 Well, there was a couple of really tough acts to follow. Roberto and Marshall had some just awesome research. That's one of the reasons it's so great to be here at MIT and be able to participate with them. David was just talking about how there's a revolution underway. And in our book, The Second Machine Age, we explicitly harken back to an earlier revolution, the Industrial Revolution, which really literally bent the curve of history. As an economist, if you look at what happened for millennia, for centuries, in terms of the basic living standards of people, the answer is essentially nothing. There was people living standards didn't change much until a certain time and a certain place. That time was the late 1700s, and the place was right here in England where the Industrial Revolution set off a fundamental transformation in our prosperity. And that was because of a set of technologies, in particular, the steam engine. And so it's good to be here where that first revolution really got underway and to start talking about the new revolution that's underway and the one that's involved in The Second Machine Age. When you look at what it was that made the first machine age so successful, it's because it transformed one of the two key ingredients that's needed to really change the world. So if you think about what it takes to change the world, one of the things is a power system. The world is literally made of atoms, mostly. And if you want to change the world, you need to be able to move them around, transform them, shift them. They're held in place by powerful forces. And in the first machine age, we learned how to use technology to do a lot of that. Instead of having human muscle power, animal muscle power, we used machines, incredibly powerful machines. First the steam engine, later electricity, the internal combustion engine. These are what economists call general purpose technologies. And each of them provided more and more power to move and transform things. And it's important to have that. And that's what led to a roughly 50-fold increase in living standards since the late 1700s. But it's not sufficient. You also need a control system. You need to figure out how to arrange and organize those things that you are moving and transforming. So in the Industrial Revolution, we set forth a set of technologies that dealt with physical power. And they were mostly a compliment to humans because the humans were the control system. We were the ones that managed how that physical power was deployed and how it was used. And I mean compliment in a very specific technical economic sense, in the sense that the technology made human labor more valuable. And I don't just mean that intuitively. You can measure how much more valuable human labor became because wages rose. So the value to an entrepreneur of hiring one additional worker went up. And that was reflected in the fact that they were willing to pay more for that labor power. Now we are, as I mentioned, in the early stages of a second machine age. And Andy and I have been trying to study and understand this. And one of the special characteristics of this second machine age is that we are beginning to use technology not just to deal with the physical power, but also the control system. Not just to automate muscles, but also to augment our brain's mental power. And the general purpose technologies of this revolution include computers, software, machine intelligence, and other technologies. It has some similarities to the first machine age. It's leading to an increase in output per person and increase in productivity, more wealth creation. It has some differences as well. One difference is that many of the technologies are evolving dramatically faster. There's an exponential improvement in the power of computers. We call it Moore's Law, which means that a child's PlayStation today is more powerful than a military supercomputer from the late 1990s. Computers get better faster than anything else ever. Now, in the first machine age, there are also improvements in the technology. And we went back and charted, for instance, the improvement in the efficiency of the steam engine. That also had a doubling time. But the doubling time was about once every 70 years. So it took 140 years for it to become four times more efficient. So a very dramatic difference in the exponents. Another difference is that as these technologies evolve, it's not as clear whether or not there are going to be complements or substitutes for human labor. You can see some ways in which they make human labor more valuable, but you can see other ways in which they really replace the need for humans in the loop in many situations. And if you look at the data, it's a bit troubling, because median wages, rather than going up, have stagnated and even begun to fall, suggesting that entrepreneurs, managers, the market is not willing to pay as much for that additional worker on the margin the way they did for two centuries previously. So let's dive in a little bit more closely and try to understand what these technologies are and how they're changing the economy. Let's go back, not way back to the 1700s, but let's go back to just a little about 10 years ago, when we were first beginning to look at some of these issues I was teaching a course at MIT. We used this terrific book by Frank Levy and Dick Murnain. And I had a great discussion with my class about what were the things that were uniquely human. Humans could do well that machines could not do. And we brainstormed and we came up with a list here. It's a fun exercise to do. We redo this exercise