The Technology Trap | Carl Benedikt Frey | Talks at Google

Thanks very much. It's a real pleasure to be here, and thank you all for coming despite the nice weather. I was slightly tempted to sit down on the terrace myself, so it's really nice to see that so many of you are here. Back in 2013, Michael Osborne and I published a paper in which we estimated roughly 47% of jobs are at risk of automation. And a question that we got quite frequently since then is that, is it something— there's something different about this time, right? Because we've been through periods of technological change before. And what I would like to do today is to take you through some of the history about mechanization and jobs, because one thing that I What I found quite extraordinary when researching this book is that technology has progressed enormously over the past 200 years, but the debate surrounding its effects has not. And as many of you will have noted, we are living in a period of automation anxiety where few people feel enthusiastic about the tornado that's coming for the office. And in fact, majority of Americans now think that governments should impose restrictions on how many machines businesses can install and implement in production. And this seems puzzling from an economic historian point of view because if you take a really long view, what you see is that economic growth looks like this hockey stick, right? It was slow and stagnant and for millennia and only took off in extraordinary fashion around 1800. And the standard story behind this is that this was roughly the time when the mechanized factory arrived, that with its extreme divisions of labor and the implementation of machinery in production allowed us to produce more with fewer people. And as a result of that, we are roughly 40 times richer today, adjusted for inflation, than we were at— in 1750 at the dawn of the first industrial revolution. And needless to say, that if anything understates the transformation that has taken place, because the consumer basket that you can buy with that income today is so much more diverse than it was in 1750, right? People back then could only look at the lives of the wealthy in envy, who had servants to do the most tedious things for them. Today we all have access to an electric servant in terms of dishwashers, washing machines, vacuum cleaners, and so on that relieve us of a lot of tedious work, not to mention other minor inventions like the automobile and antibiotics. And if that isn't evidence of progress enough, consider the fact that producing those goods and earning those higher incomes is much more comfortable today as well. Back in 1900, a lot of people still worked in coal mining. Explosions and cave-ins were part of everyday working life. Lung disease came as part of the work package. And today most of us work in air-conditioned offices. I was just told that you have a massage room next door and a gym that you can go to during the lunch break if you like. So it's quite extraordinary how much working conditions have improved, and in large part because of technology, right? Even occupations that still exist today have been transformed beyond recognition. Back in 1900, a farmer would have walked the fields with nothing more than animal power, and, and today that same worker can sit in, you know, an air-conditioned tractor and listen to the music of his or her choice. So there's no doubt that a lot of progress has been made. I'm here to tell you that unfortunately that is just half of the story. The Industrial Revolution created the foundations for the modern world we live in today, but getting there involved painful transition. In 1844, Benjamin Disraeli, before he became Prime Minister of Britain, published a novel called Corningby, in which one character remarks that, I see cities peopled with machines, certainly Manchester must be the most wonderful place of modern times. The same year, Friedrich Engels published his Conditions of the Working Classes in England, which was written during a stay in precisely Manchester. Engels, needless to say, had a very different take on the matter. He argued that machinery only served to downgrade people, to deprive them of their jobs and incomes, and as a result of that, he argued that it was not in the interest of the working classes for mechanization to progress. Now, Friedrich Engels was clearly wrong about the future, but as the economic historian Bob Allen has pointed out, he was actually right about the past because for 7 decades, even as the British economy took off, wages were stagnant or even falling at the bottom end of the income distribution. The wage data for this period is admittedly not that great, but even if you look at other sources of data like consumption, you see that spending on non-essentials, for example, declined during this period. And if you look at biological indicators of well-being, like heights, you find that the cohorts born in 1750 was actually taller than the cohorts born in 1850. So it seems that this had also an impact on people's incomes, their nutrition, and life expectancy more broadly. And the reason for this was in large part the mechanization of industry, the fact that machines replaced middle-income artisan jobs with low-paid jobs often performed by children in factories. In fact, early spinning machines were specifically designed to be tended by children. It was only after 1840 when steam steam power became more pervasive in production that the adult population regained the comparative advantage in production. But it took time for people to learn new skills. It took time for production to be standardized across factories so people could credibly threaten to leave their job for another factory job, which increased their bargaining power. Labor unions, of course, played a role as well, but that was actually much later. So the upsurge in wages during the later stages of the Industrial Revolution, as I argue in the book, is best explained by the industrialization process itself. But getting there took a long time. Now, one puzzle to many economic historians has been Why would people have voluntarily agreed to participate in the industrialization process if it reduced their utility? The simple