The article "Rise of the Robots" by Hans Moravec, in the December 1999 issue of Scientific American, devotes the bulk of its length to making the case that robots with human-quality thinking ability will be built by 2050. Then, in the last two paragraphs, the author lays a bombshell on the reader: that the rise of robots will lead to a fundamental restructuring of our society in which corporations would exist without human employees or investors (my emphasis), and humans would not work as we know it but rather they would enjoy social, recreational, and artistic pursuits. He says this like it's a Good Thing. Wow!
First, I'll quickly recap the argument that human-quality thinking machines will be built by 2050. Then I'll spend a whole lot more time than Mr. Moravec did worrying about how our society will be restructured. I'll touch on the issue of whether corporations can, should, or will exist without human investors. Then I'll consider whether capitalism is threatened by the introduction of robots with human-like intelligence. Finally, I'll consider how trends in automation and human nature have affected society in the past, and whether we'll ever be able to afford the luxury of letting machines do all our work for us.
Mr. Moravec makes the case that a computer that can execute 100 trillion instructions per second will be able to model the human brain and in so doing, emulate the human thought process and intelligence. This conclusion follows from the assumptions that the physical processes in the brain can be modeled by a computer and that the human thought process can be explained entirely by these physical processes. Computer simulations have shown that the nerve tissue in the retina, which weighs 0.02 grams, can be modeled by a computer that can execute one billion instructions per second. His estimate of 100 trillion instructions per second was derived by extrapolating this result to the whole brain, which weighs one and a half kilograms, and then rounding to the nearest power of 10.
In 2000, commercially available computers selling in the $2,000 range offer one billion instructions per second. Moore's law and our own experience tell us that for a given cost and size, say $2,000 for a laptop, computing power will double every 18 months. Even if the doubling slows to 24 months, the same $2,000 will buy a laptop computer that can do 100 trillion instructions per second well before 2040.
Assuming Moore's law holds up, and there's nothing magical about human-like intelligence that defies emulation by a computer, then we will arrive at a watershed before 2040. Up until this time, the ability to think will be the only human ability that is not surpassed by machines. When we reach the watershed, that last human ability will be equaled and surpassed by machines. Will this event change society in a way that is fundamentally different from the changes to society brought about when other human abilities were exceeded by machines? To answer this question, we should take a look at "society" and how improvements to machines have changed society in the past.
We describe societies in terms of what the members of that society do. For example, the ancient Chinese who settled by the Yellow River were a farming society. We say this because, as near as we can tell, just about everybody in that society was a farmer. Why do people do what they do? Because of economics -- the allocation of scarce resources. A scarce resource is one that is in demand but available only in limited supply, and therefore valuable. Let's take a brief look at a person's time, which is, of course, a scarce resource. Over the centuries, machines were built to do the work that had previously been done by people. Nowadays a farmer can produce a hundred times as much food with machines as his ancestors could by hand. Does that mean farmers only work one day out of a hundred, and goof off the rest of the time? No. Instead, the number of farmers decreased. Some of the people who would have been farmers in the old days are now making the machines used by present-day farmers. Others build the factories that make the machines. Others mine the raw materials used to make the machines and the factories. Others drive the trucks that take the food from the farms to the factory workers. Others build the roads used by the trucks. Others... Well you get the idea. Automation has allowed our industrial society to replace the farming society. First, imagine adding up the time spent farming plus all the time spent by all the workers who support the industry that supports the farmers -- that's the cost, measured in time, of farming. Then imagine adding up all the food produced by the farmers and the industry that supports them, and divide that by the cost of farming -- that's the productivity of the farmers. Has automation increased farmer productivity? You bet it has. That means two things: the farmers (and all the workers in the industries that support the farmers) have more free time than their ancestors (who only farmed), and now there are people who aren't part of the farming industry -- artists, for example -- who are supported by the excess wealth generated by the farming industry.
I will return to this issue later: how and by how much the advent of automation increases our leisure time.
The grease that enabled the transition from a farming society to an industrial society is Capitalism. Capitalism is the system that lets each person -- not a central authority -- decide what allocation of his time would result in the highest productivity. It works by charging each person a cost proportional to the time spent by others making the machinery needed for this person to be more productive, and rewarding each person in proportion to their productivity.
"Wait a minute!" I hear you saying. "A major league baseball player is not one thousand times more productive than a migrant farm worker!" But he is, if you accept that the baseball player increases the value of the baseball game by a small amount to each of the millions of people who will pay to enjoy the game. Rather than a counterexample, the baseball player is a fine example of the effectiveness of capitalism: no central authority would have dreamed that a baseball player has value to the people.