periodically. But three categories that we thought seemed to be pretty important. One was autonomous mobility, just being able to move around, whether it's in a room or out in the world. Fine motor control, you can pick up a pen or a fork from your table very easily. That's actually quite hard for a robot to do. Another category is language and complex communication. We speak, we interact, we write. And the third category is unstructured problem solving and pattern matching. Machines we acknowledged were quite good at solving structured problems. But the unstructured ones that weren't well defined tended to be very difficult for machines. And we thought of some different examples of things that just had a lot of these unstructured characteristics that used lots of sensory data, fine motor control, that wouldn't be very suitable to automation anytime soon. And one of our favorite examples was actually driving a car or truck through heavy traffic. We didn't see any likelihood that that would be automated anytime soon. I'm looking around. I hear some laughs and chuckles over there about how stupid we were. Because of course, you know now that there are self-driving cars. So we really underestimated. In fact, here's a picture of Andy and me. We are grinning kind of like idiots, like we sometimes do when we learn something new. In this case, we just drove, or I should say rode, in this self-driving car that's behind us a couple of years ago from Cupertino or from Mountain View, California, up to San Francisco and back down again on Route 101, the highway with an autonomous vehicle. So we were happy to be proven wrong and realized that the technology is advancing a lot faster. And some of the categories that we thought wouldn't change anytime soon were, in fact, changing a lot. And this is part of what sparked us to think a little bit about self-driving vehicles. And I want to open this to questions, and I'm going to do it right now. Try that Sturro Drive. Yes, Sturro Drive? Yeah, that's a lot harder. In fact, this is a good point that Route 101 in California is an easier environment than Sturro Drive in Boston. Just any kind of driving in Boston is a lot harder. Because not only are there no structured rules, none of the human drivers follow the rules either. And it may be a long time between when cars can drive on highways to driving in downtown London or downtown Boston. So these things are evolving, but I certainly don't want to give the impression that this is a completely solved problem. But it is worth kind of assessing where we are doing an update on what we thought was true 10 years ago and what's true again. So let's come back to that list of interacting with the physical world, language, and problem solving. And I want to make the case that in all three of these categories, there have been just dramatic changes, dramatic improvements, and machines can do a much better job in each of those categories than we anticipated a while back. And this is part of what's sparking this acceleration of the second machine age. So let's take them each in turn. The first one, interacting with the physical world. Our colleague, Rod Brooks, has started several robot companies, iRobot, now Rethink Robotics. And this is one of the robots that he has. It's called Baxter. And Baxter is an amazing robot. It can work side by side with humans. Traditional industrial robots are really dangerous. If you put your head in the way of one of its arms, a traditional robot might just not even notice and knock your head right off, literally. Baxter is aware of the surroundings and is gentle and will not hurt people when they work alongside it. It can do a lot of simple tasks. Like here, it's working an assembly line, putting some stuff in boxes. It seems like a very simple routine task, but it's something that millions of people all around the world do, basically, this exact task, including people in Massachusetts, a relatively high wage state and place in the world. There are people doing this task. And Baxter works for about the equivalent of $4 an hour, according to Rod Brooks. And they just came out with a new version that's about twice as fast, 10 times more sensitive in terms of its motor control called Sawyer. So there's been some very rapid improvements there. And you can see on the horizon a next generation of robots that have even more mobility and control. For instance, this is Atlas, about six foot tall, six foot feet tall, 300 pounds. And Atlas is the robot that's being used currently in the DARPA Robotics Challenge. The previous DARPA challenge was to make a self-driving car. The new one is to make a robot that can walk up to a vehicle, get inside of it, grab the steering wheel, put its feet on the pedal and drive a short distance. It's more like a golf cart than a car, but it's sort of beginning to get the feel of it. It walks across a field of uneven cinder blocks and steps across them without falling. It reaches down and picks up a hose, attaches a nozzle to it, goes up a flight of stairs, turns a lever, goes through a door. A series of simple tasks like that that are sort of an obstacle course for robots. And Atlas cannot yet do all those things quite well. It's been able to do some of them. But if you talk to Gil Pratt, who runs DARPA's program, the DARPA Robotics Challenge, he says that the teams are way ahead of where he anticipated they would be at this point. And he seems quite confident that it will achieve those goals. And if you're a manager or an entrepreneur, it doesn't take a ton of imagination to think about, huh, if a robot can do all those steps, what