answer is that they didn't. They petitioned to Parliament to block the introduction of labor-replacing technologies, and they frequently rioted against the mechanized sector. mechanized factory. Some of you will have heard about the Luddite riots, but they were only part of a long wave of riots that spread across continental Europe, Britain, and even China. And how did the British government respond to those riots? Well, not too infrequently, they sent up troops against the rioters. And in fact, the army that was sent against the Luddites was larger than the army that Wellington took against Napoleon in the Peninsula War of 1808. So I think it's important to remember that a lot of revolutionary technologies also spread a lot of political revolutionaries along the way. What began with the introduction of the first machines factories and ended with the construction of the railroads also ended with the publication of the Communist Manifesto. Now, I'm not here to predict a socialist revolution, but I do note that levels of income inequality have increased rapidly in recent decades and have been approaching levels not seen since the Industrial Revolution of the 18th and 19th century. And needless to say, what's driving this is not spinning machines and steam power, but it's a new technology. It's essentially computerization. And the first electronic computer was developed at the University of Pennsylvania in 1947, but it was probably too large to be fitted into this room, and as a result of that, it had virtually no impact on the labor market and productivity growth. It was only around the 1980s, with the introduction of the personal computer, that A, companies were able to restructure supply chains to take advantage of cheap labor in countries like China, and B, that companies became able to automate a lot of routine rule-based activities that have disappeared in offices and factories. And this automation has been very good for some of us, right? So most people in this room, we are able to work more productively with the help of computers, and we can reach a global marketplace to export our ideas. But it hasn't been that great for everyone. And if you look at prime-age men with no more than a high school degree in the United States, their wages have actually fallen for 3 consecutive decades. They have been worse off in the labor market due to, in large part, automation. Now, we have already seen a backlash against globalization, but automation and globalization have had very similar effects on the unskilled and their communities. Where the robots are is also where many of America's problems are, and a variety of studies have shown that when jobs dry up, you get a lot of social problems in those communities, such as rising crime, faltering marriage rates, and, and you even see what economists Anna Case and Angus Deaton have called the deaths of despair. A lot of people dying prematurely because of alcohol-related causes, liver disease, and suicide. And not too surprisingly, these are also the electoral districts that are most likely to opt for radical political change. Our studies show that much of the 2016 election actually played out in electoral districts which adopted robots very extensively. And if you want to explain why 3 key swing states like Michigan, Wisconsin, and Pennsylvania, who had voted for the Democratic candidate every single election since 1992, all of a sudden ended up voting for Donald Trump. Automation is definitely part of the story. Now, the worrying part is that we have seen nothing yet, because past automation was essentially based on top-down computer programming, with a person specifying what the technology should do at every given contingency. We are now in an era of bottom-up machine learning where new technologies can essentially infer the rules of the game themselves through, A, data tapping into our experience, and trial and error. And at some risk of oversimplification, this is what's driving everything from recent developments in autonomous Driving to Google Translate. Now, when I make this presentation, I'm sorry to say some people sometimes point out that, you know, Google Translate is not perfect. But at the same time, it's important to remember that every revolution in technology began with imperfect technology. The early steam engines were merely used to drain coal mines, and even that didn't do particularly well. And we've seen exponential improvements in many of these technologies in recent years. The second thing to point out is that we are not perfect either, right? This is a study of Israeli judges and their decision-making during the day, and what you see here is that we make a high share of favorable decisions just after we have morning snack, and then, you know, the share drops down during the morning and bumps back again as insulin levels go up after the lunch break, and so on and so forth, right? And perhaps we can actually use algorithms to make better decisions in some domains where we perform poorly. Thirdly, I think it's worthwhile pointing out that much automation doesn't happen by replicating exactly what the human worker does, right? We didn't automate laundry work by building robots that, you know, walked out from the kitchen into the forest, chopped down trees, carried wood and buckets of water into the home, heated it on the stove, and then performed the motions of hand washing, right? We did that by inventing the electric washing machine. In the same way, we didn't automate away the work of lamplighters by building robots that are capable of climbing lampposts. And engineers have, at least historically, been very good at circumventing some engineering bottlenecks by simplifying tasks, and I suspect that is likely to continue. That being said though, there are still certain domains in which computers perform poorly. Simple things for us, like deciding what's a piece of rubbish lying on the floor and what's a really important document, is still something that, you know, is very hard to explain. And there are no general-purpose robots that are able to do all of the things that your cleaner does, for example. When it comes to more complex tasks like creativity and complex social interactions, we also see