Voters Still Matter (mostly)

November 17, 2022

Previous | Next

This is admittedly an odd title for a post about a democratic country, especially one that has traditionally regarded itself as the leader of the free world. The idea that voters matter is one of those things we should not have to say, like Black Lives Matter or human rights matter. But since the election of Donald Trump in 2016, many serious and well-informed people have warned us that American democracy is not as secure as we might think.

Four years ago, Steven Levitsky and Daniel Ziblatt published a book called How Democracies Die. They pointed out that in recent decades, most breakdowns of democracy have occurred not through military coups, but through democratic election of leaders who used the power of their office to promote authoritarian rule. All societies produce an extremist demagogue from time to time. In the most democratic countries, they usually don’t get elected. The threat to democracy arises when “fear, opportunism, or miscalculation leads established parties to bring extremists into the mainstream….” First those established parties fail to stop them from being elected; then they fail to stop them from violating democratic norms.

Levitsky and Ziblatt warned that Donald Trump looked like such an authoritarian leader, and they did so two years before the January 6, 2021 assault on the Capitol and the serious attempt to overturn the results of the 2020 election. The House committee investigating those events has made a strong case that the defeated president and his associates resorted not only to legal court challenges—which they lost for lack of evidence—but also to illegal means like pressuring state election officials to “find” more votes, arranging for fake electors to come forward, persuading the Vice President to exceed his legal authority by refusing to accept certified state results, and sending a mob to the Capitol to impede the certification process.

These events raised legitimate concerns that having failed in 2020, election deniers could win control of key positions in battleground states, putting them in a position to interfere with the 2024 election. They could use unfounded rumors of election fraud as an excuse to suppress votes in heavily Democratic areas or reject the results if the Democrat won the state. Once back in power at the national level, who knows what other measures MAGA Republicans could take to weaken democratic institutions and consolidate their power. I have not been as pessimistic about American democracy as some, but I certainly shared these concerns.

Democracy held

I am relieved to report that the voters blocked the worst outcomes in most cases. In races for governor, prominent election deniers went down to defeat in Pennsylvania (Doug Mastriano), Wisconsin (Tim Michels) and Arizona (Kari Lake). In races for secretary of state, they lost in Michigan (Kristina Karamo), Nevada (Jim Marchant) and Arizona (Mark Finchem). Although election deniers did win some races, they were generally in solidly red states where Republican candidates could win fair elections anyway.

The 2022 midterms were the third big loss for Trump and MAGA Republicans. They failed to win the presidency or either house of Congress in 2020; they failed to overturn the results of that election; and they failed to set the stage for the subversion of future elections, at least for now.

Another reason to be hopeful is that the voting and ballot counting went much more smoothly than many of us expected. Many of the losers conceded, and few besides Trump called the process corrupt. No angry mobs appeared at polling or counting places.

Many Republicans are now blaming Trump and his minions for costing them the big midterm victory they expected. We hear a lot of talk of “moving beyond Trump.” The response to his announcement of another campaign for the presidency has been less than enthusiastic, and his prospects of winning another term seem to be getting dimmer all the time.

Dangers ahead

Threats to democracy remain, however, especially at the state level, and those threats also affect the strength of democracy at the federal level. Gerrymandering remains alive and well in many states. The narrow Republican victory in the House of Representatives is probably due less to persuading more voters to vote Republican, and more to redrawing voting districts to squeeze more seats from the voters the party already had.

Consider my own state of North Carolina, which is one of the most gerrymandered states. The state has a nearly evenly divided electorate and a divided government: a Democratic governor but a Republican-controlled legislature. Donald Trump narrowly won the state in 2020 with 49.9% the vote. Republicans have been working for years to turn their legislative advantage into permanent control of the state. They have succeeded in redrawing election districts so that they can get about two-thirds of the seats even if they win only half of the overall vote. This remains true even though the state Supreme Court rejected their latest—and probably most extreme—redistricting based on the 2020 census.