are the kinds of jobs that are being done in a warehouse or a factory or elsewhere that we could put together some combination of those tasks and it would be useful, economically useful work. Another big category I talked about was language. And today, if you walk down the street, you see somebody and you see them talking to their phone. And in many cases, they literally are talking to their phone, not on their phone, but they're talking to the machine and they're expecting the machine to understand what they're saying. And they expect the machine to answer them back in English or whatever language they're speaking in. That's just kind of weird and amazing. I mean, think of it, 10 years ago, that would have been considered science fiction. Again, I don't think the machines are quite perfect at understanding what we're saying, but we're in the middle of, I call it, a 10-year period. Tom Mitchell at Carnegie Mellon describes this period we're in where we went from machines just not being able to understand this at all to machines being routinely expected to understand our language as we speak to it and carry out actions based on that. There are machines that are translating language. Microsoft Skype program, you may have heard, will now translate from German to French or Spanish to Chinese or perhaps someday from British to American. I'm hoping we'll be able to understand that. And those technologies are also evolving quite rapidly. Again, not perfect, but serviceably well. One of the most amazing things, I think, is the way that they're even beginning to generate quasi-original text. Companies like Automated Insights or Narrative Science are doing that. Here's a new story about the world champion New England Patriots football team. And to translate that, world champion means that they've defeated all the other football teams in the world that are based in the United States. And football means, well, it's not what you think football is. So it's a different kind of sport, but what's interesting is not this story that's sort of just a run-of-the-mill kind of new story, sports story that people read thousands of times a day. It's who the author is. Someone by the name of Automated Insights. That's right, a robot wrote this story. It took the box score, took the statistics of what happened in the game, and it inferred from that what the storyline was of the sports game. And they are generating thousands of these stories for professional sports, for college sports. There are versions of these algorithms that are being used for earnings reports and accompanied by magazines like Forbes. So you're starting to see machines actually writing. And again, I want to emphasize, these are not sophisticated novels or research reports. They're very simple, basic stories, but it is a threshold that's being crossed here. And what's the third category I talked about? It was unstructured problem solving, answering unstructured questions. Could give you lots of examples, but maybe my favorite one is just the game show, Jeopardy. You may be familiar with this game. They can ask questions like this one that involve literature, sports like the previous slide, news, geography, science. You never quite know what the question's going to be. Often the questions are a little convoluted. They involve puns. So it's hard sometimes even to understand exactly what they're trying to get at. Something that humans are quite good at, like this guy, Ken Jennings, was the world champion, won 75 games in a row. Andy said there was a rumor that he was genetically engineered from birth to play Jeopardy. I'm not sure that's true, but he is quite good at that. But they had IBM Watson play against it. And the father of IBM Watson, Dave Ferrucci, showed me this most remarkable class. Like he showed it to my class at MIT. And I want to share it with you. If you chart how accurate Watson or a human was at answering questions and how many questions they tried to answer, you get this kind of a chart. There's little blue dots up there, little mosquitoes buzzing up there. That's the data from all the human Jeopardy champions, all the people who ever played Jeopardy on television and won the game, the ones who were really good. You can see that they tended to get a lot of the questions accurate. That's why they were champions. They also are pretty aggressive, often answering more than 50% of the questions. So that's what humans did. How did Watson do? Well, actually not so good at first. When Watson first started, it was dismal, getting 10, 20, never more than 50% of the questions right. Watson's aligned instead of a dot because they have a little dial. They can make it more aggressive or less aggressive, only answer things that's really confident in. But even then, when it's confident, it still isn't all that good. But Watson had one feature that no human had. And that was the ability to learn at a ferociously fast pace. So they fed lots and lots of information into Watson. And every few months, they came out with additional information and a new model. You can see Watson just progressed at a scary rate. One point they gave Watson the entire Wikipedia to read and understand all the correlations in there. Can't really read it the way we read it, but it read it well enough to see correlations of basic facts. Another point they asked Watson to read something called the Urban Dictionary. If you guys may know what that is. It turned out that was not such a good idea. Watson started answering some of their questions with profanities, so they had to erase that part of the memory. Don't let a six-year-old