that technologies are still performing quite poorly, right? And I think this is well captured by the Turing test, where chatbots try to convince human judges of them being a person. And some people argue that there was a breakthrough a couple of years ago when one chatbot actually managed to convince 30% of human judges of it being a person. But it did so by pretending to be a 13-year-old orphan Russian boy speaking English as a second language with no understanding of English culture. And if you think about the variety of much more complex social interactions you do in your daily jobs, I think it's almost inconceivable that you will be replaced in those tasks in the near future. Now, a key question that emerges from this is then how intensive are our jobs in tasks that correspond to these engineering bottlenecks to automation? According to our estimates, quite a few are actually not very intensive in such tasks. Roughly 47% of jobs, we estimate, are at risk of automation, and that's no longer just production jobs. It's everything from transportation, logistics, warehouses, retailing, construction, and so on and so forth. And when we published this study a couple of years ago, we actually published a very detailed list of 702 occupations and their relative exposure to automation. And you can imagine that quite a lot of people dug through this list and tried to come up with silly examples. And my friend Ken Kukiera at The Economist used to tease us because we found that fashion models are highly exposed to automation. These models on these pictures actually don't exist. They've been created by generative adversarial networks which have, you know, generated them from thousands of pictures, and they're already being used by some companies like Dior. At the same time though, we did say that technology is clearly not the only factor that drives decisions to automate, right? When Nissan produces cars in Japan, it relies heavily on robots. When it does the same thing in India, it relies heavily, more heavily on cheap labor. When tractors were first invented, they were long not adopted because there weren't any operatives that could actually run the machines. And the New York Times noted in 1880 in 1918 that the tractor is too good a machine to be put in the hands of a poor operator, and to spur adoption, colleges were actually created that provided training courses for tractor operators to spur adoption. There are essentially restaurants today which are vending machines which kind of cut out the waiter from the process, and that may be good for certain events like going for a quick lunch, but we are unlikely to prefer having an automated waiter for a pleasant dinner, right? So consumer choices also matter a lot for decisions to automate. And even if Google Translate became perfect tomorrow. Human translators still, or for certain documents, you still need certification. So unless we certify Google Translate, those jobs are not going to be replaced tomorrow. And lastly, as I pointed out earlier, adverse public opinion can matter as well. These are dockworkers who went on strike in Los Angeles last month because of the introduction of machinery in the harbor in which they work. And I think that a big difference between today and the Industrial Revolution is that people today actually have, at least in the industrial West, political rights. And I think that the economist Leontief was onto something when he suggested that if horses could have joined the Democratic Party and vote, what happened on the farms might have turned out differently. They may have used their political rights to bring the spread of the tractor to a halt. And the case of the Luddites was essentially hopeless, right? Because in 19th century Britain, property ownership was still a requirement for voting. So their voice was essentially left unheard. Now, I'm actually an optimist about our long-run future. I'm 100% certain that artificial intelligence are going to make education and medicine better. It will rise productivity, boost some people's wages, It's going also to create a lot of new jobs that we can't even imagine today. And it's already created new jobs in recent years. LinkedIn recently did a survey on their platform where they looked at the fastest growing occupations. And among those you find jobs like iOS developers, social media interns, big data architects, but also Zumba instructors, and you find jobs like digital marketing specialists and also Beachbody coaches. And I think this reflects a broader trend that we are actually seeing in recent years with tech jobs clustering in skilled cities like London, like Stockholm, in the Bay Area, and so on, and those people demanding more specialized in-person type of services. And the result of this is that more people are clustering in these places, and that has been great for cities like London, but it hasn't been great for certain people, people in the countryside who have seen jobs vanish due to automation. And this has created very much of the divide that we are seeing today between these hubs for technology and the rest, essentially. And I would predict that unless we do something about this, these trends are likely to continue. The book is called The Technology Trap, and the technology trap actually refers to pre-industrial times when workers frequently resisted the implementation of new technologies and governments, fearing social unrest, essentially sided with angry workers. And as a result of that, new technologies were often resisted and often outright blocked, and this meant that economic growth was also slow for a very long time. The question I ask in the book is, could we return to this technology trap in which governments tax robots and break up technology companies and essentially try to hinder the spread of automation technologies? The simple truth is that I'm far from sure about that. But I do think that it is a real risk. And I think to harvest the long-term gains from artificial intelligence and other technologies on the horizon, we need to think through the short-term dynamics quite carefully. And I think that's probably a good place to open up for questions. So thank you all for coming, and great to be with you. Thank you.