In the midterms, NC Republicans hoped to win a supermajority that would enable them to override any vetoes coming from the more democratically elected governor. They succeeded in the state Senate, but fell just short in the House. But they did win a Republican majority on the state Supreme Court, which makes that body less likely to interfere with gerrymandering in the future. If that weren’t enough, NC Republicans have also brought a case before the Supreme Court of the United States arguing that state courts do not even have the authority to review legislative decisions relating to voting. This “independent state legislature theory” has not convinced the high court in the past, but appears to have gained support since Trump appointees have joined the court. (Oral arguments for Moore v. Harper are scheduled for December.) Despite being an evenly divided state, North Carolina is perilously close to establishing one-party legislative rule, without the normal executive or judicial checks and balances we associate with democracy. Arizona and Wisconsin are other states teetering on the edge of authoritarian rule. Florida may have already gone over the edge.

Even if election deniers have lost in most statewide races, they have often won in House districts, especially those gerrymandered to favor Republicans. The Washington Post counted 150 election deniers among the roughly 220 elected House Republicans. On January 6, 139 Republicans voted against certification of Biden’s election. One can easily imagine an even larger number doing so next time.

As welcome as it would be, the departure of Donald Trump from the scene would not necessarily turn the Republican Party into a defender of democracy. Even if the party does not need him as its leading voice, it still needs his supporters, many of whom were well on their way to being radicalized by right-wing media long before Trump rode down the escalator in 2015. MAGA supporters will constitute the largest and most aggressive bloc within the Republican-controlled House. They can exert great pressure on their leaders and more moderate members to cater to their anti-establishment impulses, perhaps by impeaching President Biden or his cabinet members, conducting endless investigations and hearings based on unfounded conspiracy theories, shutting down the government by refusing to pass a budget, or forcing the government to default on its obligations by refusing to raise the debt ceiling. What a majority of voters would actually like Congress to do might count for little in this scenario.

Voters have narrowly warded off the most immediate threat to democracy, but much remains to be done. American democracy requires two parties (at least) willing to engage each other with honest debate on issues, compromise when necessary to achieve a result the majority can live with, and accept defeat gracefully. The temptation will always exist to resort to undemocratic means to have one’s way, but responsible leaders will resist it. The voters must have the information they need to identify politicians who do not respect the rules, as well as the voting power to defeat them.


The AI Economy (part 4)

November 4, 2022

Previous | Next

Here I discuss the Epilogue of Roger Bootle’s The AI Economy, in which he raises deeper questions about the resemblance between artificial intelligence and human intelligence. Many technology enthusiasts imagine a point where AI becomes so advanced that it equals or surpasses human intelligence. Victories of computers over humans in challenging mental tasks like playing chess are only the beginning of what they think is coming.

The popular term associated with this turning point is the “Singularity”:

The Singularity is now generally taken to mean the point at which AI acquires “general intelligence” equal to a human being’s. The Singularity is so important, not only because beyond this point machines will be able to outperform humans at every task, but also because AI will be able to develop itself without human intervention and this AI can therefore spin ever upward, out of our understanding—or control.

In this scenario, humans would not only be replaceable in most forms of work, but could be enslaved or even annihilated by the intelligent machines, as some science fiction movies imagine. Opinions on the Singularity range from the belief that it is inevitable and imminent, to the belief that it is “eons away” (Noam Chomsky) and perhaps impossible in principle.

Enter philosophy

Whether the Singularity is possible depends on whether a mechanical artifact can ever be considered the equivalent of a human being, or if the distinction between the two is ultimately insurmountable. Bootle insists that this is a philosophical question beyond the scope of computer science. The discussion brings up the old mind/body problem, the question of whether what we call “mind” is ultimately reducible to a mechanistic and material system that we can emulate in a dead machine.

[I]f a group of technicians believes that they can venture into this most difficult territory of the relationship between mind and matter without reference to the contemplations of philosophers over the last 2,000 years and without even considering the religious view, I wonder if they can really appreciate the depth of the issues.

Bootle also seems to realize that one cannot dismiss the problem simply by rejecting traditional religious supernaturalism or early modern mind/body dualism. Even if mind is not a distinct immaterial thing, it may have qualities quite different from a mechanical brain. A key question is whether human intelligence is reducible to computation. Although he does not get into detailed philosophical or scientific arguments, he shares his own intuition that it is not. He attaches great significance to the fact that human intelligence is embodied in a living body that is capable of emotion.