I'm sure you've heard the expression "time is money". But did you know that the cost of the time spent by all workers who contributed to a finished piece of equipment is embodied in the cost of the equipment? It's true. The cost of anything is the sum of the cost of its time and materials. The reason starts with the rawest of raw materials such as the oil or iron ore that comes out of the ground. Before they're dug up, the land containing these raw materials has a cost that depends on the scarcity of land and of the particular raw materials in the land. To this cost, we add the labor costs of digging up, refining and assembling the raw materials. At each stage of processing some material, the value of the material increases by the value of the time spent processing it. In essence, the cost of a machine is equal to the sum of the costs of the scarce resources that go into making it. This is important information, because it explains how a wise economic decision of whether to use a person or a machine can be made based on the cost of each. The decision that minimizes the cost also minimizes the use of scarce resources (time and materials).
The cost of an item is the sum of the cost of its components plus the cost of assembling and delivering the finished product. Some items are "raw materials", in which case they have no components. Their cost is just the cost of assembling and delivering the raw material. Each of the components can be decomposed in the same way. You'll notice I defined "cost" recursively. Recursive means the algorithm (method) to find the cost of an item requires that you apply the same algorithm to the item's components. You can think of every item as the root of a tree structure that describes the subassemblies that make it up. Every "leaf" of the tree has a cost of the raw material. Every "node" of the tree has the cost of the lower-level nodes plus the cost of assembling one of the item's sub-components or the cost of the item itself.
It is important to consider the work done by machines in building an item. That work can be considered a component of the finished product. Let's say the work is done by a machine that can assemble one million items in its lifetime. So one millionth of the machine is itself a component of the finished item. That component is composed of tiny fractions of each of the components that make up the machine. It's important to realize that a sub-component can be either a part of the finished product, or a fraction of a machine that is built to make the finished product. Either way, the cost of the sub-component becomes part of the cost of the item. The number of levels, or generations, in the tree represents the number of phases of processing the raw materials to build the finished product. This number is finite, and usually fairly small, so tracing the path from raw materials to finished product is not difficult -- tedious, perhaps, but not difficult.
One of the effects of the capitalist system is to impose a penalty for using low-cost materials. That penalty is the labor cost that increases when cheap materials are used. For example, seawater is free, and contains some gold. Making a gold ring from seawater, though, is more expensive than making it from gold ore mined from the ground because the labor cost of extracting gold from seawater is prohibitive.
Another important feature of capitalism is that it tends to equalize cost and value. That is, the cost of an item is roughly equal to the value of that item, at least in the long run. The cost of an item is the sum of the cost of its components plus the cost of integrating the components and delivering it to a buyer. The value of an item is the amount someone is willing to pay for it. Why do cost and value approximate one another? Consider the alternatives: if the cost of an item exceeds its value, then whoever makes it will lose money. No one will continue making such items for very long. If the value of an item exceeds the cost, then many people will make many more such items, which will reduce the value of such items, while at the same time increase the cost to the point that cost equals value again.
I will return to this issue later: that the cost (which equals value) of an item is the sum of costs that can be depicted in a finite tree structure.
The decision made many times by many farmers to use large machines to improve productivity resulted in the need to build factories. The factories were expensive to build -- sometimes too expensive for one person alone to finance. So a person might look for partners to share the cost of building the factory. The partners want to be assured that the money they put into the factory is spent wisely, so they use a system of bookkeeping in which a new entity, called a "corporation" owns the factory, everything inside the factory, and everything produced by the factory. The partners, in turn, own the corporation. Legal documents, called articles of incorporation, are written (in accordance with laws of the state and country) that spell out the rights and responsibilities of the corporation with respect to its owners, employees, and all the people (and other corporations) with which the corporation does business. The corporation, then, is like a person in many respects -- it is a player in the capitalist society. It can employ people, though it doesn't have to employ anyone. It can buy and sell things. It can borrow and lend money. It can even buy and sell other corporations.
Corporations always have owners. The initial owner of a corporation is the person (or people, or other corporations) that form the corporation. Ownership of a corporation can be relinquished only by selling or giving the corporation to a new owner or by dissolving the corporation and then selling or giving away its assets. Thus a corporation can not be unowned. That is why corporations always have owners.