So if you were then advising policymakers on what action should they be taking knowing these trends, what would be the big policies that you think they should be prioritizing?

I think the keyword there is in a way big, because everybody thinks that you have a big challenge and you need a big policy change to sort of solve it. And unfortunately, I don't think that there is this one big policy and that to deal with the challenges that we face is more of a package of many things that individually may sound minor, but that collectively can make a big difference. So to take one example, I grew up in southern Sweden close to a city called Malmö, right? And Malmö was a city that specialized in building ships, an industry that went down in the 1990s, and the city was actually doing quite poorly for a long time. And essentially it only took off again with the construction of the bridge to Copenhagen, which meant that people could stay put in Malmö where housing was cheap, commuting to Copenhagen where there was a greater abundance of new jobs, spend their money locally in Malmö, which gave a boost to the local service economy there, and essentially created the this virtuous cycle. And I think that one of the key drivers of both income and wealth inequality has actually been that there are significant constraints on housing supply to places like London where these new jobs are emerging. And a lot of people can actually not afford to live in places that, you know, provide most opportunity. And that is the reason why some people are now leaving the Bay Area for places like Austin, because it makes it harder to actually attract talent. So I think that a lot can actually be achieved by building more in, A, in places that are expanding, and B, connecting them through investments in smart infrastructure Because people ideally want to stay put in their communities, right? And if you can, you know, connect declining places to expanding ones, a lot can be achieved. Now, you know, looking into the future, you know, there may be better transportation technologies than we have now. There's currently a feasibility study being done of connecting Cleveland and Chicago with a Hyperloop. So that's currently a 6-hour drive. If that was to be successful at some point, that would be a 28-minute commute. So 2 labor markets would essentially be one. And I also think that a lot can be achieved by investing more in early childhood education. So we don't know exactly what the skills of the future will be, but we know what some of the hurdles to acquiring new skills are. And if you have early deficits in math and basically reading, you're much less likely to learn new skills later on. You're much less likely to go to college, and you're much less likely to participate in civic activities, and so on. So I think by investing in early childhood education, particularly in deprived areas, a lot can be achieved as well.

If you ask people in many of those towns, they'd say, we don't want to be a dormitory town to a big capital city. We want good jobs back like we've had before, like farming jobs, fishing jobs, manufacturing jobs, would your argument be, well, that would effectively be like the Luddites were trying to do, hanging on to jobs that will no longer exist? Or would your argument be you need to invest in those other kind of jobs as well?

Yes, I mean, I don't believe in conserving old jobs. And, you know, I mean, bringing old jobs back is essentially conserving the past. And if If we would have done that for the past 100 years, a lot of people will still work in coal mining. And coal mining communities might have been better off then in a way than they are now. But on average, you know, we're so much more better off because of the changes in the labor market and the changes in technology we've seen over the long run. And I think, you know, in the short run, a lot of people um, or are likely to be made worse off by changes in technology. I think a problem, problem when economists speak about the short run is that we are usually not very specific about what that means, and whether it's 20 hours, 20 days, or 20 years kind of matters. During the Industrial Revolution, that was a long time, but in sort of the grand scheme of things, right, Later generations have been made so much better off as a consequence of that. And if we value later generations as we value current ones, then we should think that it's still in the greater good. I was interested by the robot map, if you will, showing kind of that correlation, if there is a correlation, because I guess the loudest explanation that I've heard about why kind of the, the US shifted was offshoring of jobs. Do you think that there's maybe a lack of awareness among those that are kind of at greatest risk of automation? I know you showed kind of the 47% statistic in the LA Harbor, but do you think there's a great recognition of that? And do you think that kind of narrative is going to get louder over time? Yes, I think first of all, I mean, offshoring is enabled by technology as well. So, you know, to disentangle the two is tricky. But it's absolutely clear that both offshoring and automation has contributed to deindustrialization. And if you look at the United States, you see that manufacturing employment peaked in 1979, and the manufacturing employment share has fallen steadily. since then, but the output share of, of the economy has actually made— been constant over the same period of time. So there's still a lot of stuff being produced in the US, but with fewer people. And I think, as I mentioned, globalization and automation have both played a significant role. I think looking forward though, the rise of China has already happened. We've already seen a lot of offshoring. Most people today actually work in non-traded sectors of the economy that can't be exported. So they are relatively shielded actually from future globalization, but they aren't shielded from automation, right? So there are roughly 3.5 million cashiers in the US. which are exposed to technologies like Amazon Go. There are another 3.5 million truck, bus, and taxi drivers which are exposed to autonomous vehicles. So there are a lot of jobs that are not offshoreable but are automatable. And I think that is also one of the reasons to be concerned that it may actually shift more towards technology. Facebook recently made the argument against the breakup of Big Tech that if Big Tech is being broken up, somebody else, as you just mentioned, in this case Chinese company, would take over. Do you believe that we reached a point where globalization trumps the power of single governments to stop, or in this case globalization would encourage or accelerate the spread of automation? So one of the key reasons that the Industrial Revolution happened in Britain was not that there was an absence of resistance to technological change, right? Quite the contrary. And the reason that it happened in Britain was that governments for the first time sided with innovators and pioneers of industry. And the reason for that in turn was in large part actually early globalization and competition between nation-states. So the craft guilds, for example, which sort of forcefully resisted new technologies, became weakened by globalization because their political clout, their didn't extend beyond the city in which they worked. So as a result of that, the threat from below was essentially reduced. And because governments increasingly realized that, you know, their military muscle depended on their economic strength, they also invested more in mechanization and allowed it to thrive. And I think that essentially goes to your question in the sense that, you know, we are concerned about an outside threat or competitor gaining from, you know, imposing essentially new technologies potentially on their population, even if there is resistance to them. Now, In most modern democracies, you know, we are more worried about the short run than, you know, the Chinese government is, for example. And, you know, whether we have Jeremy Corbyn or Boris Johnson in office, they all worry about being elected in the next general election. So they are much more likely to take, you know, the democratic process, or the threat from below, if you like, more seriously than, let's say, the Chinese government is. So, I mean, I take the point that, you know, China may have an advantage in that regard. But I still think, you know, it's a difficult selling proposition to voters and say, look, you know, it may be your jobs that are at stake, but we are still going to implement these technologies because of, you know, competition with China and some other place.