[E]motion is a key part of how humans make decisions and also a key part of their creativity. This is a completely different realm from computation…

Perhaps our ability to engage with the physical world, to encounter and understand it, and to be, is rooted in the fact that we are incarnated. If that were true, then it would prove impossible to create, artificially, out of nonbiological matter, what we would recognize as intelligence.

In other words, a mind such as ours is more than a piece of software that could run just as well on a different platform, a dead machine instead of a living body.

Although the computational theory of mind and the mechanistic conception of reality on which it is based have been dominant ideas of the Machine Age, some scientists are open to alternative perspectives. Bootle cites the work of physicist Roger Penrose, who founded the Penrose Institute to study the distinctly creative qualities of human intelligence.

Creative experience

Although I am not a philosopher, my own studies in philosophy and science have led me to similar conclusions. One important source for me has been Charles Hartshorne’s “philosophy of shared creative experience.” In an essay with that title, he begins by saying, “In every moment each of us accomplishes a remarkable creative act. What do we create? Our own experience at that moment” (Creative Synthesis and Philosophic Method, p.2).

Consider what your mind is doing as you read these words. Can you process them just by applying an abstract algorithm, a computational procedure? Don’t you have to interpret each word in the light of a multitude of past experiences, some as recent as the preceding words, but others as remote as much earlier life experience. (Doesn’t it matter whether you’ve ever heard of Hartshorne, or have some idea of what philosophy is?) Many people might read these words, but you are having a unique experience by constructing a cumulative synthesis of your prior experiences up to this moment. Hartshorne says, “By no logic can many entities, through law, exhaustively define a single new entity which is to result from them all.” Also, you can be conscious of what you are doing, especially after philosophers have encouraged you to reflect on it!

This line of reasoning draws the distinction between living thinkers and machines quite sharply, since even our smartest machines experience nothing! Never mind, say the AI enthusiasts. Programmers just have to find the right logic, the right algorithm, the right wiring, and a network of dead circuits will acquire consciousness and the capacity for experience. A human brain is nothing but a network of connections anyway, isn’t it? Or does the fact that it is composed of one of the marvels of the universe, the living cell, make a difference? (I am also open to the possibility that smaller, naturally occurring parts of nature can experience something, the philosophical position known as panexperientialism. Bootle touches on that by suggesting that “mind, in some form, is at the the root of the universe.” Some distinguished philosophers and scientists, such as William James and Alfred North Whitehead, have taken that position, and some interpretations of quantum physics support it.)

Machines are externally programmed, but without feelings and life experience they are not truly self-motivated. They can learn, but their learning is limited by what their programmers want them to learn. A robot can simulate being happy to see you, but it is not actually happy to see you. A robot can display caregiving behavior under the right conditions, but it cannot really care about a child or love the child unconditionally. Machines can assist us in fulfilling our human aspirations, but they can hardly lead us, since they have no aspirations of their own. Without true self-motivation, they cannot be truly autonomous from their designers and programmers. Those who make the machines must not abdicate responsibility for them or avoid accountability to society for what they get them to do.

Time will tell if Bootle’s skepticism about the Singularity is justified. In the meantime, Bootle’s reasonable assessment of the possible economic impacts of artificial intelligence is probably more useful than more speculative dreams of technological utopia or dystopia.


The AI Economy (part 3)

November 2, 2022

Previous | Next

Part 3 of Roger Bootle’s The AI Economy is called “What Is to Be Done?” Here he addresses some of the policy questions posed by the advances in artificial intelligence and robotics.

Regulation

Since Bootle does not accept the most pessimistic or apocalyptic visions of the technological future, he does not see any good reason for government to discourage or impede the adoption of these new technologies. He does not support a “robot tax,” based on the assumption that robots will deprive humans of the opportunity to work or deprive government of needed revenue from employment and income taxes. He expects new forms of work to become available, and incomes and tax revenues to rise along with enhanced productivity.

However, he does think that new regulations will be needed to protect the public from some of the misuses of technology. For the most part, these are problems that are already in the spotlight, such as cybercrime, massive collection and storage of personal data, and the increasing surveillance of populations by authoritarian governments.

