How the Industrial Revolution invented modern computing

Who's in the Video David Alan Grier David Alan Grier, PhD, is a professor, writer, author, and speaker on issues of technology, society, and organizations. He is the author of several books including When Computers Were Human,[…] Go to Profile Part of the Series Full Interview Explore series

This content is for subscribers only.

Become a Member Login How the Industrial Revolution invented modern computing “The process of systematizing, correcting errors, finding approximations, and making them work as civil systems that was what really drove me to start looking at human calculation and what was the foundation that it laid for the modern computer age.” ▸ 1:23:36 min — with David Alan Grier Description Transcript Copy a link to the article entitled http://How%20the%20Industrial%20Revolution%20invented%20modern%20computing Share How the Industrial Revolution invented modern computing on Facebook Share How the Industrial Revolution invented modern computing on Twitter (X) Share How the Industrial Revolution invented modern computing on LinkedIn Sign up for Big Think on Substack The most surprising and impactful new stories delivered to your inbox every week, for free. Subscribe

Before computers existed, people performed massive calculations by hand where error, repetition, and standardization shaped the outcome. We tracked comets, mapped nations, and solved problems of scale. 

That legacy of manual calculation shapes how we live today;  our modern algorithms and the shaping of predictive models. Dr. David Alan Grier explains the unexpected link between the Industrial Revolution and artificial intelligence.

DAVID ALAN GRIER: I'm David Alan Grier. I am currently a writer and author on issues of technology and industry and things of that sort. In the past, I have been a computer programmer, a professor, a software engineer, president of the IEEE Computer Society. I am the author of the book "When Computers Were Human" and also the book "Crowdsourcing for Dummies," among others. [typing] Chapter 1 - Computers and the Industrial Revolution Why is computing part of the Industrial Revolution? The Industrial Revolution is about systematizing production. And it's about producing goods of uniform quality, if not uniform design, at the lowest possible cost for the largest possible market. If you want a date that's easy to remember and just nails things down, you go with 1776. And that's useful for my purpose as a writer, because that's also the year that Adam Smith's "The Wealth of Nations" is published. And the start of that book is about the description of industrial processes, how we came to them, and how we used them to start building uniform products that would have large markets. That would increase the wealth of nations. And those first chapters deal with the division of labor, the specializations of tasks, and the systematization of work. That book was highly influential, not only in the industrial group, particularly in London, in the cotton producing and in the pottery producing fields in northern London, but also amongst the scientific crowd. Because they also had things that were large problems that needed systematic approaches. Astronomy was the first. We had used astronomy for navigation, but there was a question of what was out there and how did things behave that was being addressed by people purchasing telescopes, or financing telescopes, and then setting up a staff to collect observations. And they would night after night go to the observatory and map the heavens. And in the process of mapping the heavens, it doesn't take long to realize the data problem they generated. Suppose it takes a minute or two a night to get one star located. Okay, that means you can get a couple hundred, maybe even a low thousands, of stars a night. But the figures that you record depend upon the hour of the day which direction the Earth is facing, or your telescope is facing. And the time of year, where the Earth is in its orbit, you'll get different measurements of the stars at different times of year. That means you create a massive pile of data that needs to be reduced to absolute coordinates, to a location somewhere fixed in space. That requires a lot of work and a lot of arithmetic. And it required these observatories which could do the recording with a staff of two or three astronomers to have a large group of people to help them reduce these data points to something that was absolute so they could start fitting them into their map to the heavens. That was a repetitive job. There was a lot of it. And the issue that they all faced was how can you do it for the least amount of money. Part of what we think of as high tech in computing and programming is also the task of systemization, of regularization, of taking a complex thing that could be done many different ways and putting it in a form that can be marched through in a fixed series of steps. At base, no one really needs to know the return date of Halley's comet. There are certain few scientists for whom it helps explain the universe. But it's mostly just a reminder that this object has been seen every 75 or so years throughout history and has been recorded as such. But at the end of the 18th century, a group of French astronomers, knowing that it was coming, asked could we figure it out? It was a test of the science that had been developing before them, of the theories of people like Copernicus and Galileo. And the question was could they get a date? Could they get the date that was closest to the sun? And could they do it mathematically? And that's actually a tough problem at some level even now, although we have all the programs to do it, because it involves the location of several big oblogics and needing to move them through space while you're tracking the comet around the solar system. And what they did was divide the labor. And that becomes a key theme in computing, a theme that was worked out by human beings and relatively modest mechanical devices before we started putting electronics to it and rushing off into programs and artificial intelligence and all the rest. We worked out the problems of computing because it was divided labor. And for that first return of Halley's Comet, they had two people working on the location first of the Earth and second of Jupiter, because Jupiter is a major influence on the motion in the solar system. And then another astronomer tracked the comet, which is pretty small and has little impact in terms of moving other things around in its orbit. And they figured out how to do that and do that repeatedly, repeatedly and in a way that they could double-check their work. And that's sort of another one of the themes, that it's not just the brilliant insight, it's not just the algorithms. It's figuring out how to find mistakes, how to find mis-additions, miscalculations, and even misapprehensions, misunderstandings of what's going on. That process, which took about a month, really set the stage for the things that were to come. In particular, it set the stage for the nautical almanacs. As I said, you and I have lived all of our lives without knowing the next return date for Halley's Comet. And our lives will go on. And when it comes, it'll be a big party, and I hope it's better seen than the last time when it came in the '80s. But it's just a party, it's just watching an odometer flip. The bigger problems, the things that involve production in terms of our living our lives, of feeding ourselves, clothing ourselves, providing shelter, we need to produce goods and services. And the key thing that was involved there was trade and ocean-going vessels. And knowing where they are. Once you're out of sight of land, how do you know where you are? The ancient astronomers, pre-17th century, had a lot of various methods that were somewhat ad hoc that sort of involved, "This is spring, we know this star is over our destination, we go out, we point the ship at the star, and we pray that we get there." They didn't know where they were on the Earth's globe. And that required an understanding of how to compute longitude and latitude. And latitude is fairly easy to get, at least in the Northern Hemisphere. Longitude is a lot harder, and it requires knowing where stars are relative to a fixed point on Earth. And that knowledge, you have to codify in a book, and you have to do it multiple years in advance, because you give these books to captains and turn them loose across the oceans, far away from wherever that fixed point might be, which in the early days of exploration were one of three points, London, Amsterdam, or Paris. To produce those books accurately, you needed a system, you needed a system that allowed you both to do the calculations, and then undo them in a way that was different, because one of the very early pioneers of calculation, Charles Babbage, discovered what he called Babbage's Rule, which is two calculations done the same way by different people will tend to make the same errors. There seems to be, in the process of hand calculation, mistakes that