just surf the internet unsupervised. When Dave showed this chart to me, you could see Watson was already better than most of the human jeopardy champions. These are not run-of-the-mill people. These are champions. Well, there's still some people up there at the top right, like Ken Jennings, who on average did much, much better. But a few months later, they went on national television. As many of you may know, they went ahead and beat Ken Jennings and Brad Rudder, who was the next best person in the world, quite handling. More than the two of them put together three times as much as either one of them, so they won $77,000. But the point, if you really think about it, was not to win $77,000, it was to take this technology and apply it to lots of other areas. So today, that technology or related technology is being used in call centers. For instance, there's a call center in South Africa, which, when you think about it, what is a call center? It's a question answering system. You ask questions and you expect whoever's answering on the other end to tell you the correct answers. Watson is powering that call center. They have a version of it that's helping with legal questions. There's asked versions of Watson that help give investment advice and financial advice. There's several versions that work for medical diagnosis. You can go up to it. This is kind of fun. And describe your symptoms. And Watson will give you a differential diagnosis of what is likely the disease that you may have because it's causing those symptoms. If Watson is not today the best diagnostician in the world, I wouldn't be surprised if it is within five years. People like Vinod Kosla, the venture capitalist, legendary venture capitalist in Silicon Valley is making exactly that bet. And you can imagine combining marrying Watson with Siri. And now you have a medical diagnosis that's available to billions of people, perhaps for free, on a mobile device worldwide. So those are some remarkable changes. And I could go through some more technologies, come to our annual conference or go to our website and you can read other descriptions of some of the things we see happening. But let me now turn to what all these changes may mean for the economy. What kinds of changes do we expect to have happen? And the first set of changes, I think, are just very, very good news. And that is that we can have vastly better medical diagnosis and just vastly more goods and services than we did previously at a much lower cost. In fact, a cost approaching zero in many cases. This newspaper ad somebody sent me was from the early 1990s. And the interesting thing about it, all these devices for sale, there's a computer, a phone, a video camera, a regular camera, a music player, there's an answering machine, if any of you guys remember what an answering machine is. All those things that cost hundreds or thousands of dollars, collectively, are now available for free as bundled in as part of your smartphone. Most of them, I shouldn't say all of them, but most of them are available. And then there's a whole set of other products that you couldn't even have gotten then. A GPS, we talked about earlier, or a way of diagnosing skin disease, things like that, that are new applications. And again, they're available at no additional cost once you've paid for the basic hardware or the internet connection. And we've been trying to do some research to understand what is the value of all these free goods that are now available in the mobile economy and the digital economy and the internet. And first, we just looked at one slice, which is free goods in the internet, like Wikipedia. What would the value of those be if you had to pay for them? Well, it turns out it's about $300 billion per year. And that's interesting because that's a number that does not show up in the GDP statistics. GDP measures the value of all the goods and services bought and sold. So if something has zero price, it has zero weight in the GDP statistics. But that doesn't mean it has zero value. Wikipedia is probably more valuable to me than Encyclopedia Britannica was, with all due respect to Encyclopedia Britannica. Wikipedia has about 10 times as many articles, it's more up to date, and the accuracy apparently is comparable. So that's good news, we're getting lots of bounty, free goods and services that aren't measured. We have to update our statistics. People like Roberto Rigobon, you heard, are showing the way to how we can get more fine-grained data to understand better what's happening in the world. But that's not the whole story. If you look at some of the main characteristics, statistics that economists keep track of, it shows an interesting pattern. So first let me show you these four, productivity, employment, median household income and real GDP. And for much of the past economic history, since the Industrial Revolution, those have been rising in tandem, tightly coupled. And there's sort of an implicit social contract that as the economy got bigger, we would share in that, in particular for instance median income. The income at the 50th percentile would grow alongside. But more recently they've become decoupled. Andy and I call this the great decoupling. Productivity and GDP growth have continued to grow, but median income has stagnated and in fact is lower now than it was in the 1990s. Employment is also falling, the share of workers falling, the employment population ratio is falling. So this is a break from history. This is different from what we saw in the first machine age. What's going on? Why is this happening? Well, to be frank, we