A very tricky question is who will be legally responsible for the decisions made by AI systems. The algorithms they use are so complex that determining how a particular decision was reached can be extremely difficult. But someone must take responsibility for decisions that impact people’s lives. “The algorithm made us do it” cannot be a sufficient excuse. Social media companies are grappling with the question of whether their algorithms contribute to political radicalization by feeding their customers more and more of the political messages they “like,” no matter how inaccurate, hateful or defamatory. As I write this, Elon Musk has just bought Twitter and vowed to weaken its rules about acceptable content. (The self-appointed “Chief Twit” has also set a despicable example by tweeting unfounded conspiracy theories about the assault on Nancy Pelosi’s husband.)

Education

Bootle imagines a transformation of education, but what he expects is not what the most ardent admirers of artificial intelligence often envision:

AI enthusiasts gush about the way robots and AI are going to transform a particular aspect of society—in this case, education—but they completely misjudge the nature of that transformation. Supposedly, teachers are bound to go the way of taxi and lorry drivers, set to be made redundant and consigned to become yet another lump of unemployed skilled workers having to scrabble around for some new way of earning a living. Meanwhile, the subjects that need to be taught are revolutionized, with traditional arts subjects thrown out of the window, to be replaced by STEM subjects only.

Bootle fears an overemphasis on science, technology, engineering and math, to the exclusion of a broader education. This is consistent with his view of humans as creative thinkers and decision makers, assisted by technology but not replaced by it. He quotes David Kosbie of Carnegie Mellon, who says that “with AI taking over routine information and manual tasks in the workplace, we need additional emphasis on qualities that differentiate human workers from AI—creativity, adaptability and interpersonal skills.” Bootle’s view is also consistent with his expectation that workers will take some of the benefits of higher productivity in the form of more leisure. People need education for life, not just for technical proficiency. He agrees with those who would add an A for Arts to make STEM into STEAM.

This does not mean that the organization and methods of formal education will survive intact. The traditional lecture method may well be replaced by more efficient methods of imparting information, but that doesn’t mean that most teachers will have nothing left to do. As with other creative forms of work, teaching can be technologically assisted without being technologically replaced. If teachers are to work more creatively with students, we may actually need more teachers, but with smaller classes.

Bootle describes the higher educational system as an “expensive disgrace.” Too many students pay too much to receive a degree that gives them an advantage over other job applicants without necessarily enabling them to do a better job. Here he draws on the work of other critics of “credentialism,” such as economist Bryan Caplan (The Case against Education). Bootle admires Switzerland, which has one of Europe’s leading universities in Zurich, but also more vocationally relevant alternatives for young people who don’t go to university at all. He believes that “the bulk of educational time can be spent at home or in the work place, interspersed with occasional visits to the physical ‘seat of learning.'” Higher learning need not be packed into four years of frenetic activity, but should be spread over a lifetime.

One issue on which Bootle and Caplan disagree is public education. Caplan would separate school and state, leaving education to the private sphere. Bootle uses the economic concept of “externalities” to argue for a large public role. The costs and benefits of an economic exchange are not limited to the individuals who engage in it. When a company pollutes the environment in order to produce a product, the costs are borne by society. Similarly, when an educator well serves a student, the benefits are shared with society. A democratic society has good reason to incentivize activities with positive externalities and disincentivize those with negative externalities. If left to their own devices, private firms have a limited incentive to invest in their workers’ lifetime learning or retraining, since the firm doesn’t own the workers for life and recover the full cost of its investment. The need for more education in the technological era will probably increase the government’s role in financing education.

Income distribution

Bootle expects artificial intelligence and robotics to cause some disruptions to employment, but he does not think they will be severe enough to require drastic policy measures.

There is no reason to believe that the Robot Age implies the Death of Work. There will be plenty of jobs for people to do… Even so, it is possible that the jobs that are available for many people are temporary, insecure, and badly paid.