trip up everybody. Not all the time, but there's a tendency that if one person makes the mistake, the next person will make it. So you need to approach the problem in a different way, and in particular, in a different way that exposes errors. And that was thrashed out in the late 18th century in the nautical almanac in London, and in a comparable publication in Paris. They figured out not only how to divide, because there were multiple objects they wanted to get the locations of, so they divided that amongst a variety of people, and then ways of undoing those calculations in a form that exposed the errors. And that was a key innovation on their part. And that led to these publications, and that leads to the age of exploration and the start of global trade, global oceanic trade, trade that requires boats to go out of sight, and that of course leads very directly to industrialization in the modern world. Same was true with surveying. In the United States, there was, by the early 19th century, there's the problem of figuring out where the United States actually is, whether it's bounds, whether it's limits. And that involves surveying first the coasts, and then moving in. And that again required, they used nautical almanacs, again, to figure out the locations of places. But a great deal of calculation, because you're basically laying down across the earth a bunch of triangles. And you can figure out where two of the corners of the triangle are, and from those two corners, you can get to the third, and then from that third, that gives you a new triangle, and you can keep marching off across the land. That, particularly when they got to California, proved to be a hard thing to do, that it was very difficult to get all the calculations you needed to get that vast space well surveyed. And in particular, that was needed because you had a bunch of people who thought they owned land there somewhere. And by thinking that they owned land, they had to know where it was. They couldn't say from this rock to that mountain, to that tree. They had to have more exact points. And the treaties that shaped the West and shaped the United States basically required the new states to figure out land ownership and the location of land. And so they all faced the same problem of how do you take large amounts of data and process it, and that required them to think industrially, to think how the two pieces went together. And so there evolves other systems to try to do that quickly, to get first approximations, to be able to get numbers close enough so that the errors on borders are relatively minor and can be worked out in local negotiations. But it's that process of systematizing, correcting errors, finding approximations, and making them work as civil systems that was what really drove me to start looking at human calculation and what was the foundation that it laid for the modern computer age. Those lessons were first approached in largely the 18th century and by people doing work by hand and doing it largely but not exclusively for astronomy or surveying. One of the aspects of building systematic processes around numbers, around calculation was that very, very quickly people started asking, "What can you do with machines?" The idea of an adding machine, a machine that could add two numbers, goes back to discard. It has a long, long history. And the idea that you can represent numbers by the turning of a wheel and then if the wheel turns all around, it turns the next wheel, that was well worked out early. However, it really wasn't part of a process. So something that you could rely on. And when you talk about industrializations, you're basically building processes that can be done by people who at base don't know what they're doing or more accurately. They are doing something that they gain a skill at, that they understand their steps and they learn how to do them efficiently, but they don't understand necessarily the science and the ideas behind them. At the start of the 19th century, there's a great explosion in the interest of what machines, levers, gears can do. And there is a study of linkages, for example, of mechanical arms that connect together. And what they could do that, I don't think it's quite lost. I'm sure there are mechanical engineers who will give me a lecture on this. But it's something that has certainly vanished from our daily contemplation and contemplation even of non-specialized students. And one of the people who got fascinated with this was an Englishman named Charles Babbage. He's a start of the 19th century. His father was a banker. Because his father was a banker, his father was also involved in the Caribbean trade. And that meant he was familiar with boats going off to the Caribbean to collect largely sugar and bring it back to the United Kingdom. And it's at a point where systematized shipping is starting to really take hold. It really doesn't kick in until the start of the industrial age, where boats run on a schedule. And you know the schedule and you know more or less the date they're coming. Babbage attends Cambridge. He learns astronomy, gets fascinated with it, fascinated with mathematics. And then starts pondering, how can you mechanize mathematics? What can you do with it? He had written a couple of things while he was in college that in many ways was filling up spaces and filling up shapes with smaller versions of itself to see how you could approximate them. And what systems would work and what wouldn't. And sometime during that early period, he got involved with an article almanac. He ends up being on an advisory board for it. And this is when he discovers his rule that two people doing the same calculation the same way tend to make the same mistakes. The second job that's starting to come up and he gets involved and interested is the understanding of insurance, of probability behind it. In particular, insurance is at some level a savings account, nothing more than that. And it's savings that you get to cash in the savings account when you die or if it's a health insurance, which is later in the century, when you get sick. That means the people running the bank need to know how much money they are likely to pay out. And the understanding of that, the mathematics of that, the mathematics of probability, is just really starting at, again, it's an industrial revolution start, but it's really starting to take hold in industrial life at the start of the 19th century. And it involves the creation of another kind of table called mortality tables, which tell you how long someone is likely to live given that they've lived this long. And they are highly dependent on data and on what people. You don't do them for a huge population because there are differences in the way people live, in their diets, in the work that they do, in the kinds of families that they have. So you tend to do it for smaller groups and that means you have to do a lot of these tables and you have to process a lot of data. Babbage pondered it and he began to realize that for both astronomy and for these insurance tables, there was a common kind of calculation that could be very useful, but was by the standards of the time, very time consuming. And it's basically fitting a curve to data. You have these points and you want a curve to go nice and smoothly through them so that you can make estimates between the points and you can go project out beyond the end of your curve to figure out where it might be going. And it dawned on him, on Babbage, that that calculation was in effect could be reduced to a lot of additions, a very large number of additions and subtractions. And that meant that he could chain together a bunch of these adding machines and produce a machine that could indeed do that by just grinding away at a crank or as he thought of it, be produced by steam. He was working in the very early age of railroads and steam engineering and he saw his machine as very much in the heritage of locomotive design. And so he produced a machine with very large gears and numbers and the dials on it were made out of well machined brass. And the process was probably beyond the engineering ability of his time. His lead engineer pushed British engineering substantially and advanced at both working for Babbage and later. And it was also probably the wrong concept to build from. Machines are very often done as metaphors. We build a machine to do something like this. And Babbage was building a machine to do calculations like a railroad engine. And railroad engines were not completely safe and secure then. And they were still learning a great deal about how to build them. Babbage never got his working. He built a number of models. He built several that demonstrated the proof of concept. He could fit not the curve he wanted, but a much simpler curve to data and in a machine that got through the calculations we know at least once. And then he built a lot of rough parts for the machine he wanted to build. And because of that Babbage and his ideas left a legacy about what they were trying to do more than what they actually did. What