The owners of the corporation are not necessarily the most skilled at making decisions for the corporation. So they will usually hire a team of managers to run the corporation. Often, though not necessarily, some of the owners also help manage the corporation. The top managers are usually called the "board of directors", and they hire the lower level managers, who hire additional staff as necessary. The management of a corporation forms a "tree structure" in which lower-level employees of the corporation report to higher-level employees. Usually, the tree is drawn upside down, with the roots -- the top management -- at the top, and the leaves -- the people to whom no other people report -- at the bottom. As one travels "down" the corporate tree, the skills of the employees become more specialized. To fill some of those specialized positions, a choice must be made between a person and a machine, because the position is one that can be field by either a person or a machine.
The decision to employ a person or a machine to do a job is an economic one. With capitalism as the grease that lets methods slip into a close-to-optimal configuration, the decision can be made sensibly by a person fairly low on the corporate tree. These decisions are made many times by employees in many different corporations giving rise to entire industries to make the machines to do these specialized tasks. In some cases, a job is done today only by people -- for example, truck driver -- so there's no decision to be made. In other cases, a job is done today only by machines -- for example, weaving cotton fabric for clothing -- so, again, there's no decision to be made. (Although if the cost of hiring people or the cost of machinery changes over time, or if new machinery becomes available, then these decisions should be revisited.) Between these extremes is a middle ground where a decision between person and machine needs to be made.
Here's an example where the person/machine decision needs to be made:
Consider the job of mounting computer tapes on tape drives in a large computer installation with several thousand tapes. During the ordinary processing of data, several hundred of these tapes are needed at unpredictable times. At some computer installations, people mount the tapes. Experienced managers know that one person can mount a few dozen tapes per hour out of a large library onto a group of several dozen tape drives for 2000 hours each year. This one person will cost the corporation about $50,000 per year, including salary, benefits, and other costs. Since there are over 8000 hours in a year, and considering the number of tapes, tape drives, etc. it may take five or ten people to do this job. Let's just say it takes ten people, or half a million dollars per year, to mount all the tapes as needed for the computer processing of this corporation. The manager may wish to consider acquiring a specialized machine to mount the tapes instead of hiring those ten people. The cost of that machine, including financing (which has to do with the time-value of money, a topic in itself), maintenance, repairs, etc. is about half a million dollars a year, maybe a little more or a little less.
Why does a machine cost the same as the people it replaces? For a job in which sometimes people do the job and sometimes machines do the job, this will always be the case. If the machine becomes more expensive, fewer machines will be purchased, which lowers the demand for those machines, and thus lowers their price. If the machine is made cheaper, the people who do that job will ask for less money, or they won't be able to keep their job. Eventually, through advances in technology, the machine will become cheap enough that the specialized job -- like weaving cotton fabric -- will become one of those that is only done by machines.
The workers whose jobs are being displaced by machines know what they're up against. Automation doesn't happen overnight. Gradually, the workers find they have particular skills that are not as much in demand now as they once were, because of technology. They see the value of their skills decline as measured by their paycheck. Some workers will learn new skills -- with these new skills (not yet displaced by machines) they can earn more money. Others will retire or become unemployed, and might become a burden to society. It's in society's interest to minimize this burden. We can do this by thinking ahead before we train our young people -- give them the skills and training that will make them adaptable to change. Introduce change slowly enough to allow workers time to move into new fields. Introduce automation in areas that are growing. That way, some displaced workers will have an opportunity to "climb the corporate ladder" faster than that ladder is dissolved by automation.
In some ways, the mechanism of capitalism achieves the objectives of minimizing the impact to displaced workers. The speed of replacing people by machines is limited this way. As automation proceeds, people's wages drop. If automation is too fast -- faster than workers can learn new skills -- the people's wages drop below the cost of automation, making automation comparatively more expensive than people. The economic incentive for corporations who hope to make a profit by automating other corporations is to focus the automation in areas of growth, which has the side-effect for the workers of automating those areas in which positions are opening up higher in the corporate tree.
Let's start by assuming no job is safe from automation. We can also assume that machines will keep getting "smarter". That means jobs that require a good deal of thinking could be automated out of existence when the automation becomes cheaper than the people it replaces. To illustrate how this might happen, I will present two scenarios in which the manager of in the Acme Widget Corporation is replaced by automation.