The idea of a Universal Basic Income (UBI) provided by government has been kicking around for many years, but the possibility that technological change may destroy many jobs has sparked new interest in it. If Depression-level unemployment were to become the norm, the economy might need a UBI to sustain aggregate demand. Although that sounds like a liberal idea, some conservatives like it because it may enable firms to pay lower wages, and also simplify the government bureaucracy that administers the complicated system of benefits we have now. Bootle does not expect the level of unemployment that would require a UBI; nor does he consider it very workable. If the UBI is truly universal and set relatively high, it may spend too much tax revenue on people who don’t need it. If it is set too low, it provides little help to the people who need it most. Bootle agrees with economist John Kay, who said, “Either the level of basic income is unacceptably low, or the cost of providing it is unexceptionally high.” In 2016, Switzerland defeated a UBI proposal in a referendum. In 2017, Finland experimented on it with a small sample, but decided against extending it.

I was surprised that Bootle discussed UBI at great length, but rejected the idea of a Basic Jobs Guarantee (BJG) out of hand. From an economic standpoint, I would think that paying the unemployed to do some useful work, while hopefully developing their job skills, would be more cost-effective than paying them to remain unemployed. Modern Monetary theorists advocate BJG as an economic stabilizer, since the program would expand in times of economic downturn, but contract in times of fuller employment. Bootle observes that it has not generated much support yet on either side of the political divide.

What can be done for workers and families who are struggling in the emerging economy? Bootle points out that even if economic inequality is here to stay, new technologies can raise the average income, benefitting people at all levels. He also hopes that new technologies will help “improve the efficiency of public services, and hence deliver better results for the same money, which would disproportionately benefit the less well-off.” But he repeatedly returns to the issue of education, where he believes that public investments have the best hope of reducing inequality.

Conclusions

At the end of the book, Bootle restates his main conclusions:

  • The technological breakthroughs of our era will have consequences comparable to those of earlier phases of the Industrial Revolution. They will raise productivity and the return on capital, destroy some jobs but create others, and generally provide higher incomes and more leisure.
  • Smarter machines will do many things, but humans will still do some things better or at lower cost.
  • The AI revolution will liberate people from jobs that have “taxed their spirit and eroded their strength and enthusiasm,” leaving them with more freedom to be fully human.
  • The best thing the state can do to help is to fund public education, with a greater emphasis on lifetime learning and retraining.

Many readers who are familiar with the more apocalyptic visions of science fiction writers and other technological visionaries may wonder how Bootle can be so optimistic. How do we know that smart machines won’t replace humans in more dramatic ways than he envisions? Bootle realizes that in the end, defining the difference between a human and a machine requires a deep dive into philosophy and science, opening up fundamental questions that lie beyond the scope of his economic perspective. He wisely leaves these questions for his Epilogue, and even there only hints cautiously at the ultimate answers. I will comment on that section in my last post.

Continued


The AI Economy (part 2)

October 28, 2022

Previous | Next

For this topic, my posts correspond to the three parts of Roger Bootle’s The AI Economy: Work, Wealth and Welfare in the Robot Age. Part 2 of his book is titled “Jobs, Leisure, and Incomes.”

Work and leisure

He begins with the impact of technology on the division of life between work and leisure. New technologies are often labor-saving devices, but the labor they save can either be channeled into other forms of work or enjoyed as leisure. Industrialization has had some of each effect, and Bootle expects this to be true of artificial intelligence and robotics too.

Industrial work has often been rather dreary, and many workers have felt that they are little but slaves to machinery. Not surprisingly, a shorter work week has been a traditional goal of organized labor. “Between 1870 and 1998, in the highly industrialized countries the number of hours worked per annum per employee fell from 2,950 to 1,500.” Divide that by 52 weeks, and it implies a reduction in the average work week from about 57 hours to 29. If 29 hours seems low, remember that it averages together full-time and part-time workers, many of whom might prefer either fewer or more hours. It may also seem low to Americans because our work hours are at the high end among industrialized countries, most of which have more vacation time than we do.

Although the long-term trend toward leisure is very real, it can be exaggerated. In the early twentieth century, John Maynard Keynes envisioned a reduction of the typical work week to 15 hours. Many social scientists have been surprised by the long hours many professionals work today.

Economists have looked for a relationship between hours worked and personal income, but the relationship is not simple. On the one hand, economists see an “income effect” in which some people see less need to work as their hourly income rises. On the other hand, economists see a “substitution effect” in which the substitution of leisure for work imposes the greatest opportunity cost on the highest-paid workers. They use this second effect to explain why so many well-paid workers seem so driven by their work.