he was trying to do was systematize calculations so he could handle large scale in both cases calculations, fitting curves to data and things of that sort. He had contemporaries that were intellectually in terms of direct connection, far more important, George Bool. George Bool wrote a book called On the Laws of Thought. We now know those laws of thought as Boolean algebra, which is the fundamental tool underlying all of the analyses that design electronic circuits and computers. That connection has been well known and understood forever. Babbage vanished for a bit and also the exact intellectual connection is not as clear unless you look at the systemization, the building of an industrial process, the work to lessen the cost of computing and to move it into a bigger environment. In that sense he had a tremendous impact and remains an important figure to this day. Almost 30, 40 years after Babbage did it, a Swedish father and son built one of these machines using Babbage's idea and built it and it was fully functional using clock technology. And clocks are much smaller, they're much easier to work with at that scale, they require less energy and there were lots of standardized parts they could borrow from. In particular, escapement clocks, which were the dominant technology until not that long ago, have an important role in science writ large and in computing specifically. Clocks were one of the first sophisticated technologies that had wide distribution in the 18th century. The escapement clock, clocks with a pendulum with a little gear and a little thing that goes click, clock, click, clock, tick, tick, tick, tick, tick. Technology that were novel, they were often packaged in ways that made them in effect a luxury item, something that the rich would have. But very quickly they became common and they allowed industrialization, for example, by setting times when a factory would be open. You don't need a clock if you're running a farm necessarily. There might be a few things where knowing an hour or knowing a minute is useful. But for the most part, you're following natural rhythms of the day and those natural rhythms have a very wide margin of error. You can miss them by 15 minutes, 20 minutes and you will be just fine. If you're setting up a factory, if you're bringing people together to work together, to collaborate together, and if in particular you're dividing a task up into a linear process, which was one of the very first things that happened. Adam Smith writes about making pins and needles. And in particular it's a process of cutting the wire, sharpening the wire, putting a head on it for a pin, putting it in the paper that you're going to sell it with and moving it on. It's a linear thing. You have to do one step before the other. And if you're dividing it up across different people, it means you have to have people for all the steps. And that means they all have to be there at the same time. And you really needed clocks to be able to do that. Scientifically, it made a lot of measurements more interesting, in particularly the measurements of location. You needed that to determine longitude. You needed that to determine location on the earth. Because of that, it became an important technology for disseminating ideas, for bringing them to people who became fixated with clocks with how they worked, of how they could improve them, of what they could do with them, of clockwork mechanisms that became part of other technologies. In computing, the clock, the tick-tock of the escapement ratchet, became part of computing because it became a drumbeat that stepped through the basic mechanisms of calculating devices. As you would go, you just wouldn't let your machine run wild. You would have a device that was going tick-tock, tick-tock, tick-tock, that was controlling the other elements, the other parts. And that allowed them to think about machines that could do calculations in a systematic orderly way. It gave them a method to control them. But more importantly, it gave them a technology to build off, to think about, to say, "Okay, we know this works for doing time. How does it work for collecting data? How does it work for timing additions? What can we do with it?" And so these clocks, clockworks and clock mechanisms are deeply imbued in industrialization, deeply part of calculation, and part of the tools that people used to think about how to build the next generation of devices and how to use them. Chapter two, the power of standardization. Standards are a key part of the industrial world. Building things to standard models, standard parts, goes almost all the way back to 1776. The one that we usually point to is the Remington factory in Connecticut and the tools that they used to build standard firearms. But the real standardization process, the one that has created the modern age, is much newer than that and has a much broader scope that involves the mathematics of calculation, that involves data much more profoundly, and in turn influenced how we calculate and how we work with data. The fundamental goal of all of this is reducing cost, reducing effort, and also transferring ideas to the least well-trained individuals. If you're going to make large-scale data processing work, you need to have people involved who are not as educated as yourself, who can deal with a part of it, who can follow instructions. If you look at the original calculations done for nautical almanacs in the late 18th century, the 1780s and 1790s, they are done by individuals who would lay them out on a piece of a paper, each used their own method, each had little tricks for identifying things, for remembering the exact calculations they were doing. There would be a collection of additions and subtractions, a few multiplications, and if you could at all avoid it, no long divisions. There is a collection of them that belonged to one of the human computers, as they were called, that worked for the British nautical almanac who lived in the United States, and he would do his work here and send them across via a boat to London, where they would be incorporated in the nautical almanac. He had his own way, and he also had a rather cramped hand. He could not write on a straight line. And if you go over his materials, you have to be very careful to make sure you're on the right line of the calculation, and you haven't jumped to the wrong one. He could make that work, and he was quite successful at it. Not everyone could decode that writing. By the start of the 20th century, you see people doing it on graph paper with little squares so that all the rows line up. You start seeing people identifying the operations line by line by line, particularly in the mid-19th century in England when they were doing title recordings. We did those here as well, but they were ahead of us in systematizing them. We did that a little bit later in the 1870s for weather. Weather requires lots of data, because if you look at the weather channel, if you look at your phone and watch the clouds move across, there are clouds that will hit one part of our metropolitan area that will miss other parts, and they won't be that different, that far apart. So there's a lot of data you have to have to be able to get accurate weather forecasts beyond just the broad statement that it might get this hot for this metropolitan area on this day. When we set up a network to start doing that, again, we had people collecting data. There was more data to collect. It was a little more subtle about how you get rainfall and precipitation and humidity and required a certain amount of arithmetic, and it required that you be able to work with people who had not gone to college, because you weren't going to get enough of them if you restricted yourself to just college graduates. And that forced the weather people and scientists as a whole to ask about standardization. How do we follow the industrial patterns and move things into ways that are always done the same way, that always use the same methods, that always produce the same results, and that anyone can interpret? In terms of the modern sense of standardization, I really pin it on World War I more than anything, and if you wanted one individual who was more responsible for it than any other, who was a major leader, it's Herbert Hoover. Hoover was an engineer. He was a mining engineer, which means the big thing he's building is tunnels and walls and things to hold up tunnels and concentration plants, which are basically pools and other big things. But he also grasped that engineering methods could be applied to other products if there was a standardization. And he worked in an era when things that we take for granted, like bolts, you know, you have a nut, you have a bolt, you put them together and you can screw them, and they work because they're the same. We're in standardize. You buy them from one manufacturer, they're not going to fit from another one. And he argued that there were a large number of things that if they were produced to standard forms, would increase the scale of our industrial processes, would