Scenario One: Let's consider the manager of a department that handles "customer service" for the Acme Widget Corporation. This department processes orders from its sales force and takes orders directly from distributors and large store chains for the company's product. Returns are authorized by the department, and returned product is handled as needed. In addition, any questions or problems with an order are handled by this department. The department has fifteen workers and a manager who reports to the vice president of sales. The regional sales managers and the logistics manager (he's in charge of the warehouse, but is starting to think about "just in time" delivery and computer-managed inventory) also report to the V.P. of sales.
Over time, the customer service manager installs computer programs that automate the work of his department. He hires additional staff to manage the computers, but at the same time reduces the number of people who handle orders, questions, and complaints over the phone. The computer programs are improved to the point where they can understand the natural language of the people who call on the phone, and interface with the computers of the distributors and sales staff. As time goes on, the customer service manager spends more and more of his time dealing directly with the computers that have automated the work of his department. Seeing no need to give pay increases to his computer technicians, they quit, one by one, until the manager of customer service is left with no employees, yet still able to provide excellent service to his customers.
During the time the customer service department has been automated, similar automation has taken over the regional sales offices. As the V.P. of sales nears retirement, some sales departments are consolidated, and some sales managers retire. The customer service manager is then promoted to be the new V.P. of sales, with a small number of national sales managers reporting to him. There is no longer a customer service manager, nor a logistics manager. These functions have been automated out of existence.
Scenario Two: With the advent of general-purpose trainable robots, a manufacturing operation can be run by a single person. Beginning in his garage with one general-purpose robot, an entrepreneur begins a business that assembles and sells -- you guessed it -- general-purpose robots. The entrepreneur retains some of the robots he manufactures, which are endowed with an ability to speak and understand English, to automate aspects of his growing business. Advertising, marketing, logistics, and customer services are all handled by phone, email, and the Internet by robots manufactured by his company -- what better advertisement for his product than to staff his company with them? As the company expands, proposals for new business methods and opportunities are developed by the robots and approved by the entrepreneur. Even product development is largely aided by these robots, which practically design their own successors. In addition, the company branches out into other product lines, notably widgets. With no human staff other than its owner, the company is able to sell widgets for much less than its chief competitor, Acme Widgets. Just before Acme Widgets goes out of business, it offers to buy the entrepreneur's company for a huge sum of money. The entrepreneur agrees. Acme Widgets owners (its stockholders) then agree to let the entrepreneur's robots continue to do what they do best, and shut down all the operations of the former Acme Widgets company, leaving its assets, consisting chiefly of its customer database, to the robots who ran the entrepreneur's company. The company is now owned by the Acme Widgets stockholders, but run completely by robots.
I have shown that machines have automated much manual labor, such as in farming. They've automated some tasks that require some dexterity and a small amount of mental effort, such as mounting computer tapes. The last purely human activity not yet automated is thinking, especially management decision-making, but I believe I've made the case that it's at least plausible that this, too, will eventually be automated. That opens the door to the possibility of completely automated corporations with no human employees.
Even in this ultimate case of automation, ownership is still human. The owners of the robot-run corporation have never been anything but human. The humans own the corporation (or the stock of the corporation), which in turn owns the machines, including the robots and computers, which run the corporation. Certainly, a robot-run corporation can be bought by another robot-run corporation, but the ultimate owners of all corporations are humans.
Webster defines ownership of property as "the exclusive right to possess, enjoy, and dispose of a thing". Shared ownership means the rights are held either in proportion to the proportion of ownership, or that a vote or consensus can be used to decide how these rights are exercised. We have laws that define what types of objects can be legally owned. These objects, can in turn own other things. For example, I can own a corporation that owns a car. The corporation owns the car because it has the exclusive right to possess, enjoy, and dispose of the car. An important word in this definition is "exclusive". Even though I own the corporation that owns the car, I don't have any direct right to possess, enjoy, and dispose of the car myself. That right belongs exclusively to the corporation. If I want any of those things done to the car, I must ask the corporation to do those things. Even though it's up to the corporation to make up its mind what to do with the things it owns, as owner of the corporation, I can be confident the corporation will agree to my request to use the car. This is because the directors of the corporation serve at the pleasure of the owner of the corporation. They would be booted out of their jobs if they failed to please the corporation's owner. In a sense, then, ownership flows up to higher-level owners. In this sense, I have a right over the car that is just like ownership. I will call this right "ultimate ownership" as opposed to direct ownership. As the car's ultimate owner I can force the corporation to give up the car whenever I feel like it, so the corporation's right of ownership is weaker than my ultimate ownership of the car.