were really puzzled when we saw this, we were like, wait a minute, this doesn't fit with our intuition. We see this wondrous innovation at MIT and elsewhere. It's creating all these new possibilities. How can it be that median income goes down in the face of all this wonderful innovation? But facts are stubborn things and we looked at it and we realized, well, this is really accurate. People who have been pointing these, this falling median income are not wrong. You can tweak the statistics a little bit and you can adjust for household income, but the basic story I just showed you doesn't change. So how can that be? And the aha moment came when we were talking about this when we realized that while digital progress does make the pie bigger, there's no economic law. There's nothing in any textbook that says that everybody or even that most people will automatically benefit. It's possible that the pie can get bigger and some people don't participate. Now, it may be a very small group that doesn't participate, but in theory, it could be more than half the population that isn't joining in that bounty. And the data suggests that recently we haven't created the kind of social and economic and technological system that has created the shared prosperity that we had in earlier eras. And the economic term for this is biased technical change. And let me quickly describe to you three kinds of biased technical change. I wanna get into some time for questions and answers as well. And this biased technical change means the technology is helping some groups more than others. The first one that's been studied quite at length is skill biased technical change. And that looks at the differential effects on people typically with more education or less education. Sometimes it's more training or experience as other ways of dividing it, but it compares those different groups. And recent technologies have affected them quite differently. Here's a terrific graph from our MIT colleagues, David Otter and Daron Asimoglu, who studied this in some depth. And really, if there's been lots of people, we wrote some papers on it and others. And this collective literature basically describes how in the early part of the century there was a rising tide that lifted all boats. Then there was the oil shock and recession and everyone took a bit of a hit. But since then, you can see things are spreading out. People who've gone to graduate school or college graduates have resumed the increase in incomes. But people who are high school dropouts or don't have as much education are worse off in terms of wages in absolute terms. This is true for males. There's a similar pattern for women. All the whole thing is skewed upward a little bit. But if you control for demography and other factors, the key characteristic that shows that comes out is this change in education. So if any of your kids are thinking of dropping out of college or not going there, you might want to show them this chart before they make that decision. There's clearly been a change there. And it's not just one particular set of industries that, driving this, we have a whole set of different industries, starting with some of the more digital industries, into some of the other industries like retailing and finance that have been very much transformed. Manufacturing, logistics are clearly in the midst of a big transformation. Professional services and others are sort of on the cusp. We don't see it quite happening in education yet, but you can see that the beginnings of it are there. And maybe someday, even domestic help and other categories are all being transformed. The second type of bias technical change that we're observing is what we'd call capital bias technical change. Another way of dividing the world is between capital owners and those who provide labor. And for much of history, there was sort of a dance there where they kind of bounced around, but sometimes one or the other would have a little rise or fall. But you can see in the past decade or so, the labor share has been falling quite precipitously. Meanwhile, profits are at record highs, both in absolute terms and as a share of GDP. So things are becoming much more skewed. And again, you can imagine that technology is part of the story here. When someone like Terry Gao replaces thousands of workers, he says he's planning to replace as many as a million workers with industrial robots that can increase output, maybe do things in terms of efficiency and quality. But it also shifts the share of revenues that are being paid to labor to workers versus being paid to capital. And that's a transition that's happening on a worldwide basis. And the last kind of bias technical change, the way the technology can affect different groups differently, is what we call superstar bias technical change. It's perhaps most evident in some of the media and entertainment industries, whether it's Lady Gaga or Beyonce or others, they're able to reach global audiences now, digital technologies. You see it also in software. Here's Scott Cook. Maybe not quite as recognizable, but he's a billionaire, in part because he's the co-founder of Intuit that makes programs like TurboTax. And they have codified some processes that used to be done only by humans, but now tax preparation is often done by software. And if you can codify something, you can digitize something. And if you can digitize something, you can make copies of it. And 10 copies, 100 copies, a million or 100 million copies. And each of those copies has some unusual characteristics. They can be made for almost zero cost. They can be transmitted