allow us to make more products for more people at lower costs, and expand our ability to do things. And that led to a variety of institutions, the National Institute of Standards, then called the National Bureau of Standards, being one of them. And private standard organizations, my former organization, the IEEE, the Institute for Electrical Electronic Engineers, has a big standards group that deals with electrical standards. There is the American National Standards Institute, which deals with a lot of other standards, including a lot of standards involving computers. If you have everyone going off and doing their own thing, that can be very good, because it means you're exploring and figuring out what works and what doesn't work. But at some point, you've got to start focusing on production, at least for certain problems. And in those cases, you need standards. And as the computing age built, they more and more started looking at standard ways of doing calculation, standard ways of expressing algorithms, standard languages for expressing programs. And all of these became a key part of computing that would have been a lot more expensive and a lot slower if it hadn't had the standards there. In our modern age, one of the things that companies do, that individuals do, is ask, "How can I use these to solve my problem? How can I build and extend and expand them?" So at this point, it's so deeply connected with computing that there's very little you can do without understanding them. Yet at the start of the 19th century, they were at best vaguely understood and vaguely used. We live in a standardized world and we live in an industrialized world, and we industrialize lots of things. And one of these things that we do, we produce at large scale and low cost, is education. Prior to the early part of the 20th century, really in the late 19th century, there were some efforts, but education was entirely a local affair in the United States. It was something that a lot of people weighed in and provided books and sample curriculum. There were publishers primarily in New York that did textbooks that were commonly used, readers, for example, that helped kids grow up. But if you think about it, that outside places where there was large and easy transportation of children, in an age before there were school buses and paved roads often, classroom was one room teaching. It was a group of kids together with one teacher who was there for a short time and did the best they could and moved on. 20th century starts to move that forward. It does go back into the 19th. But Carnegie's work really comes in the 20th and it's largely dealing with higher ed, with colleges. And I suppose technically he's Carnegie. This is the founder of what becomes U.S. Steel. In colleges, we have a standard amount of education that'll get you a bachelor's degree. 120 for a bachelor of arts, 146 hours of instruction, that'll get you a bachelor of science. You take a certain number of classes. The classes have to have a certain distribution. They have to meet for so long. These standards were developed out of work done by the Carnegie Institution. He funded studies of a wide variety of subjects that asked what's the best way to teach them. The most famous is the 1910 study of medical education. That's the Flexner Report. The Flexner Report went and looked at how medicine was being taught to doctors. What kind of things were being taught? What were the different practices? What were the best practices? And how can we produce the greatest number of doctors with uniform education at the least cost? And in the middle of that report is a line that refers to how we must devote our education in medicine to people who can devote their entire careers to medicine. And a lot of deans flagged about it at universities and said, "Oh good, we don't have to worry about women anymore because women take time off. Women have families, women have other duties." And prior to that, there were roles for women in medicine. They dealt with what you might expect, babies, pregnant women, and dying old people. But they were part of the medical community and the Flexner Report shoved them out the door. And if you go to the early age of the suffragette movement in the United States, the women who picketed the White House, the women who wore the yellow sash, the women who lobbied state governments for the vote, they were overwhelmingly the daughters of doctors because they had time, they had money, and they had a grudge because they weren't going to get a position in the medical community, perhaps other than being a nurse. And what happened after the Flexner Report, this is wandering away from standards for a moment and back to computing. The sciences looked at women who had been studying medicine, who had been part of medicine, and largely said, "Your nice people, please go away." Chemistry said, "We can't have women study chemistry. It will hurt their ability to have children and probably hurt the men's ability to have children too." Biology was not the field that we know it today, nor was physics. Mathematics, being somewhat desperate for enrollments then as now, said, "Women will take them. They can always be high school teachers." And in an instant, the teaching of high school mathematics becomes feminized, that women see an opportunity and they grab it. And they also grabbed jobs and calculation and other things because medical education was being standardized. This applied to other forms of education as well. But it produced a standard form of a university that we in the United States understand and know how it works. But the roots of those standards, there are a number of other organizations that have contributed to it, were in a series of reports that the Carnegie Institution put together in the early part of the 20th century. In fact, there was a time that the college course credit was called the Carnegie credit. That has largely passed from memory. But there were institutions that said, "We will become richer. We will be more productive. We can give more to more people if we do it in standard ways." And of that, Carnegie was important, Hoover was important, but so were a lot of other organizations. And computing benefited in it because without it, it scribbles on a page. With a standardized sense, you can start getting people to do it, who are not trained in it, who have not the education that the directors of the project have, and that greatly increased the power and the use of computational methods. Chapter 3, Computing the Human Experience Right now, where we're sitting in Washington is the old patent office. And back in the 1850s, that was the place the first telegraph was connected in the United States. It was from here up to Baltimore. And the first message that went across it was the famous "What hath God wrought?" And it was just people who were madly in love with technology, going, "My God, we can talk to Baltimore, and it's like it's right there, not whatever it is, 25 or 30 miles up the road." We don't remember the second question. The second message was, "What time is it there?" Because at that point, every city had a local time that was measured by when the sun was directly overhead at noon. And Baltimore is about seven and a half minutes ahead of Washington. And one of the things that it told the scientists who were in the room, and it was a political group, but there were also a number of senior U.S. scientists at the time, was that you could collect a lot of data about where things were by measuring time differences. And they immediately set up a process, and this again points, I think, to the industrial connection. A process for determining differences in time between cities that could be done by people who basically were just following instructions and really didn't know what they were doing beyond counting clicks on a clock. They could sit there and they could ping messages back and forth. They did not use the word ping, I should say, at that point. But send clicks back and forth, count clock motions, and at the end of the day, they would write that down. And that piece of data would tell you the difference in time. That was used very quickly to collect lots of data about time differences all over the East Coast and very quickly over the United States. And again, it had to be systematized and organized to figure out how it worked and what were the differences, and how could that be used for commercial reasons. And the big one that started to come was, of course, railroad trains. Because most places in the United States in the 19th century, you had one track between here and Baltimore, or one track from Baltimore to Philly, and you had trains running both directions. And if they came together, that was bad. They had places to pull out, but how do you know someone's past you? How do you know if you're reading the time schedule right? And getting a unified time, which was in place by the late 19th century, helped eliminate a large number of train wrecks and delayed trains and other things that disrupted production, that disrupted the