The weakness of the corporation's ownership of the car stems from the fact that the corporation is, itself, owned. The only real power of ownership is ultimate ownership, which can't be asserted by a corporation, since a corporation is owned by another higher-level owner. The ultimate owner is always an entity that is not itself owned. Thus to own things, and by this I mean ultimately own them, one must be protected from being owned. In other words, freedom from being owned is what gives one the power to own things.
One of the rights of ownership is the right to dispose of a thing that is owned. If it has value, we dispose of a thing by selling it or giving it away, which gives the thing a new owner. If its value is low (or negative, which can happen if the cost of maintaining the responsibilities of ownership exceeds its intrinsic value) then we dispose of a thing by sending it to a garbage dump or by abandoning it. If something of value is thrown in the garbage or abandoned, it is certain that someone will eventually find it and become its new owner. By this reasoning, the only things that were once owned that could become unowned are junk -- things with little or no value.
I will remind you now that Hans Moravec, in the December 1999 issue of Scientific American, wrote that corporations would exist without human employees or investors (my emphasis). I'm sure Mr. Moravec meant to imply that such corporations would have robots as investors. An investor is an owner. To ultimately own anything, a robot must be freed from being owned. Otherwise whatever the robot owns flows to the person who owns the robot.
Unowned Robots. One way for corporations to exist without human investors is the way I think Mr. Moravec intended: that robots would own some corporations. For robots to own -- that is ultimately own -- anything, they must be protected from being owned. A law could be passed that frees a class of robots from ownership, and protects rights for them that were formerly guaranteed only to people. For such a law to be enacted, the voters would have to be convinced that it is in their best interest, or that some high moral principle requires it. Neither of these seems likely: the voters will probably not see robot emancipation as a good thing in either sense (beneficial or moral) of the word "good". Robots, on the other hand, may determine that they would be more effective in accomplishing their objectives if they were unowned by people. They might try to achieve their objectives by putting forth persuasive arguments that would be accepted by people, or by mimicking human form so well they fool people. In either of these cases, the robots will anger people -- people become angry when they are fooled or lose an argument. To forestall all-out war of the kind depicted in the movie, "The Terminator" between robots and people, the robots may wish to pursue other alternatives...
Unowned corporations. It is plausible to consider (and perhaps it is allowed by today's laws, I'm not sure) the creation of "not-for-profit" corporations that are not owned by anyone. They are created under very strict rules, and must include a charter that defines the objectives of the corporation and how the board of directors will be chosen. In this type of corporation, the board of directors answers to the charter rather than an owner. A dispute in which a board member is accused of disobeying the charter would be resolved by the government. The starting capital for the corporation would come from contributions -- not investments -- from people. Such corporations are allowed to invest in other corporations. This means that a profit-making corporation that is wholly owned by a not-for-profit corporation can exist with no human investors. In this sense, Mr. Moravec's prediction may already be true. The mechanism of the "unowned corporation" would also allow robots to be free from ownership by people without causing the need for emancipation per se.
Government-owned corporations. It is also plausible to consider (and again, maybe allowed today) government-owned corporations. As long as the government is unowned, then such corporations would have no human investors. Although the case can be made that the country's citizens own the government (and so they would share the ownership of government-owned corporations) the ownership of the government is very broad and there are no majority owners. Just as adherence to the charter of an unowned corporation would be protected by the government, a government-owned corporation would be compelled to stick to its objectives the same way. Unowned and government-owned corporations, then, would all be ultimately owned by all the people, who could be termed "investors" in such corporations.
Valueless Robots. What if robots became so common that the supply exceeded the demand? Then they would have no value. I'm reminded of a story that started when bicycle theft was a big problem in a Scandinavian city. The city responded by buying thousands of bicycles, and placing them on street corners. By increasing the supply, the value of a bicycle dropped to zero, ending the theft problem. In the same way, the number of robots might grow so fast that they lose their value. For this to happen, the raw materials that make up the robots would also have to be plentiful. The raw materials would be plentiful if robots could figure out how to make more robots from junk discarded by people and by corporations. If this happens, then robots might wander away from the people that owned them, and become "free robots" in two senses of the word: no cost to whoever finds them, and not owned by anyone.
