office work. In the end, it became a very skill-based technology that improved productivity mostly for skilled managers, university graduates and so on.
DER SPIEGEL: Does this reflect a certain image of humanity?
Acemoğlu: The popular thinking is that humans are imperfect machines. There is a huge demand from Silicon Valley for such theses. People like behavioral researcher Dan Ariely tell everybody how humans are so prone to make mistakes that you cannot trust them with important tasks. While Ariely is now disreputed …
DER SPIEGEL: … because he is said to have manipulated data during his research on the subject of honesty, of all things …
… but such ideas sort of morphed into the Silicon Valley worldview: Ordinary people are unreliable – except, of course, for the few geniuses. That’s why the geniuses have to design technologies to overcome these imperfections of the masses, very often by monitoring workers strictly or by just taking tasks away from human employees.
DER SPIEGEL: The ultimate goal is not to help humans, but to replace them completely?
Acemoğlu: Not everybody in Silicon Valley is thinking this way. But it definitely is a very strong current. When you reward people in terms of prestige and jobs when they design programs that reach human parity, then you encourage more and more people to work on that type of technology.
DER SPIEGEL: In your book, you compare tech entrepreneurs like Mark Zuckerberg and Bill Gates to John D. Rockefeller, one of the notorious robber barons of industrialization. That’s a harsh comparison.
Acemoğlu: Figures like Rockefeller were very innovative. Rockefeller was standing at the forefront of the innovations. But he was also a ruthless and profit-crazy monopolist. Pretty much the same holds for many of today's tech barons.
DER SPIEGEL: You’ve suggested the breaking up of corporations like Facebook and Google. What is supposed to get better as a result?
Acemoğlu: With the right direction, technology and innovation can bring enormous benefits. But actually, Google, Facebook and Microsoft have not been friends of technology and innovation. They have acquired so many competitors and killed them.
DER SPIEGEL: So, why don’t politicians take tougher action against them?
Acemoğlu: Most politicians have put blinders on. They persuade themselves that increasing inequality is either inevitable or OK in the name of progress. A lot of politicians have indeed signed up for a version of extreme market fundamentalism without regulation and guardrails. One of the reasons why libertarian-type ideologies are so attractive is that they remove any discussion of otherwise very difficult social tradeoffs. The thinking goes: Whatever the market produces is good. All we need to do is let the market work.
DER SPIEGEL: Is regulation of Big Tech feasible at all? Or might regulation deprive the sector of its innovative power – or even drive the companies abroad?
Acemoğlu: China is, in fact, showing that regulating AI is indeed feasible. The Communist Party has been very successful in clamping down on tech. To be clear, I’m not advocating for the Chinese style of regulation. They are doing it for the supremacy of the party. But if authoritarian regulation is feasible, then so is democratic regulation.
DER SPIEGEL: What could we do to turn things around?
Acemoğlu: There is not one, single tool. But, for example, our tax systems today are biased against labor and in favor of capital. While in the U.S. the average tax rate on labor income is stable at 25 percent, effective taxes on investment in equipment and software have fallen from 15 percent to five percent over the past 30 years.
DER SPIEGEL: What will happen if nothing changes?
Acemoğlu: Our future will be very dystopian if we make an enormous share of the working population irrelevant. That would create a completely two-tiered society. We should do everything in our power to avoid that.
DER SPIEGEL: Mr. Acemoğlu, we thank you for this interview.
