The topic of artificial intelligence’s looming impact on the US economy is one both Hillary Clinton and Donald Trump studiously ignored during the 2016 campaign for President, choosing instead to focus on job losses from bad trade policy, despite the fact that the vast majority (at a ratio of about nine-to-one) of good-paying manufacturing jobs that have disappeared in the last 30 years disappeared because of new technology, not because they were shipped overseas.
This silence comes despite warnings of AI’s potentially disastrous impact, even by technologists and scientists like Stephen Hawking, Bill Gates, and Elon Musk, not to mention alarm bells sounding in the halls of academia and offices of economists.
Amid that debate, during the closing days of the Obama Administration, the Executive Office of the President released a policy paper to very little fanfare, titled, “Artificial Intelligence, Automation, and the Economy.” The US political leadership was (and remains), curiously asleep at the switch after the paper was released.
Repeated requests by The Technoskeptic to the House Subcommittee on Research and Technology yielded not a single call back to comment on the new policy. Requests for comment from a former technologist that served in the Obama Administration’s Office of Science and Technology Policy (OSTP) were met with silence. Similarly, we received no reply to a recent request directly to the OSTP to discuss the policy paper, to determine whether the Trump administration accepts or repudiates it.
In the meantime, like much of the rest of the federal bureaucracy, the OSTP at the White House is now critically understaffed. According to an April 2017 article in Motherboard, of the 114 positions in OSTP, more than 70 are still not filled. And of the positions that are filled, none is specifically tasked with the tracking the issue of artificial intelligence.
Eisenhower Exec. Office Bldg., home of OSTP
Unless the Trump administration issues new policy guidelines or Congress gets involved, every federal agency will nominally follow the Obama administration’s AI strategy. When it comes to thinking about AI, right now our goverment is, ironically, on autopilot. This means executive agencies will be making assumptions and shaping outcomes based on a policy that no national politician seems to want to discuss.
Since our highest political leaders don’t seem to care about the issue (perhaps because the numbers of Senators and representatives are mandated by law, and so they never need fear being replaced by a piece of software), it might behoove the average American to have at least a basic notion of the approach to dealing with AI our leaders have quietly adopted. Because hidden in the painfully dry economic jargon that make up the 55-page report are assumptions and intentions that, if acted upon, may lead to horrific outcomes for many, many Americans.
One of the biggest assumptions underlying the whole paper is that we have some ability to predict AI’s effect on our economy, despite the fact that there has never been anything like AI before. AI is not just smarter computers. It is computers and software and robots so advanced they will not only replace low-skilled jobs like driving a truck, or medium-skilled jobs like manufacturing and construction, but high-skilled jobs like software development, law, medicine, and so forth, ad nauseum.
Instead, the report clings to the conventional economic wisdom that AI is good as it will increase labor productivity, and “labor productivity generally translates into increases in average wages.” This completely ignores that in our current era of high-tech, wage growth has been flat for decades in the US. And even though high-tech has already resulted in the loss of more than six million manufacturing jobs in last 20 years, the report asserts “AI should be welcomed for its potential economic benefits.”
The paper also incorporates a vast range of estimates, which makes it hard to see how it can be considered usefully prescriptive. It alleges AI will cause the loss of between nine and 47 percent of jobs in the next two decades. One is a minor annoyance; that many jobs are routinely lost and replaced every few years. The other is a potential social and economic disaster that could mean the return of breadlines, even sharper political divides, and unemployment riots that could open the way to creeping authoritarianism.
The report even concedes that mass unemployment scenarios are not unlikely. Artificial intelligence, it says, may result in “a superstar-based technological change, where the benefits of technology accrue to an even smaller portion of society than just highly skilled workers. The winner-take-most natures of information technology markets mean that only a few may come to dominate markets. If labor productivity increases do not translate into wage increases, the large economic gains brought about by AI could accrue to a select few. Instead of broadly shared prosperity for workers and consumers, this might push towards reduced competition and increased wealth inequality.”
For readers who aren’t policy wonks, that scenario translates to, “The few folks who make the AI will get the money you used to get paid to do your job. You are, in other words, quite likely screwed.”
The report carries an implicit assumption that many new AI-assisted jobs will be created. Although how numerous those jobs might be, and whether they will be present in the numbers to make up for the millions we know will be destroyed, was not addressed. It is treated as an item of faith that somehow new jobs will appear, because they always have.
The report claims that AI will create new kinds of work, delineating one potential growth area as jobs that use “augmented intelligence,” in which, for example, AI does the hard work of diagnosing an illness, but a health care worker still tells the patient their options and administers the treatment. Not necessarily a bad thing for Americans, who suffer the highest health-care costs in the world…but one likely to put a lot of doctors and nurses—and medical schools—out of business (and really, who doesn’t love interacting with diagnostic software when your life is at stake?).
iStock.com / Petmal
Other jobs it predicts arising out of AI are development of new technology, although why a lot of new jobs would be generated for those roles when we already have software writing software right now, is also not explained.