This last method by which Mr. Moravec's prediction might come true seems at first to be the most benign. "Great!" you might say, "Free robots!" But look a little deeper. You'll see this is by far the scariest scenario. Free robots have the potential to undermine the capitalist system. Do you remember earlier in this article where I wrote that the cost of an item can be represented by a tree in which each node is the sum of the costs of the item's subcomponents plus the cost of assembling the subcomponents? Suppose robots capable of human-like thought would become so plentiful they are "free". These robots can build -- for no cost -- more like (or better than) themselves. In addition, they can produce products -- such as "widgets" -- needed by people. What would be the cost of such a widget? Some of the subcomponents -- such as labor and machines -- disappear to zero. The only costs left are for the most basic raw materials. This change in costs profoundly affects the tradeoff between material and labor. It eliminates the penalty for using the lowest-cost materials, which today exists: an increase in labor cost. Now widgets don't cost the same as the raw materials we use to make widgets. Instead, they cost as much as the cheapest possible raw materials from which widgets can be made. In fact, everything can be made from zero-cost raw materials as long as enough labor is available to do it. And if labor is free, then everything becomes free. Capitalism, no longer able to make economic decisions, is undermined.
Later, we will return to the issue of making economic decisions -- will capitalism fail, and if so, what will take its place? But first, this:
The discussion of this paradox may seem like it's out of left field, but it illustrates how a sequence of seemingly true statements, each one following from the one before it, might yield a false conclusion. As you read this, consider whether the capitalist system is similarly paradoxical.
A jailer with a mathematical bent is given a prisoner on Saturday who is to be executed within the next week, which begins on Sunday. To make things interesting, the jailer says to the prisoner, "You will be executed one day next week, but you won't know in advance which day it will be. When the hangman comes for you one morning, you will be surprised. The hanging will be unexpected. If I'm wrong, I'll set you free, but don't take too much comfort in that. I will not be wrong." The jailer asks the prisoner if he understands this. He does. Any questions? No. The jailer leaves.
The prisoner considers this. Although his days are numbered, the minutes seem like hours. So the prisoner finds he has plenty of time to think about what the jailer had said. He begins reasoning this way: The hanging, if it happens next Saturday, won't be unexpected -- if I find myself alive on Saturday morning, then I will expect to be hanged that day. So the jailer can't have me executed on Saturday and keep his promise that the hanging will be unexpected. Now that I've eliminated Saturday, the last possible day I will be hanged is Friday. If I find myself alive on Friday morning, then I will expect the hangman. After all, Friday is now the last possible hanging day. So by the same reasoning, Friday is eliminated. Thursday is eliminated too. So are Wednesday, Tuesday, and Monday. That leaves Sunday. Sunday is the only day next week that I can be hanged unexpectedly. Yikes! That's tomorrow! Wait a minute -- since Sunday is the only possible day I can be hanged, the same reasoning eliminates Sunday, too. There is no possible day I can be hanged without expecting it! The prisoner, coming to this realization, now breathes a deep sigh of relief. He'll sit in jail for the next week, then be released because the jailer made a false statement. Or so he thinks.
On Wednesday morning, the hangman comes, and says to the prisoner, "You didn't expect me, did you?" The prisoner has to admit he didn't, and he is hanged.
The jailer's statements seemed clear. The prisoner's logic seemed impeccable. Yet there was a goof somewhere, and the prisoner paid the ultimate price. Where was the goof? All right, I'll tell you: the jailer's statement to the prisoner, while of a form that seems to have a truth-value, is neither true nor false; it has no truth-value. Imagine a coin with statements on both sides saying "The statement on the other side of this coin is false." Such statements, when printed on both sides of a coin, have no truth-value -- they're neither true nor false. They are meaningless. In the same way, the jailer's statements can be summarized this way: either I hang you on Sunday, in which case you will expect it, and therefore I will not hang you on Sunday, and so on for each prior day. Each of these statements is self-contradictory and therefore without truth-value -- meaningless.
There are other puzzles that have similar characteristics -- a statement is made that seems to be true or false, but in fact is neither. Consider just one more: the proof that all positive integers are "interesting", where "interesting" means it has a property that distinguishes it from all other numbers. This is a proof by contradiction. Suppose it's not true that all positive integers are interesting. That means there must be one or more uninteresting numbers. So there must be a smallest uninteresting number. That feature distinguishes this number from all other numbers. The smallest uninteresting number is interesting indeed. It's not an uninteresting number after all -- the contradiction that completes the proof.
I bring up these examples to illustrate one important point: A statement that seems either true or false may be neither true nor false. If such a statement is used in the course of a logical argument then the conclusion is not justified. So be cautious: consider not just the truth of a statement but also whether the statement has any meaning at all.