One question I have is one of the kind of things that components of the processes that are happening today is the environmental challenges and limits that we're reaching. How do you see this play out within the kind of processes that are occurring right now? And do you see this different from before?

Sorry, you're asking about the environmental challenges related to automation?

Yeah.

Okay, so I mean, I don't have a strong view on that. So I think my— there's a tendency of new industries to be, you know, less carbon intensive than past industries. And I think as the economy dematerializes, And that is going to be less of an issue with new innovations. I mean, if we— if you speak to our group of geoengineering at Oxford Martin School, they actually think that to reach the 2% target, you need, you know, technical solutions, because even if you stop carbon emissions tomorrow, you're not going to reach it. So I think sort of stopping the technological clock due to environmental concerns, which I take to be slightly implicit in your question, maybe it was not, uh, it's, it's not a great idea. But—

My question was more about whether you think that the limits that we're reaching, um, are, um, kind of configuring a different scenario to what was happening in the previous Industrial Revolution when those limits just simply weren't there, because we had like all of the resources to use. And I was wondering whether you see this factor play out within the processes that are occurring right now. But you then said you kind of answered, so thanks.

Thanks.

No, but there were actually some significant limits. So I mean, the early Industrial Revolution in part also happened in northern England because it was close to coal and transportation costs were still relatively high. That was actually a constraining factor in the early days of industrialization as well. I think you mentioned it in the introduction of the book, that there was also the impact on communities. So for example, during the Industrial Revolution and what we have witnessed until now, probably social order is very similar, right? Roughly, right? I mean, like, people moved into factories, they still work together, the community was much more stronger. Well, actually now with technology, I think there is a much more shift to individualism. the concentration of power even more in less people. How do you compare kind of the societal impact of technology during the Industrial Revolution versus what we were just through? So I think in that regard, the Industrial Revolution was mainly more disruptive in the sense that a lot of people— so we still, you know, people still worked, but few people actually had jobs, right? Most people worked at their own pace and at home, often, you know, close to their relatives and children, and, you know, everybody was under one roof. And what the factory city essentially did is to take a lot of people from various places across the country and bring them all together in rather unhealthy, aesthetically unappealing environments, and, you know, at wages which hardly compensated for those working conditions. So I think In that sense, the Industrial Revolution might have been more disruptive than what we've seen today.

You've touched a little bit about the policy level and some of the things you could do there. What would be your advice at the individual level if you're speaking to somebody who is facing, say, imminent threat to their role, maybe due to the Amazon type of situation of an automation of checkout or something? Is it simply reskill And also, is there a role the government ought to be playing in terms of helping them find that next role in terms of that reskilling journey?