transportation of goods and services. So all of these things are deeply, deeply imbued in computing, and computing is so tied with the issues of production, of getting things done quickly, easily, systematically, and of capturing human experience. In the early days, the human experience was largely capturing natural experience. What's the time difference? Where's the locale? When does the sunrise here? When does it set there? But by the end of the 19th century, they're starting to look and capture human data, human activity, the kind of things that people did and worked and behaved so that they could use that for making policy decisions, for making marketing decisions, for deciding how much to produce. That became the interesting thing. And again, as just you went from doing one comet to doing nautical almanacs to doing mapping the sky and to mapping the earth, this was another big jump in the amount of data collected. The census was done every ten years, it's in the Constitution, and for the first hundred years or so, it was largely an office that appeared every ten years. They hired a bunch of people and told them to run around and count, and then they got all the stuff back in Washington, tabulated it up into how many people were in each city and each state, and published it, and it was straightforward and fairly easy. By the 1840s, they're starting to ask questions, how many people live in your household, how many men, how many women, how many children. They began looking at issues of public health, of disease, of general healthiness. Landowners, what kind of work people did, their ethnic background, certainly a slavery built, the problem over slavery, built in the 1850s, that became an issue. And that meant we were collecting more and more data, and the censuses took longer and longer to do, and by the 1870 census, that takes most of a decade to complete. The 1880 census is never really finished. It still isn't really finished to this day, there are publications, but there's data out there that we haven't officially worked with. And as 1890 began to approach, the head of the census said, "We've got to do something. We've got to systematize it. We have to industrialize it. We have to reduce the cost." And so he put out a request for someone who could build machines that could count people. And the process that came out that won the contract formed a company that was known then as the Hollerith Tabulating Company, becomes something we all know a little better, IBM. And they proposed one that would send people out to the field, not with a pad of paper and something to write things down, but with a punched card and a little machine that she would punch holes in it with. Those cards could be taken back and they could be put on a machine, and the machine would count up how many women, how many men, how many people over 45, how many people living in this city. And it sped the process so greatly that by 1893, the preliminary census was completed, and it was being widely advertised and widely used. 1893, there was a big world's fair in Chicago, and the results and the machinery that produced those results were on display and of central exhibit at that fair. People were talking about the results in particular. At that fair, there was a historian who got up and said, "By reading the recent census reports, I can conclude that we no longer have a frontier. America is a settled country." And that produced a change in perception of ourselves, and that we only reached it because we were able to handle that scale of data. They were, as is so commonly the case, in love with their machines. They were very proud of them. They were very proud of what they could do, of the results that they could obtain, and the speed with which they were able to tabulate the population of the United States, to the point where they started falling in love with the punched cards. There is an assistant director of the census who wrote this stunning document about how he could hold up a card and he could see the person in the card. Now, I have no doubt that people got used to them enough and said, "Yep, that's the manhole, that's the womanhole, that's the kidhole. I can quickly scan and guess what kind of person this is." But it went beyond that. It was an enthusiasm, as I said, that's common with new technology, that he would hold it up and he felt that when he was putting it through the tabulator, he was working out the future of that person. And the bell which indicated that the card had been processed was the bell calling that soul to judgment of heaven or hell. It's sweet and it's endearing, and it also shows how as we start approaching our life through data, through the large processing of data, we start seeing ourselves in new ways and we start seeing something that we couldn't see before. We couldn't see before, we get a hope, an intuition, and that that intuition provokes an imagination and a love that we didn't have before. Now, that often comes crashing down, but nonetheless, there was a time of falling in love with large-scale processing that came from seeing ourselves in new ways that were driven by data. Chapter 4 - How Computers Change Us How long have we been going? I'm a professor, I can talk for hours. In the 1940s, there was again a lot of computation that was needed for the war, in particular bombing, hitting things, and specifically shooting down aircraft. Shooting down aircraft is basically like duck hunting. The hunter must search his lead and aim ahead of the duck, if he is to hit it. You want to get up there, you want something to explode near the aircraft, and if it explodes near the aircraft, you get a hit. And unlike duck hunting, you don't set your dog after it, but the effect is quite similar. But because of the great altitude and speed of a bomber, the anti-aircraft gunner cannot rely on dead reckoning. His leading must be a careful mathematical calculation. That's a hard mathematical problem, because you've got a moving aircraft, you've got a gun that is probably moving on a ship, or if it's on fixed land, it's swinging, it's moving back and forth. You're computing part of an arc, which makes it harder. They had methods from the First World War where they could estimate what the whole curve looked like, but they were hitting something on the ground or a ship that was fixed. And that was easier. Shooting things down out of the air is tough, and getting where you want your shell to explode is tough. What was then the Army Air Corps invested a lot of money in figuring out how to do that, and in particular at the University of Pennsylvania. They employed a couple of machines that had been built there to simplify it and just produce reams and reams and reams of data that were taken to another site and reduced into instructions that you could give to pilots, to gunners, and other soldiers. The group in Philadelphia, in working with their machine, which was what we now call an analog machine, it used amplifiers and tubes to draw an arc that could have been drawn on a screen. It was usually drawn on a piece of paper. One of their numbers realized that you could do it more generally by doing the actual calculations. You could draw the curve by stepping along the line, and from that you could do more than just do the curve for a shell. There are other problems that you could solve with it. They pitched this to the Army and set up a process to build this machine, and this is a key moment in American computing history. This machine, which would get the name E-NIAC, Electronic Numerical Integrator and Calculator, was a very direct precursor of the modern computer, and it operated in a substantially different way, but nonetheless a very large number, the first generation of computer engineers, got their training on that machine. And by first generation, I mean really the next eight to ten years. In doing that, one of the leaders was a mathematician from the University of Michigan, Herman Goldstein, and he, one of the train stations around Philly, noticed a scientist that he knew from Princeton named John Von Neumann. Von Neumann was at the Institute for Advanced Study, which is an institution of higher ed that has no students, takes no tuition, gets famous people to come and just think. Einstein was one of the people there. He went up and introduced himself, said he was working on calculations, and Von Neumann said he was very interested. Could he come and look? Von Neumann was a mathematician very skilled, could very quickly grasp ideas. He had been interested in calculation for a long, long time, and had studied some of the computing groups, particularly the one in London, those doing the Nautical Almanac. And with the leader of that group, the two of them had sat down and worked out things that were really the precursor of what a program would look like. What did it mean to program a device? What did you systematize calculation to a set of instructions that would always be performed the same way? Von Neumann goes to visit the ENIAC, which is at the University of Pennsylvania, and gets very excited