Complicating the picture are all the non-financial reasons that people work: because they find the work intrinsically meaningful, enjoy winning in a competitive environment, enjoy the social relationships on the job, and so forth. Also, some long hours may be involuntary, imposed by employers as a condition of employment. Once upon a time, the wealthy were a more distinct leisure class, but today’s high-paid workers face more pressure to work, whether they like it or not. “Those wanting to work more are typically low-paid waiters or cleaners. Those wanting to work less are typically highly paid doctors or other professionals.”

In this complex situation, the effects of smarter machines are not likely to be simple. On the one hand, people who lack technical skills may have even more trouble finding as much work as they want. But people who are comfortable with new technologies may find the high-tech society liberating. They can turn many tedious tasks over to labor-saving devices, while enjoying the good pay and reduced hours that come with enhanced productivity. Many of the qualitative benefits of meaningful and creative work can also be achieved in unpaid activity, such as creative hobbies and volunteer work. As usual, Bootle emphasizes the positive potential of the AI economy rather than its dehumanizing possibilities. “Individuals themselves and society as a whole will be able—and indeed will need—to choose the right balance between work, rest, and play. And by ‘right’, I mean the balance that best suits them.”

Bootle reports that the countries with relatively low work hours, particularly northern European countries like Denmark, score higher on measures of personal happiness. A further reduction in work due to advanced technologies is not necessarily something to fear. That’s assuming that a lot of human work remains, and that people who want paid employment can get a decent piece of it.

Future jobs

Bootle complains that too many prognosticators lack the imagination to envision a world in which human activity is different but not necessarily less rewarding. They underestimate how much we humans will adapt to our changing technologies, as we always have. We will do less of some things but more of others, creating jobs that we can hardly imagine today. Who in an earlier phase of industrialization imagined the internet or the job of software engineer?

Bootle is highly skeptical of complete machine autonomy, believing that humans are much more likely to be assisted than replaced by technology. He discusses the possibility of self-driving vehicles at some length, concluding that computer-assisted driving is much more feasible. (I love my car’s smart cruise control that adjusts my speed to the car in front of me and keeps a safe distance.) Completely driverless cars have a lot of drawbacks: inability to anticipate every possible situation, very high cost, the threat of being taken over by hackers, and the deterioration of human driving skills so you don’t have them when and if you need them. The dangers of autonomous weapons systems are even more obvious. One of the manpower challenges in the military is that so-called “unmanned platforms” actually need to be monitored by technically qualified and sensible people.

The impact of artificial intelligence can be different from what people might expect based on past experience with automation. It is no longer the case that mental work is irreplaceable by machinery, while manual work is most vulnerable. Routine mental work like translating, interviewing or calculating is at risk, but “many skilled, manual jobs look safe for the foreseeable future. These include plumbers, electricians, gardeners, builders, and decorators.” A model for the worker of the future will be a creative person with the skill to work with a robot. “The losers in the AI economy will be those who are doing essentially robotic jobs and/or are themselves irreducibly robotic when dealing with other people.”

Although some jobs will no doubt disappear, Bootle sees many areas of increased worker demand—health care providers using artificial intelligence to diagnose conditions and robots to assist surgery; more technically assisted caregivers “as people get richer and older”; organizers of leisure to provide interesting and creative experiences off the job; human relationship experts; and of course, technicians to design and program the robots and help people learn to use them. He believes that “new jobs that barely exist today will spring up and multiply. Accordingly, I see no reason why the robot- and AI-infused economy of the future cannot be accompanied by full employment.”

Impact on inequality

Economic inequality within developed countries has been increasing, and the unequal distribution of technical skills is one reason why. Globalization aggravates the situation, since it puts workers with less education and skills in competition with similar workers in poorer countries. Some observers think that additional technological advances can only make things worse. Yuval Harari says, “In the coming century or so, humanity will divide into two classes of people—the gods and the useless.”

Bootle also discusses the potential for greater inequality between businesses. Producers of a popular digital product or service can dominate markets around the globe. This creates “winner-take-all” markets where millions of consumers can choose the global leader over local competitors. On the other hand, cutting-edge technologies can also be very disruptive. A startup company or new product can rise up to challenge a market leader, just as Apple challenged IBM in the 1980s. Entrepreneurs can rise from obscurity to fame in a few weeks.