Right. So it's a great question. I'm an academic, not a career advisor, but broadly speaking, I think you do well in focusing on the things that machines do poorly. So I think, you know, a lot of work that involves complex social interactions, creativity, for example, is a good start. I do note that, you know, computer science, we struggle to keep PhD students for the first 6 months, right? Because they get hired by Google and other companies, and it's clearly, you know, a skill set that's enormously in demand. Broadly speaking, I mean, also if you think that data is the new oil, it's a cliché, but nonetheless, Statistics and machine learning generally is probably a good idea, but I think people need to figure out for themselves what they enjoy and what they're good at as well, so I don't have a very strong opinion on that.

I see technology as something that essentially humans created to make their lives easier, so even if there was a limit imposed by an authoritative body, on technology progressing, then humans will always find a way to make their lives easier. Would the question not be about, say, the distribution of resources? And I only mention this because you mentioned the Communist Manifesto earlier in your presentation, but isn't that essentially one of the questions that Marx answered? When we reach a certain level of technology, that resources have to be distributed. Obviously, he envisioned a world where, you know, robots are working in the farms and people are frolicking about picking flowers and stuff, but—

Yeah, so Marx thought that people would have to acquire the means of production. He didn't think that wages would rise as mechanization progressed. So things happened quite differently. Well, I mean, I'm not sure that I agree with the basic premise that people just innovate to make their own lives easier. So I think a lot of innovation is happening in businesses today. And it's happening to meet the, you know, certain consumer demand. But I do agree that, you know, it's very much a distributional issue. If you look at a country like Saudi Arabia and a country like Norway, right, they both have oil, they've essentially, you know, black gold coming out of the ground. But they're very different economies, they're very different societies. It's more widely shared in one place, than the other. And obviously that is due to the result of institutions and culture and factors that have nothing to do with technology in itself. So clearly that's right.

You drew the parallels between the rise of automation in certain parts of America and the rise of the Republican— or the switch to the Republican vote. So I'm further drawing the parallel between the rise of populism with that. Are there any kind of parallels in what happened with the Industrial Revolution, and is there any sort of light at the end of the populist tunnel that we seem to be going down at the moment?

So sorry, if there is a parallel with the rise of populism?

With the rise of, with the automation, how did that pan out in a similar way back in the Industrial Revolution? Was there a shift in politics and how that responded to that, and then how did it evolve once the things had kind of established themselves?

Right, so not really. So I think one key difference is sort of actually in the sphere of economic ideas. So a lot of people at the time of the Industrial Revolution still believed that the world was Malthusian, right? And that was the idea that, you know, larger, higher incomes would only result in larger populations with no benefit to the average people in per capita terms. And as a result of that, it was also believed that any sort of redistribution was essentially meaningless because it would only translate into a larger population. And this idea was actually very popular. It led to the reform of the Poor Laws in 1834, which essentially, you know, took away the safety net that had existed. So the policy direction in that regard was more sort of the other way around. I think today there's actually most people realize that there are things that governments can do to smoothen the transition. And, you know, the other big difference is that even with the Reformer Act of 1867, property ownership still remained a requirement for voting. So most people were essentially politically disenfranchised.

You made the point that perhaps if horses had had the vote, then maybe the Industrial Revolution wouldn't have played out as planned. You could say that some of the beneficiaries of technology change have the vote today, but many don't. And so in advanced Western democracies, how does this all play out?

Sorry, I'm not, I'm not sure if I follow your question.

In a world where everyone had— where in advanced economies where most people have the vote, but Perhaps many people won't be beneficiaries of technology change. How does this play out in a democracy?

Yes, I think that that's the key point, right? So I mean, if people are not happy with the verdict of the market, they can opt against it through various mechanisms. And I think that somebody who is sort of tapping into this concern, like Andrew Yang in the United States today, is gaining a lot of traction on that. So I think it's extremely hard to predict how this is going to play out. And that's sort of also one of the key points of the book, because this 47% study that we published back in 2013, in a sense, you know, created a literature in which people were trying to predict exactly how many jobs that were going to disappear or appear by 2030 and so on. And we merely looked at the potential scope of automation. I think the pace of it is much more complex. It depends on what we do. It depends on acceptance for new technology. And much of that sort of is going to play out in the sphere of political economy. And we struggled to predict the outcome of the general election in the very same day. So I think we need to be rather humble about our ability to predict such outcomes.

Brilliant, thank you. With that, please join me and John in thanking Carl Benedikt Frey.

Thank you.