AI: Good, Bad, or…
Semi-serious question: what does AI have in common with the fighting in the Middle East? Fundamentally, the answer is little or nothing, but the media nonetheless insists on treating the two in the same way. According to the headlines, each will either benefit the country mightily or bring catastrophe. Nothing in between is on offer. Of course, a guess where the fighting will lead is fraught. Wars are that way. But there is a basis for guessing on the impact of AI and the robotics associated with it. History and logic both suggest that neither extreme is right, that the effects of AI will land somewhere between paradise and hell and probably more toward the positive side.
Why AI Will Neither Save Nor Destroy the Economy
The media choice is truly binary. Those on the heaven-on-earth side see AI ushering in a future of great societal wealth, soaring productivity, increased leisure, and among other desirable developments, freedom from life’s most odious tasks. Those on the hellish side paint dystopian pictures of mass unemployment, a dispirited and restive population in which a few benefit greatly and others live purposeless lives in penury or on a generous dole. Even developers of these technologies—among them Dario Amodei, Sam Altman, and Elon Musk—worry over employment destruction and advocate for a government guaranteed income to care for the coming unemployable millions.
Many of these fear-filled assessments reach hyperbolic proportions. Notable is a recent X post from AI startup founder Matt Shumer. His piece, entitled “Something Big is Happening,” notes that AI is already a better coder than his firm’s best human coders. He then extends this insight to declare that those presently involved in most other forms of “cognitive work” will lose out. These others include lawyers, accountants, writers, journalists, customer service workers, public relations executives, even medical doctors, and more. The Peterson Institute for International Economics forecasts a “roster of trillionaires” and otherwise mass unemployment, adding to this ugly picture an expectation of “artificial viruses.” The prestigious Brooking Institution avoids the darkest aspects of these sorts of forecasts but nonetheless sees an urgent need for a guaranteed income to deal with widespread joblessness. Brookings suggests that the firms creating the AI and benefitting from it should fund such a scheme. These are only a sample of the kinds of horrors some expect from the new technology.
Yes, AI Will Displace Workers. But History Tells a Different Story
There can, of course, be little doubt that AI and robotics, like all technological revolutions of the past, will displace some workers. Maybe the coders of which Mr. Shumer writes will need to apply their keen intelligence and fine talents to other employments, much as once-lauded efficiency experts had to when the first factory robotics appeared in the 1960s. Perhaps superior AI robots will displace miners and others in dangerous jobs as well (and incidentally fulfill the dream of storied United Mine Workers union boss, John L. Lewis, who insisted that no person should ever have to go down into the pits.) The list of displaced occupations will no doubt extend beyond these examples. Among others, it might include administrative assistants, receptionists, some journalists, junior-level graphic designers, financial analysts, language translators, travel agents, researchers, law clerks, and others.
But as has happened in all past technological revolutions, the new technologies will also create new opportunities largely unforeseen at the early stages of the revolution. Like other technological breakthroughs, AI and robotics will create great efficiencies and reduce costs that will raise real incomes elsewhere in the economy so that more people will demand more goods and services in different and unforeseen ways. These demands will create more jobs for displaced workers as well as for others, and if history is a guide, more jobs than existed before the technological wave.
Events have followed this pattern with each past technological wave. From the steam-powered machine revolution that enabled mass production to the electronics revolutions of radio and television to computers and the internet, people had an easy time identifying the sorts of jobs threatened by the new technologies. And that picture troubled them. What has always been more difficult, and few could see early on, were the business and employment opportunities that were yet to emerge. AI and robotics should follow this well-established historical pattern. A complete recording of this historical record would go beyond the space available in this article and certainly try the patience of readers (as well as the author). A few examples should suffice.
The classic case, and in some senses the most dramatic, comes from the machine revolution in England in the early nineteenth century. Mechanical looms and spinning equipment could produce more cotton and wool cloth more rapidly and at lower cost than the hand techniques previously used. Those involved in the earlier techniques were displaced and tried to stop the machine revolution by sabotaging the equipment. They called themselves Luddites. Despite their strenuous efforts, the use of machinery grew, efficiencies ensued and the price of cloth fell so that more people—many of middling means—could afford much more clothing than previously. The textile industry grew to meet the increased demand, ultimately employing more people than previously, while the increased wealth from more jobs and greater cost savings enlarged job prospects and incomes across the whole economy, including even in agriculture, which enjoyed what those involved referred to as the “wool boom.”
Several intervening technological waves later, the 1960s saw the first introduction of robotics into factories, what at the time was called “automation.” The robots replaced line workers and as mentioned earlier, the need for highly paid efficiency experts. Early in the process, this technology engendered many of the same fears current today. A joint study, entitled “The Triple Revolution,” by a large group of academics, including several economics and scientific Nobel laureates, claimed that automation had severed the once-secure “link between jobs and incomes.” President John F. Kennedy warned that automation threatened the “dark menace of industrial dislocation, increasing unemployment, and deepening poverty”—a more eloquent expression of today’s worries.