anywhere on the planet almost instantly at the speed of light. And each copy is a perfect replica of the original. You can't tell the difference between the copy and the original, they're the same thing. Free, perfect, and instant are three adjectives that we didn't use for most goods and services for the preceding couple hundred years or couple hundred millennia, for that matter. But they're standard for digital goods. They're just automatic. And they lead to some unusual economics. One of the things they lead to is power laws. Andy and I and Michael Spence, the economics laureate, wrote about this in an article last year where you get a power law economy, which means that there's a small group of people that winner take most, that dominate the industry, that dominate the income, dominate the revenues, and a longer tail, like the one that Marshall showed earlier of people who are also contributing, but it's not nearly as clustered toward the middle as it is for other types of goods and services. And it's not just media and music, but Mark Andreessen has said that software is eating the world. It's coming to all those other industries, to manufacturing, finance, retailing, all the other ones I listed earlier, and that means that those economics of digitization are becoming more fundamental to every industry. It's transforming the economy. It is why we're studying the digital economy. There are many aspects of that, but on the theme of biotechnical change, let me point out that we have seen a real shift, not just in income between college and high school, but up to the 1%, and it turns out the 1% have their own 1%, there's the 1% of the 1%, that's what I'm charting here, the 0.01%, and the share of income in each of those groups has been going up recently to points that were last matched just before the Great Depression, if that's any source of comfort there. So that's another pattern we see in the economy. So let me just say a couple of thoughts about what's to be done, and then I wanna open up for questions and get your thoughts. We need to think about how we can use these technologies to create more shared prosperity, inclusive innovation, more abundance, but we emphatically do not think that the answer is to put the brakes on technology and stop the bounty that's being created, but it's to think about ways that we can have the rest of us catch up with those technologies. As we said in our earlier book, digital technologies change very rapidly, but organizations and skills don't tend to change nearly as fast, and because they're not changing as fast, there's kind of a mismatch there, leaving lots of individuals behind, and also lots of organizations behind. Jan Tinberger, a Nobel Prize winner in economics, put it this way, he said there's a race between technology and education, and for much of history, education was keeping up. We would broaden that, it's not just education and the skills of individuals, it's the knowledge and skills of organizations, or organizations structures aren't keeping up. Marshall was very good at illustrating some of the remarkable new types of organizational forms that are not just feasible, but that are optimal in a world of digitization, and they're not as well understood as they should be. Marshall's work is helping to pave the way in some categories of that, but there's much more work to be done, and even when they are understood, they don't diffuse broadly, most companies aren't really adapting to them very well. And taking it up a level further to society, we have to rethink our policies and our organizations and our institutions at that level as well. In the first machine age, we made some huge changes in the way society was organized. One example is mass public education, mandatory public education, that was introduced in the early 1800s. That was a radical idea at the time. Some called it socialist or other things that seemed kind of strange to do that, but of course, looking back, it was essential to that transition when we went from 90% of people working in farms to 40% around the turn of the 1900s, and then to less than 2% today in countries like the United States. All those people didn't simply become unemployed, they became redeployed into new industries in part because they had been reskilled, their education allowed them to do jobs that they couldn't have done previously, and there's also invention by entrepreneurs, people whether it's Henry Ford or Steve Jobs or Bill Gates or many others who helped invent entirely new industries that took advantage of this labor force. That's the scale and scope of changes we're gonna need to make going forward. So I would argue that the new grand challenge for us is not simply to make a self-driving car or a robot that can sit in a car and drive it, but it's to understand that digital technologies will continue to accelerate and our skills, our organizations, our institutions are lagging, and therefore simply doing business as usual, improving technology by itself isn't going to lead to the kind of shared prosperity that I think we want. We need to reinvent our economy and our society to keep up with this accelerating change. So that's very much what the initiative on the digital economy is all about is to take advantage of this huge opportunity that technology is providing us and think about how we can accelerate the rest of society to keep up with these amazing technologies. So let me stop there and see what kinds of questions and comments you may have on that. Thank you. So we got a question over here in the back. My understanding of the issues you've