about it and starts thinking about how to expand it and develop it. And wrote a report that has since remained the controversial beginning of the computing age, because it defined what a modern computer was, and it left off the names of all the people who had been working on the ENIAC. Von Neumann abstracted what had been done there. And it dawned on him that the machine that they were really trying to build, not the one that they were building, but the one that had the most flexibility and the most ability to do things, had three elements. It had a place where you could store numbers, a scratch pad, if you will, memory that we now know it. It would have a processing unit. Didn't describe what that processing unit did. At the time, it was generally assumed that it did addition, but it could do other things. It could do the other arithmetic operations, or it could do other operations still. And that there was a third element in it, and that was a program decoder. That is to say, something that would read signals from the memory, figure out what those signals meant, and send them either back to the memory or send them to the processing device to do some activity. These three elements in multiple forms are part of every single computer we have today. Some do arithmetic, but not all of them. Some of them do just various forms of symbol processing. Some do a variety of other things, but having an instruction decoder, a memory, and a processor is part of everything. The key piece of research that laid the foundation for the Internet was the work on ARPANET. There's a lot of work out there that's done, but the first systematic network that was built using the principles of sharing lines, of splitting them up, was indeed the ARPANET. And the people who worked on that then turned around and eventually worked on the Internet. And the ideas that they work are clearly the Internet builds on the ideas that they put together for the ARPANET. It connected a group of computer research sites that were at primarily major universities. It was funded by the Department of Defense. One of the things that that department wanted to do was to build computer science into a recognizable discipline. At the time, it was not. There were very few places you could study it. If a few you could, you were considered to be studying a form of electrical engineering. My early study of it, which was from that period a little bit later, I studied mathematics because I couldn't study it. I had a father who was in computing, which gave me access to lots of stuff to learn it, but I still couldn't study it. Their goal was to build a community. And in that goal, a whole lot of other things came out that were unspoken or sometimes very actively spoken about what communication should do. The first was that there was always going to be a human-to-human element. Email was not the first application. It was probably the third. And people figured out all sorts of ways to do it on the fly before they got a good system that worked. And in fact, that was a very important early standard, that they built a system that putting a file with a message in it, it would get to the right place without you having to intervene. That took a couple years, but it got there and it was working. The second part that they grasped and articulated was that it was not only person-to-person, but that there would be repositories, that there would be collections of information that anyone could go and get. They would be called libraries, they would be called servers, they weren't quite sure. They would be organized in various ways, but they would involve searching and looking for things. And that concept of searching is crucial and we have entirely forgotten that it is. It's one thing to have access to everything and everybody. It's another thing to find the person you want and get that little bit of information that you're hoping to use. Searching was an early computer problem. If you look at the early algorithms from the late 40s, you see lots of written on efficient ways to do multiplication and clever ways of doing long division. And every now and then a little thing will bubble up and say, "Oh, if you wanted to search through a bunch of records, this is how you would do it." There are papers, I think you can find one from '49, but you really don't start seeing that kind of question until the early 50s. And it very soon becomes a major issue of how you do it and how you do it well. If you're searching for names, you know that that name is a part of a record. That's easy. If you're trying to search on qualities, that's harder. A group of Stanford graduate students in the early 60s had this great idea, which we all have borne the burden of. Wouldn't it be great to put people's qualities into a database and have it search through them and match the couples that are most alike? It was an early form of computerized dating. They had a party, they ran the program, they distributed the outcome to everyone who was there, including couples who had been dating. I believe there were some there who were even married. Needless to say, the walk home that night was not as much fun as they hoped it had been. They were excited by the process. The thing energized them incredibly. But there are people that said, "What do you mean I don't match with this person? I match with that one. What do you mean that I'm closest to these qualities rather than those qualities?" I mentioned about the Hermann Hallruth machines, where people got very, very excited about the cards and saying they could see human beings in them, see histories in them, see their future in them. That comes up in the literature again and again, each time a new technology comes that allows us to see ourselves, our activities, our data in new ways. The PC was that very much in the 70s. People saw this as a personal device, something that I would use in a way that we have forgotten was novel. The idea that you had a computer that was solely yours, that you could use it and have complete control of it, was quite new in the 70s. People had been able to work one-on-one with computers before. That's not anything special. But it's special that it was yours, that you could make it. And you see it in customizing computers and putting stickers on them. We still go on with wallpaper and other things that make it ours, that make it the reflection of us, that show our ideas are important, that our work and that our activities are positive, and that they can be reflected back to us by our machine. But in working with systems, the fundamental rule is we adjust ourselves. We adjust our thoughts. We adjust the way we work. And that process involves us adapting our thought and our habits and the way we look at the world to the way those systems are designed. We talk about how we go shopping, about how we interact with various things. These are our actions to systems that are built on algorithms and other processes that have helped discipline us. And you see that in everything you do, the way we use our phones. We didn't carry phones with us 25 years ago. We went from phone to phone, from office to home. And in the process, we learned to pay attention to it, to use it in different ways, to get certain information and put information into it in different ways. You know, we now look at it for advice on how we get home. If we're commuting by car or by bicycle, where's the busy traffic? What do we do? What do we avoid? That was an algorithm that was once considered one of the great advances of artificial intelligence. There's an important strain of AI that is building large databases and searching through them. And the search that does it most effectively for that kind of work is called A* search, and that's what we use on our phones to find the fastest way home. Where's the traffic? What can we avoid? And in doing that, we look at it and we know how we give credence to that, whether we go on the roads marked red, because that's the straight way and we don't want to be bothered. Or whether we avoid it because we just loathe sitting in traffic and we want to keep the car moving. We adjust how we think and we adjust how we develop that strategy as we begin to see how effective it is. Does it just make sense to go through the big traffic? Are there better things to do? That's a simple example, but in everything we do, everything we deal with a computer, we're fundamentally asking three questions. Are we getting what we want out of it? And if we're getting what we want, we're likely to repeat it. If we're not, we're likely to modify what we do and try something different. Is it taking too much effort? And if it's taking too much effort, we're looking for a way to do it more cheaply, more quickly, more easily. And the last one is, is it irritating me? That's part of the cost, perhaps. But it seems to be a slightly different thing of how does it make me feel about what I'm doing? Will I change my tactics because the way it's presenting information or the advice that it's giving or the response that it's making, something that I'd rather not face