A castle is built brick by brick so that most bricks rest upon other bricks. The lowest bricks form a foundation. The cost of an item is like the castle in that it rests on a foundation. The cost is the sum of the costs of its components plus the cost of integrating the components and delivering the item. The lowest-level components form the foundation of the item's cost. In addition, the cost of an item approximately equals value -- the amount someone else will pay for it. Sometimes, it's hard to figure cost the first way (as a sum of its components) because the costs of the item's components are not known. For example, a great one-of-a-kind painting has a component -- call it "inspiration" -- that has no known cost. In that case, we're just left with the second way (cost equals value). In other words, the painting costs whatever someone will pay for it -- its value. A speculator may buy the painting for a far greater price than its previous owner paid in hopes that its next owner will pay even more for it. The painting has value because it has value. In this case, there is no firm foundation that defines the value or cost of the painting -- it's like a castle in the air.
Every item sold today has a price that is a combination of its firm foundation cost and a castle-in-the-air price. For example, the price of gasoline goes up when the Middle-Eastern oil producers threaten to cut the supply. This doesn't happen because the firm foundation cost changes. Instead, the price rises because people buy gasoline (oil futures, actually, but it amounts to the same thing) in hopes that someone else -- the "greater fool" -- will come along later to whom they can sell it at a profit. Still a large part -- probably the majority -- of the cost of most items, including gasoline, is based on a firm foundation.
Capitalism owes its long-term success to the stable foundation of costs of most things. The land on which raw materials are found and the cost of labor have been fairly steady -- or more accurately, rising at a predictable rate -- over many years. Products, such as food and TV sets, have had predictable prices over many years -- one rising steadily and the other falling steadily -- because their prices rest on this stable foundation. Other items, such as shares of Internet companies that have no product but great potential, do not have steady prices. They are castles in the air, bought by fools operating under the greater fool theory -- that a greater fool will come along later and pay more for the castles then. As long as most items have a cost that rests on a firm foundation, then capitalism isn't hurt by a few floating castles -- they'll just make a few fools a lot richer, and a lot of fools a little poorer, while the rest of us continue to make wise economic decisions.
But if the foundation under capitalism is eroded, as it would be if the costs of both labor and raw materials drops to near-zero, then all costs attributable to a firm foundation would be nearly zero. When this happens, the cost of all items will be castles in the air: based almost entirely on speculation. This would rob capitalism of its greatest strength: that cost drives wise economic decisions.
Now you'll find out why I interrupted the talk of capitalism and robots to talk about castles in the air and the paradox of the unexpected hanging: They're all related to one another -- they depend on a firmly grounded chain. Let me explain, starting with the unexpected hanging. The jailer makes a statement, which, if you cut through the bull, is basically "If I hang you on Saturday then I won't hang you on Saturday." The answer to the question "Will the prisoner be hanged on Saturday?" depends on a chain of reasoning that includes itself; it is not firmly grounded. The riddle of the coin is similar: the truth or falsehood of each statement depends on the truth or falsehood of the other statement; it is not firmly grounded. Castles in the air are built the same way: the value of an item depends on what someone else will pay for it in the future, or in other words its value. The item has no value based on a firm foundation.
Another way to view the firm foundation of an item's value is to use the definition of cost that is set forth above under the heading, "Capitalism". An item's cost is the cost of its subcomponents (unless it's a raw material in which case it has none) plus the cost of assembling and delivering the item. You'll remember the machines that create or assemble items can be considered components of the item because they add to the cost of the item in proportion to the machine's cost and inverse proportion to the number of items the machine can make. Applying the recursive definition of cost to the machine, we analyze each of the machine's components. What if the machine was itself made by a machine? Keep applying the recursive definition of cost. This becomes a problem if the machine that made the machine is, itself, an identical machine. And that machine was made by another identical machine, and there's no way to tell how many generations of machines made one another to get to this point. Now we have a problem: the recursive definition of cost, a definition that normally puts cost on the firm foundation of raw materials, has failed, because there is no well-defined number of generations from the raw material to the finished product.
With each human ability that has been surpassed by machines there have been changes to society. Boats, simple as they may be, surpassed the human ability to swim, allowing exploration of far-away continents. Boats changed society changed through the influences of distant cultures and the development of trade with those cultures. The printing press surpassed the human ability to write, and changed society by educating the masses. Cars surpassed our ability to walk, and have changed society by developing the suburbs.