Although people will need new skills to participate fully in the AI economy, opportunities to acquire skills should also abound. The situation reminds me of the need for mass literacy in the twentieth century. The response was a huge expansion of public education, facilitated by rising productivity in industry and the mass movement of children from farms and factories to schools. If new technologies create a productivity revolution in services, technical training services ought to be more widely available at modest cost. A new generation with early exposure to digital technologies may acquire useful skills pretty readily.

Technological advances have the potential to bring down the cost of services in general, just as assembly-line production brought down the cost of manufactured goods in an earlier era. If so, even the relatively worse off may be able to live better than they do today. As of now, the cost of higher education, legal services and health care remain stubbornly high, but the AI economy may find ways to make them more cost-effective.

Taking a moderate position, Bootle predicts neither a major increase nor a major reduction in inequality, but he does anticipate continued improvement in living standards. He says that “it is likely that the balance of the factors listed…will enable the mass of people to enjoy increasing incomes even as robots and AI start to proliferate. And it is perfectly plausible that there will not be a significant increase in income inequality.”

Continued


The AI Economy

October 24, 2022

Previous | Next

Roger Bootle. The AI Economy: Work, Wealth and Welfare in the Robot Age. Mobius, 2021.

One of the most intriguing and important questions of our time is how advances in artificial intelligence and robotics may transform how we work and how well we live. Roger Bootle brings an economist’s perspective—and I think a healthy does of realism—to the question. He is interested in countering some of the more fanciful visions of the future that may make good science fiction, but are too unhinged from current economic realities. Some of these are apocalyptic, such as Stephen Hawking’s statement that “the development of full artificial intelligence could spell the end of the human race.” Some are quasi-religious, like Ray Kurzweil’s suggestion that humans could achieve eternal life by merging with intelligent machines and transcending our physical bodies.

Bootle acknowledges that the future is always somewhat uncertain, but he finds some projected futures, especially near ones, more plausible than others. He believes that macroeconomics can help keep the discussion grounded in reality. Although the terms “artificial intelligence” and “robot” may immediately make us think of humanlike entities, he reminds us that robots remain just “mechanical devices that can be programmed to act in certain ways.” From an economic standpoint, we should think of them as capital equipment rather than human labor.

The industrial analogy

For Bootle, the key question is, “Are the changes taking place now, thanks to robots and AI, essentially a continuation of what we have seen since the Industrial Revolution? Or are they something quite different?” He believes that the continuities and similarities are strong enough to make economic history very helpful in anticipating what is coming.

First, he points to the remarkable surge in economic progress associated with the Industrial Revolution in its various phases. Estimated global per-capita GDP was over thirty times higher in 2000 than it was in 1800. Economic revolutions had occurred before, notably the Agricultural Revolution that brought the planting of crops and domestication of animals. But they took place much more slowly over many centuries, and the gains in production were smaller and largely offset by human population increase. The Industrial Revolution was not only faster, but it involved social changes that complemented technical developments like the steam engine and electricity. Sustained capitalist investment in such technologies raised productivity by increasing the amount of capital equipment per worker. Acquisition of new skills added human capital to the equation. Increases in trade encouraged divisions of labor and economies of scale.

Although rapid by historical standards, improvements in living standards did not come overnight. According to economic historians, “it was only after 1870 that European real wages rose decisively above medieval levels, with Britain leading the way.” Another big leap in living standards occurred during the postwar economic boom of 1945 to 1973. The average worker gained purchasing power despite working fewer hours than previous generations. Then that boom ended, and more recent decades have seen slower growth and more unequal distribution of benefits.

While the Industrial Revolution boosted productivity and incomes in the aggregate, workers in all industries did not benefit equally. Labor-saving technologies destroyed massive numbers of jobs, especially in agriculture. At the same time, it created many new jobs in manufacturing, some of which eventually became relatively high-productivity, high-paying jobs like auto worker. Contrary to the fears of many social critics, technological change neither reduced employment overall nor lowered per-capita wages. Today the locus of job creation has shifted toward services, and the future of work and income depends mainly on productivity and pay in that sector.