described around inequality particularly is there are two alternative hypotheses. One is the one you've described. I think we can also go back to Marshall's talk earlier about open platforms and people capturing value that way through power wars. The other is that value is being captured through rent seeking, whether through land, commodities, real estate, particularly in this country or patents. And maybe you can comment on that second bit, please. Yeah, I think there's a lot of truth to that and I've had some discussions. Joe Stiglitz in particular, we've talked a bit about that and he's emphasized that as well as the other ones. And I think that when I have discussions with him and others, I think we agree that there's multiple factors at work. There's those two that you mentioned. A lot of people would add globalization to the mix. Many people would add changes in unionization and tax policy. So there's a mixture of different things. I mean, my view of the evidence, I see specific jobs being automated and the demand for certain types of skills, especially routine information processing skills dropping quite precipitously and that's reflected in fewer jobs in those categories, but also lower wages for the people to continue to work. I see the ability of technology to scale what's happening in certain talent or luck or insight and be able to replicate that to billions of people and that's generating a lot of wealth for those people. So I see quite visibly the mechanism by which some of this is happening, but I also see some of the rent seeking and other factors and the patent system to pick one of the ones you describe, it's not something that was given to us by God or preordained by the laws of physics. It's something that we wrote down as a way to, according to the US Constitution, encourage the useful arts and sciences or something akin to that. And the purpose was to motivate people to invent and innovate more. If, as you suggest, it's having a counterproductive effect and hurting innovation or hurting the ability of society to gain prosperity, then it's not serving that purpose and we need to rethink and reorganize some of our intellectual property laws and I think it's quite likely that many of them are due for refresh. They were well designed perhaps for the era that they were invented in, but they're not necessarily appropriate, say, in the area of software, at least not the way they're currently defined and that's one example of how we might wanna reshape our organizations and institutions around privacy, around security, around other aspects of the economy so that we continue to benefit from the technology but we can't expect that we don't have to make changes anywhere else, that would be a miracle. If the cost of information processing falls by a million fold to not have to make changes anywhere else would be very unlikely and in fact, not true. Take another question now, here's one right here, Jonathan. I enjoyed the talk very much indeed. Can I just press you a little bit on the relationship between skills and all of this? Between skills? Between skills and all of this. One's reaction is, of course, in this kind of economy where you've got plenty of opportunity for skilled people, we need more people to get to school and so forth. On the other hand, some of these superstars, these rock stars and movie stars and this that and the other, I'm sure some of them are very well educated but not all of them have PhDs from MIT and all that kind of thing. That seems to be an exact counter example, whereby actually the very thing that they would not have wanted to have done was to go to school and this that and the other, they would go out and do their stuff. So can you just say a little bit more about how you see the skills and education issue given it's so nuanced, I think? No, this is very important and thanks for giving me a chance to elaborate a little bit on it. Many of those billionaires not only didn't get PhDs but a large number of them dropped out of their undergraduate institutions and still managed to do quite well and I think that underscores the fact that there's not just one kind of bias technical change, there are at least three and they're having different effects and different parts of the income distribution. I think what's going on with the billionaires, the top 1%, the top 100th of 1% isn't well described as skill bias technical change. There's something else going on there. The ability to replicate talent globally I think has more to do with that or luck or whatever it is that you want to describe that those people are bringing to the table. Something, some special quality and even on the skill bias technical change, I showed you the chart and one thing I didn't really emphasize was that while it had fanned out, the last 10 years or so it's kind of been level. I don't know if you remember how well you remember the chart. I won't go back to it as a few slides back but it was kind of leveling out and what that means is that skill bias technical change doesn't seem to be getting worse. It's still at a all time high as it was a few years ago. It's sort of been staying at that high level but that may not be the main story and furthermore, as one of my charts suggested, the nature of the skills that are automatable or substitutable versus complemented are constantly evolving and changing. When you hear things like machine learning and other things, I think Andy may get into some of the more recent developments. You can see that the types of things that were automated in the 1990s aren't the same things that are likely to be automated