and I will go on and do something else. Those are the key parts of our working with the machine. And our goals are always, do we get something that makes it better for us? Do we improve our position, where we are amongst our friends and family? Do we see that expand and able to engage more people, would engage with them more fully, or do things that we couldn't do before? That's the second part, our function, what we do. And again, we tend to do computers to either get something more done or do what we ordinarily do, but cheaper, with less effort on our part. And the big one that has really been the combination of AI and all the social networking has been status, which is how do people view us? How do they think about us? How do they listen to us? And some people may think about computers and say, "Does this computer honor my status? Does it give me more status?" But in fact, it's how we appear to our neighbors. How do we appear to our family? And so much of our response to computers at base is connected to how we respond to our family, to our neighbors, to our friends, to our community, to our office, to the people who drive next to us. And we change our response to it, in effect, so that we are doing things better, doing them cheaper, and our goals tend to be in that social grouping. Does it increase our status? Does it improve our function? What we do? Does it give us a new position in this world? I'm not particularly frightened or concerned about it, other than I know that it is part of adjusting our thought to the machine. And at some level, it is our adjustment that is saying, "Well, I will accept this," what the machine is saying, what the machine is doing, because I can view it as a reflection of myself, as a reflection of a being. Chapter 5 When Machines Replace Humans The connection between human computers and the machine computers is complex, because it's involving the division of labor. The modern computer divides three actions out, gives them each a device. Storing numbers. Well, storing numbers is something people are not particularly good at. We forget them very easily. There was a time that my hand would automatically type out my parents' phone number or my sisters. I have no idea what any of my friends' phone numbers are anymore, because we don't do them, and we forget them, and that's that. The second part is remembering instructions. We are better at that, but again, it involves a memory which we're not good at. But the interpretation, we tend to spin things that make life easy for us, or simpler for us, or more harmonious for us. The last piece is doing the task. And in working with the human computers as the process moved forward, increasingly, that's the piece that humans were reduced to, doing the task. There's a photograph that's in my first book that is, I'm convinced that it's going to be what I am known for in the rest of my days, that shows a room in New York with 450 people in it lined up in thin, little desks back and forth, all having a piece of paper in front of them, all of them having a pen, all of them doing addition, after addition, after addition. This room in New York City was the main office of a group known as the Math Tables Project. It was the largest collection of human computers, to my knowledge, that has ever been assembled on the face of the earth, and for its time, the most powerful computing organization in the world. It was a work progress administration project. In 1937, the depression, which at that point had been going on for not quite a decade, tightened up again. It was often called the Roosevelt Recession. And the federal government was looking for ways to put urban workers back to work, particularly New York City, which had an unemployment rate higher than most everywhere else in the country. And particularly, they were looking for jobs that were like construction jobs, but inside. The Institute for Advanced Study proposed that they build a computing lab, and that computing lab would be a group of people doing calculations for projects that could benefit by having highly accurate mathematical calculations done for them. They would take jobs from scientists, they would take jobs from government offices, they would do some work on their own doing, and this would be of value. But it's a work progress administration task. It means to say that they are recruiting from the unemployed, and by 1937, they were the long-term unemployed. Largely, they were store clerks, some of them office clerks. They had some knowledge of addition, but they were not particularly well-skilled, and the WPA was the last hope. These were people who could not feed their family any other way. And so, in taking these jobs, they did so reluctantly, they knew it was repetitive, but they knew it would feed their families. They were from the Bronx, they were from Harlem, they were from the poor parts of town. But at the same time, all of them had some at least high school education. For the bulk of them, they felt the sting two ways. The WPA gave them work, but also required them to look for work. And so, they're, "Hi, we're here, we're glad to have you there. You have to be going, you have to be looking for something else." And second, that problem of babbage, that people doing the same task the same way will always get the same results, will always make the same errors, too. Hung heavily on them, and they wanted to be known because of the reputation of the WPA as being a group that produced error-free calculations. So, they put together a process that ultimately, in terms of equivalent labor, meant that each number was calculated six or seven times, never the same way, never identically. And most of the effort was involved in creating a group of numbers, then taking them apart and trying to figure out how they worked. They stretched to find any place that their work was used and accepted, particularly as the war came. Whenever they figured out a connection between their work and a military operation, they would promote it heavily. Coming up to D-Day, roughly eight months before, in September of '43, they were given a series of calculations to do that were equivalent to putting down a map, throwing coins on the map, pennies, and counting where the coins lay. They did the experiment sometimes, and then they developed a mathematical calculation from it. One of the people looked at the map that they had put on the floor to throw it and recognized it as part of Normandy, France, and speculated that they were preparing something for the invasion of Europe, which indeed they were. They were trying to figure out how to clear the beaches of mines by dropping bombs. The idea was to drop enough bombs so they'd explode the mines without blowing too much of the sand away. They worked through the problem. It's tough computationally. It was well beyond some of the mathematical analysis at the time. They were immensely proud of it, even though in terms of looking at how the D-Day planners used that information, they ultimately concluded that rather than try to bomb the mines off the beach, they would try to avoid the beaches with mines. And how much they used it directly for planning is not clear, but they loved it, and they loved that connection. They did things for radar and radar calculations for the work against U-boats in the North Sea. The big one which they did not grasp, and it's part of their files, was calculations that involving the collapse of a sphere, of having an explosion crush a sphere into a smaller sphere. The request is in their files. It has a big statement. You cannot share this. We will not explain what it is. We will not explain what it's for. We will not tell you what the units are or what's going on. Crushing a sphere, that's what you do to get an atomic plutonium bomb to go off. And it was part of the work to check the calculations at Los Alamos for the Manhattan Project. When they found it out again, this was something of great pride to them. And yet at the same time, doing it made them feel like a part of a machine. They had trouble with labor unions, with strikes, with the typical things you have, with large labor activity, and the feeling that they were barely above a machine. When the place was demobilized, some of them were brought in and asked to take computing organizations for other groups. The bulk of them chose to be high school mathematics teachers, because they had learned enough mathematics that they could handle that and there they were dealing with kids and people. And that was a much more satisfying activity. It was work for the poor, and that really is the ultimate form of that human computing. Prior to that, it had always been surplus labor. People who had the skills and had some trouble finding work, but they were part of the process. When you get to that last group in New York City, they're laborers, and it's very clear they're laborers. The substitution of machinery for labor is a huge part of the story of computation. And there are two principal motivations for it. One, reducing activities that we thought required human intelligence