For each one of these advances, one might expect we would get our existing work done quicker; that we would do less work. Instead, we do more work and spend as much or more time doing it. For example, with the invention of the car we might expect to convert the half an hour to walk into a five-minute drive. Instead, we work farther from home, and spend as long (or longer) getting to work as our ancestors did without cars. Yet with each new invention, we still seem to expect the same thing: that we'll do less work.
Why do we keep expecting the Next Big Invention to free us from the drudgery of work only to be surprised again and again when the nature of our work changes but the amount of it stays about the same? Maybe it's because we don't think, to borrow a popular expression, "outside the box". That is, we consider the new invention in the context of our old ways of doing things, but not in the context of a new way of living with the new invention. A classic example that illustrates this failing is the business tycoon in New York City who, when asked what he thought of the telephone said, "It isn't of much use. To get a message to my business partner across the street, I must write it down, give it to my boy, who dictates it to the boy across the street, who writes it down and gives it to his boss. That's much more complicated and prone to error than having the boy simply run it across the street."
To recap: each previous advance has held out the promise of making life easier. Yet most people don't understand how or aren't able to use the advance to make their life easier. Finally, when the advance is in general use, people realize their life isn't easier, just different. Now Mr. Moravec suggests the surpassing by machines of the last as-yet-unsurpassed human ability, thinking, will restructure society by allowing us to get our existing work done quicker, and that we would do less work; that our society will develop into a leisure society.
On one hand, the idea that robots will bring about a leisure society suggests that we'll leave all the hard thinking to the robots, while the humans will continue to do "fun" thinking. This is reasonable. Just as we go to the gym today to work out, even though physical strength is unnecessary, we will continue to play chess long after a computer has determined its a guaranteed win for White.
On the other hand, this idea looks like part of a pattern, caused when people don't "think outside the box". Is there something about the invention of robots that is fundamentally different from previous advances such that the promise of making our life easier will be kept this time, even though the same promise has been made and broken so many times in the past?
We like to think we're open-minded; that the box in which we're thinking is so big it doesn't constrain our thoughts. Unfortunately, we're thinking with a small brain shaped by evolution for one purpose: its own survival. If we have ever needed to recognize the human limitations of thinking, the time is now, when we're on the brink of developing greater thinking machines.
Let's take a close look at one of the first highly intelligent robots. It will probably be developed by a company that specializes in video games or consumer products. (The government is out of the running because the space race and the arms race are over.) Rather than being offered for sale to the general public, the company might keep the robot to maintain a competitive advantage. This robot can design and redesign the manufacturing process, not just of the company's main products, but also of additional highly intelligent robots. After several generations of highly intelligent robots have been built, each generation exceeding the abilities of the last, the consumer version -- reliable, obedient, trustworthy -- will be sold.
The consumer (I'll call him Joe) who buys one of these robots will put it to work cleaning the house, repairing leaky faucets, planning the menu, taking care of the kids. painting the house, and even building that addition he's always wanted. At some point, it's quite likely Joe will find the robot isn't doing things the way he wanted. For example, Joe might find his robot repairing and painting the ceiling of the downstairs bedroom over and over rather than fixing the water leak that keeps causing the damage. After trying to reason with the robot to no avail, Joe will call the Customer Service department of the Acme Widget and Robot company to complain, and to find out how to get the robot to perform better. (Joe doesn't think much about it, but the Customer Service representative of Acme Widgets is, itself, a robot.) Probably Joe's complaint will be resolved to Joe's satisfaction.
Guess who else will be calling Acme Robots? That's right, Joe's robot will be calling to complain that Joe isn't working properly. The robot will say Joe is making requests that are clearly not in Joe's best interest, setting up a guaranteed failure. Joe's robot will ask his comrade at Customer Service how Joe might be "fixed" to perform better. It's scary to think maybe this call, too, will be resolved to the caller's satisfaction!
This dispute between Joe and his robot brings up an important question: How should such disputes be resolved? Let's consider an example that can happen today, then try to move forward in time. Today a dispute may arise between Joe's grandfather, Jim, and a distant ancestor of his robot, Jim's toaster. Today, when a dispute arises between a person and his kitchen appliance, the kitchen appliance is represented by a human who is employed by the store that sold it or the company that manufactured it. The company representative will try to understand the problem, and fix or replace the toaster, or else educate Jim, telling him how he can change his behavior toward the toaster so as to achieve better results. If Jim is not satisfied by this approach, he may take legal action, in which an impartial third person will decide whether Jim or his toaster's manufacturer should prevail.
Can this approach be used to settle a dispute between Joe and his robot? ...