Is this time very different?

Some writers think that the impact of artificial intelligence and robotics will be markedly different from the industrialization experience. Bootle discusses their arguments, but does not find them very convincing.

Robert Gordon has argued that the Industrial Revolution introduced a unique set of technological developments, and nothing resembling it is likely in the foreseeable future. The slower growth in investment and productivity in recent decades is the new normal. Bootle argues that “the digital revolution needs time to play out,” and that “we are on the brink of new developments that promise to bring rapid advances: biotechnology, nanotechnology, 3D printing, robots and AI.” He expects new technologies to “reinvigorate the engine of growth which has seemed to sputter and stall over the last two decades.”

At the opposite end of the spectrum from Gordon are thinkers who are so impressed with new technologies that they expect impacts even more momentous than those of the Industrial Revolution. They usually assume that whatever humans can do, intelligent machines can do better—or more precisely, learn to do better now that they can master tasks with more autonomous learning processes. Machines of the future will replace not only physical labor, or highly repetitive mental labor, but potentially all labor.

Bootle takes an intermediate position. He notes that predictions of massive unemployment from automation remain speculative, since “many of the countries with the highest number of robots per worker (e.g., Singapore, Japan, Germany) also have some of the lowest unemployment rates.” He believes that living human beings have qualities of consciousness, creativity and emotional intelligence that remain beyond the capacity of dead machinery. He also cites “Moravec’s paradox,” that robots often have more trouble with tasks humans find simple (tying a shoe or getting a joke) than tasks we find complex (calculating a statistic or remembering the names of all U.S. presidents). Although some forms of human work will disappear, other forms will persist or expand. The areas of expansion are most likely to involve jobs where human labor is assisted and complemented by intelligent machines. That creates the potential for a long overdue productivity revolution in human services such as education. As always, some will benefit more than others, but on balance the results should be beneficial for our quality of life.

The macroeconomic impact

If the digital revolution turns out to be another phase of technological progress in continuity with earlier phases of the Industrial Revolution, then its long-run impact on the economy ought to be positive. We ought to see a period of sustained high investment, gains in productivity, and strong economic growth.

If anything like the vision of more rapid growth of productivity expounded here is realized, then the result will be a faster rate of economic growth than has been normal in the West over recent years, and a faster increase in living standards, even on the inadequate measurements that we have, such as GDP or real incomes, or real consumption per capita. Moreover, because of increased leisure time, the underlying rate of improvement in the human condition will be still greater.

Economic arguments always come with a lot of ifs, ands, and buts, however. Economists want to identify the economic mechanisms by which greater productive capacity translates into higher purchasing power for the average household. In theory, “production gives rise to income that can be spent to buy what has been produced,” which keeps the cycle of investing and spending going. Suppose, however, that most of the income is going to the owners of robots, leaving the workers too poor to afford many of the goods and services being produced. That resulting shortfall in aggregate demand then discourages new investment, so that the economy’s productive potential cannot be realized.

One possible answer is that the emerging economy will need a lot of human capital (skills) as well as physical capital (equipment). That will give qualified workers leverage to claim a reasonable return on their contribution to productivity. Workers will have to compete for good jobs, but employers will also have to compete for qualified workers. Of course, this assumes that workers can obtain the skills they need to work productively with the technology.

Bootle also hopes that other systemic factors will support high aggregate demand. He says that the aging of many populations has created a large generation of people saving for retirement, but they should spend more when they are actually in retirement. Also, countries like China and the oil producers have held wages and spending down in order to promote their cheap exports to richer countries, but show signs of increasing their demand for consumer goods in the future.

This discussion may strike the reader as a little irrelevant, given the recent inflation. The problem right now seems to be excess aggregate demand, given the recent supply shortages, tax cuts and stimulus spending to counter the Covid economic disruptions. It takes a little imagination to envision a world in which a significant surge in productivity threatens the economy with an abundance of supply but a shortfall in demand. But that was the problem preoccupying many economists in an earlier era of industrial expansion. We have been there before, and could be again.

In later posts, I will go on to discuss more specific topics in The AI Economy, such as jobs, leisure, and macroeconomic policy.

Continued