to mechanization, to something that can be repeated again and again and again very quickly. That's only part of it. The second part is systemization. That you can put it in part of a larger system that is processing information that's doing things, that's making decisions and working on a bigger set of problems. In all of these, you're in effect trying to replace expensive labor with cheaper labor. Replace labor that is less predictable, whether by labor unions or just wanting to add their own ideas, or not understanding what's going on with something that will do the same thing again and again and again. That transition is always hard because there is a cohort that feels quite angry and quite misused at what's going on. My favorite example is in the 1950s, as the American economy is growing to, in many ways, provide the same level of production for consumer goods as we did for military goods in World War II. One of the new devices that's coming into manufacturing are automated machine tools, controlled machine tools. And the ones that we see in the '50s are things that are mimicking, that are following the motions of skilled machinists. After a part has been produced using a combination of manual and electric control, the tape is played back for automatically producing successive parts. The tape programs the machine's operations. The use of record playback control in this operation is expected to double production. You would take one of these automated tools and you'd have someone machine a part or create something. And the machine tool would record that and then duplicate it, replicate the process of doing it, so it would be behaving like the skilled machinist. And skilled machinist jobs were a higher level of factory work than, for example, assembly-like work in the auto industry. In assembly line work, you are part of the machine. Henry Ford made that very, very clear in his writings. The machine is the line and you are part of the machine and you don't, because you're a part, interfere with the operation of the whole. Machinists were different. And during the '50s, as they start seeing these tools come up, they protest. And it leads to one of the key fights that we are having today about who owns data. The argument that the large manufacturers had is we build the factories. We created the factories. You would not have your skills if you didn't work in our factories. You have no place to learn them, no place to use them. Therefore, your skills are something that we can copy and we can transfer to an automated machine tool. The workers obviously saw it from a very different point of view. We learned these skills. These are ours. We have added our own identity to those skills and you can't have them. You know, in the American labor movement, if the early '50s represents a high watermark of its control over activities, this is part of the first chink in the armor. This is part of the first attack on it. Who owns the skills of factory workers? That fight, in many ways, was resolved by the second or third, or maybe even fourth generation of machine tools, because by the '70s, they're not mimicking skilled machinists. They are putting together motions and activities and machining steps from computer-aided design files that are produced by computer. So, for the most part, mimicking skilled machinists is no longer an issue in the labor movement, not the way it certainly was then. But that process, that same controversy, continues with the collection of data for artificial intelligence. Artificial intelligence algorithms, as we view them right now, rely on data. They rely on trying to capture something of human behavior, human thoughts, human activities, and put those activities, forms and thoughts, in some form that it can store and recover, and furthermore, reason from or process from. It's not sticking with what it captured immediately. It's asking, how can we extrapolate out? How can we apply it to scenes we have not seen? How can we use it for situations that are novel? And the question becomes, who owns the activities that are being captured by data? If it were a factory, you'd have that same fight that we had before. Some large capitalized firm built the factory. You work in the factory, you could not do it without that, and so you're stuck. But it's no longer a factory. It's no longer a place we go to. It's the social media. It's where we buy stuff. It's the systems we do our work with. And often the things being captured are not necessarily the primal activity going on. It's some other action that we are doing with the system. And so that fight now probes much deeper into our lives. It's asking, who owns our basic activities? Who owns our right? Who owns the way that we use computers, how that we shop, that we communicate with our friends? Who can take that and use it in a way that replicates it? It's not clear that this fight has been resolved yet. We have seen some early efforts at legislation in California, was trying to address it and pulled back from it. But that still remains. What do we have that we own of ourselves, own of our actions, own of our thoughts? And that aspect of factory work and factory labor, that contention over who owns the skills of the factory worker was the starting point for that discussion and remains the starting point, even though our concept of factory and data gathering has grown much, much bigger. When we talk about artificial intelligence, we're lumping a bunch of different technologies into one. And we are looking from a very modern perspective at a process that's been going on for close to 75 years. The programs of artificial intelligence have been bubbling around for years, and you have seen them come and go, but it has taken a certain amount of time to get the computing power to make them work at scale, to be able to build them with enough sophistication that they work across the board with many different people. That that problem of providing a system that you don't need to be an expert at it to be able to use it has been part of it a long time and part of the expansion that we see of artificial intelligence is just that. We're building systems that allow more and more people to work with them. The technologist in me sits, there's kind of a U-shaped narrative. Something comes and it's very exciting and you're writing at the high end of the U, and then there's a dip, and then there's a recovery, and the recovery is usually less than the starting point, but there is still a recovery. New ideas are always exciting and I always enjoyed engaging them, but in the course of my career I've seen time and again new ideas that work in a brief demo in a controlled environment, and when you try to build them out to engage a large community of users, you get problems and you get problems that are seriously wrong. The early translation programs from one language to another. We're at times very clever and very laughable. I don't play this game anymore of translating something into another language and translating it back into English. The canonical one, which I was able to demonstrate during my graduate career, was taking the phrases, "Do you accept credit cards?" translating it into Polish, which I do not speak, and then translating that back to English and having the response be, "Do you play poker for money?" And, you know, it got some of those words right, and it was quite fun and humorous, but it illustrated the issue of making a system that's not the basis of reliable communication. By contrast, I now use in my work translate programs all the time because I'm dealing with people in China, with people in France, and for my group in Ireland, I even translate things into Irish just to be snarky about it. And I know that that's a regular system. I know to keep my sentences short. I know to keep direct verbs. I know do not put large amounts of dependent clauses and modifying words because the modifying words will end up modifying different things. But I know that, and it's a system I work with. It's not particularly exciting. It's not as exciting as those first ones were, but it's a tool I use and that I work with. My feeling is with artificial intelligence now, we have seen some of the first bloom of this generation. And this generation really goes back to the '90s. The machine learning, the large-scale neural networks, the use of these things in generative systems where you are getting a sense of the whole landscape of human behavior. All those are very interesting and very exciting in what they can do. They fail enough that they are not reliable tools for me. And I don't make much use of them because I don't want to trust what I am doing to potential mistakes that aren't my fault. If I'm going to pay for a mistake, it's going to be mine. And thank you very much. Okay. And so now you take all this and you edit the sucker. You're the editor. Want to support the channel? Join the Big Think members community where you get access to videos early, ad-free.

Up Next Why modern fitness culture misunderstands human bodies ▸ 19 min — with Daniel Lieberman