Presidents Lyndon B. Johnson and Richard M. Nixon became so concerned about job losses that each floated, separately of course, the idea of a guaranteed income to support the anticipated mass of unemployable people (another familiar sound). But as happened with textiles more than 100 years before, efficiencies and reduced costs created wealth, expanded demand, and actually raised employment in the factories even with their robots. It is ironic indeed that those romantic about factory jobs (almost universally people who have never worked in a factory) now view the 1960s as a kind of golden age of good-paying manufacturing jobs.
The pattern has repeated since. The computerization of business eliminated thousands of clerical positions, but it also enabled employment and profits opportunities in unexpected areas such as, to take just one example, next-day or same-day delivery firms that were not feasible before computerization and that today employ millions, many at higher wage rates than the positions that computerization eliminated. In finance, the computerization enabled employment growth in services to a wide group of customers that previously were only available only to the very wealthy and at great expense.
What Nobel Economists Actually Say About AI Job Losses
It is just these sorts of considerations that prompted the Organization for Economic Cooperation and Development (OECD) in 2016 to correct an earlier paper that had failed to make these broader assessments. That earlier paper had claimed that computerization would eliminate 47% of existing jobs. The correction reduced the number to 9%. More recently, the economics Nobel laureate, Daron Acemoglu, used this same broader perspective offered by history to conclude that contrary to the many fear scenarios, AI would at most eliminate only 5% of existing jobs.
As has been the case in the past, it would take an inhuman amount of prescience at this early stage in the process to itemize the new opportunities that will arise from AI and this latest wave of robotics. It would, however, also fly in the face of experience to claim, as it seems many do, that this time things will be different. Even if the picture is far from clear or complete, some of those opportunities and adjustments are already evident as are areas where AI may not be such a good fit.
The Jobs AI Can’t Do and the Opportunities It’s Already Creating
Take, for instance, Mr. Shumer’s concern for human coders. Because AI—whatever marvels it produces and it does—can only recount already existing knowledge, it can code effectively and efficiently, but it can offer very little to help determine what apps should be coded. That requires a large dollop of judgement drawn from the complex associations and background knowledge of executives and marketing experts. That same need for judgment is still more prominent in other occupations that Shumer for some reason insists are like coding—law, for instance, medicine, writing, and especially areas that deal with people. On a simpler level, AI may give the world self-driving school buses, but these vehicles will still need an adult to watch the kiddos. Robots cannot replace police or game wardens or therapists or personal life coaches. AI can make plumbing and electronic fixtures without workers, but it cannot replace plumbers or electricians or any position that requires judgement.
Media has already revealed the kinds of opportunities that will open. By making media much less capital intensive than it was, AI has enabled people to do in the back room of an apartment what not too many years ago required a studio full of specialists and expensive equipment. As this innovation has eliminated some legacy media jobs, it has also opened opportunities and expanded employment for numerous independent operators in broadcast, writing, and journalism to name just three areas. By lowering costs, AI has also reduced the risk implicit in any startup—in media and elsewhere—encouraging ventures and employment possibilities that would never have been tried in earlier, higher cost days.
These are just random aspects of the situation and far from a complete list of AI-related opportunities and sources of employment. More clarity will certainly emerge in time and no doubt surprise the AI developers themselves. That has almost always happened in the past. When James Watt made the steam engine practical, he thought in terms of pumping water out of coal mines and had no idea of locomotives or steam ships. When Marconi invented the transatlantic wireless, he saw it as a replacement for telegraph lines. He gave not a thought to radio and all that it would earn and employ. Uber never occurred to the developers of smart phones and GPS.
Some fearful of AI have gone beyond job losses and economics. Francis Fukuyama, for instance, seeing a world of unemployed people living off a government provided guaranteed income, has called AI a “dignity killer.” The Niskanen Center has noted that AI is already “smarter than the associated human capital,” implying presumably a coming era of machine dominance. Robert A Manning of the Stimson Center has gone a step further. He suggests that AI will make humans “irrelevant” to the point of “extinction.” This sort of thinking is reminiscent of the 1960’s film “2001 Space Odyssey,” in which the spaceship’s computer, HAL, outwits the astronauts to take over the spaceship. Never mind that AI could not think up such a plot, unless, of course, it had access to the film’s screenplay, but it would also need an ego to want to take over. And for all the very real marvels of the technology, it has yet to code a lust-for-power app for itself, nor is it likely to do so any time soon.
The Full Impact of AI Will Unfold Slower Than You Think
In any case, it will take a long while before all the effects of AI are fully realized. The technology may move at lightning speed, but complete adaptation will require changes in business and government organization as well as everyday practices. Regulations, legal structures, and the like will also have to change. These matters move a lot slower than lightning. It took, for instance, 20 years from the invention of the personal computer to find them on every office desk and fully integrated into how business as well as government operated. AI may take less time, but it should be evident that its effects—good, bad, and indifferent—will unfold at a pace that will allow people and practices to adjust.