Another growth area is supervision and repair of the robots/AI that will replace countless human workers, a.k.a. “servicing our robot masters.”
Finally, we might get new jobs, per the report, as a “response to paradigm shifts,” the idea being that new jobs we cannot foresee will be created. The example it gives is that in the case of autonomous vehicles, we might need more urban planners to adapt our cities for autonomous traffic. How hiring, say, 100 new urban planners would be a net gain for a city’s economy after autonomous vehicles put 10,000 local delivery drivers out of work is not addressed.
What is crystal clear is that the people who are already the poorest are the most vulnerable to losing their jobs to AI. The report estimates AI will threaten up to 83 percent of jobs that pay less than 20 dollars an hour. How that one statistic alone doesn’t have labor economists and reporters shrieking that the sky is falling remains baffling.
Among the other fun facts in the report was that right now, labor’s share of gross domestic product is at a historic low, and the developed country where that trend is the worst is the US. Income inequality is already at a historic high in the US, as the top one percent of society now gets about 20 percent of the GDP. Oh yeah, and the productivity growth from incorporation of high-tech that was supposed to lead to higher wages? That doesn’t exist. Per the report, productivity growth has dropped off dramatically since 2005. And a final fun fact: the US now spends half of what it did 30 years ago on retraining programs that might help workers cope with a job lost to technology.
The report would have us believe that the way to deal with the flood of AI-driven job losses is education, retraining, and economic aid for displaced workers. To support that assertion, it compares the impact of technology on the job markets of several high-tech Western industrialized nations. According to the comparison, there are high-tech countries that have maintained high wage levels despite the increase tech and automation, but the countries that maintained wages and employment had a much more robust social safety net and retraining program. Germany, for example, has kept wages and employment high, but also spends six times, as a percentage of GDP, what the US does on such support programs.
So while the report rightly says we need to have better unemployment insurance and much more extensive retraining, it also requires that people get used to “lifelong learning” (ignoring the fact that a human’s ability to master new skills has been clinically proven to decline with age). Even were those programs put perfectly in place, it implies the solution to the crisis of AI is to be constantly getting on and off a training treadmill for a person’s entire working life.
Employment stability, what there was of it, evaporates like a morning dew after a hydrogen bomb. Workers are left in constant fear of job elimination and the need for retraining, during which they get paid little or nothing. Which is going to feel like some version of hell to a lot of people…and that is the best case scenario. This, ultimately, is where the report is fundamentally most flawed.
Let us put aside for the moment that a significant chunk of the population would have either no interest in, or aptitude for, such compulsory lifelong retraining. Where would the money for the retraining and much larger social safety net come from? On that topic, it is wildly optimistic at best. It does suggest the obvious, that the companies that make and use AI be taxed at a higher rate. But can anyone assert with a straight face that any Congress would increase spending in this arena by six times? And that the money would come from the politicians’ biggest political donors, which are—you guessed it—high-tech companies? The same high-tech companies current sitting on a stash of 2.5 trillion dollars in profits in overseas banks, waiting until this administration, or the one that follows, declares a corporate tax holiday, so they can repatriate the funds back to the US? Do we have any realistic chance to reach funding parity with Germany and other high-tech economies if we depend on squeezing the taxes for it out of Apple, Google, and Microsoft, who have already proven to be, not only masters of technology, but of tax avoidance?
And what if we don’t get the funding to help cushion the worst of the transitional shocks of AI, and yet we still embrace AI, invest in accelerating AI research, and do other things to increase the adoption of AI as the report recommends? What outcome will we get?
Without the absolutely crucial support to help displaced workers of all skill levels cope with loss of their jobs, haven’t we just knowingly set the stage for a political and economic catastrophe, the likes of which ignited WWII and other horrific, global conflicts? What happens to a country of 300 million guns without enough jobs to go around?
If the report’s authors did a little better job of explaining how the increased use of AI might help us avoid that nightmare scenario, instead of marching us squarely into it, one might be a little more sanguine about how AI might be a boon to our economy.
What we have, instead, is 55 pages of techno-Pollyanna fantasy. Despite acknowledging the many ways in which AI could turn out to be an absolute economic disaster, unless we get almost everything to do with preparing for and managing AI right, the authors still assume, with few facts in evidence, that AI will be a net positive.
Having read and thoroughly digested “Artificial Intelligence, Automation, and the Economy,” one cannot help but wonder if future generations will not read the same document and shake their heads in amazed disgust.
“How,” they will say, “could the authors fail to see that even at the time the report was written, let alone in our far off future, AI was already exacerbating the conflict between the need for millions of meaningful, well compensated jobs, and the financial interests of tech barons and AI-mongers?”