now. So skill bias technical change is evolving rapidly. There are other things going on especially at the top of the income distribution. It's not a simple one story solves everything. It's one that requires a research agenda to be frank. Do we have time for another question? David, are over here, yes? Hello, yeah. I was wondering if you've expanded your research to some international aspects. In particular, I was wondering about China and India where it seems to me that the educational segmentation is probably even greater than the US and maybe the social issues are many times more serious. I share your impression. And one of the things that we've been doing is expanding that agenda. And one of the reasons we're here outside the United States is to start getting some more international perspectives. It's partly data constrained. Thankfully, Roberto and others are giving us better data in other countries. To some extent, it's the old story of the drunk in the lamppost that we go where the light is and the illumination and we get a lot of data from the United States. We'd love to have better data from China and other countries so we could analyze this more in depth. But one kind of data you get from just visiting the places and seeing what's happening on the ground and we've done that and we've gone to those factories. And you may have also seen there are lots and lots of people working side by side, working on lots of tasks there. Far more labor intensive obviously than most American or British factories which are often essentially lights out, no labor. But also, I think, much more vulnerable to Baxter and that kind of automation because precisely because they're doing a lot of these simpler repetitive tasks. To be quite economic about it, the strategy of having lower wages is very effective for competing against Germany or Britain or the United States. If you've got workers that are essentially human and they can do the task more cheaply, you can do some labor arbitrage and beat your competitors that way. But that strategy of low wages that helped bring many Chinese out of poverty and get them a lot of the big share of the world's manufacturing, isn't a particularly good strategy for competing against Baxter or other robots that are getting twice as efficient every year or two. I wouldn't wanna be competing with Baxter based on lower wages. And that makes them in many ways much more vulnerable and Terry Gao, an entrepreneur, is seeing that kind of transition there. And ultimately, the last thing you said, I just wanna underscore, it may be that the second machine age has an even bigger impact on some of those countries, China and India or others, because they're gonna have to make that transition to the new kinds of skills that are needed in the second machine age even faster. They, in some ways, start at a different point than the more developed countries. And I think we could get one more question. Who are you pointing to? Right here, okay. Thank you. No, I see very well what you are explaining, new technology making certain job obsolescence, right? And then you need to really blow it. I see all of that. My question is, is there also something more profound going on due to interconnection? When you interconnect many agents' system together, you have typically the emergence of new behaviors at a different scale, and so I would like you to comment on that. Yeah, thank you. That's one of the most important phenomena that's going on. I touched a little bit on digitization. Another thing we talk a bit about is networks and the fact that we now have not millions or hundreds of millions, but literally billions of brains connected. And they not only can access, Wikipedia and all the other knowledge sources, but perhaps more importantly they can contribute to it. And they can share ideas with each other and create new kinds of organizational forms. Marshall mentioned Uber and there are a whole lot of other things with mobile phones, for instance, are enabling in the developing world. The word emergent behavior is a good one because it's a different kind of economic organization than the traditional market or traditional hierarchy. It's much more of a networked form of organization. And a big part of our research agenda going forward is to understand better these different kinds of organization. And we don't think there's just one. We think there's multiple different kinds that are being invented and evolving quite rapidly and it's a new kind of economics, but the network aspect of it is I think ultimately going to be at least as important as the digitization and as the artificial intelligence aspect of it. So I'm looking at the clock here and I'm afraid I'm out of time. It's been really good to have a chance to do this Q&A. I enjoy that very much. Another great pleasure at MIT is interacting with the incredibly smart colleagues and you've heard from a couple of them. The one who I've been spending the most time on, revealing my preference of who I've really been able to work quite well with, is my colleague and co-author and co-founder of the Initiative on the Digital Economy, Andy McAfee. He's not just one of the smartest guys that I know there, but also selfishly, he makes me feel smarter when I'm with him and that's always great to have in a colleague and I think he's gonna make you feel smarter when he shares with you some of his insights. He has the ability to cut to the core issues and really illuminate him in a way that even I can understand and I think you guys will be able to understand as well. So please join me in welcoming Andy McAfee.