By Kenneth Anderson
Advances over the last couple of years in automation and robotics are drawing attention on account of their implications for the economy and especially employment. It’s not clear whether this is owed so much to an actual acceleration of technological advances as to the public noticing that the giants of information technologies – Google, for example – have decided to start placing big bets on robotics. These include self-driving cars, the “Internet of Things,” assistive care robots for the elderly and disabled, commercial drone aircraft, and many other emerging, but also potentially disruptive, automation and robotic technologies.
Tech companies likely would have done much of this anyway. There’s a convergence of interest in these technologies from many different directions. Still, the aging of the Baby Boomer population bears noting as an indirect economic driver for new machines, to make aging Baby Boomer care and maintenance more possible and affordable.
Mobile, connected, supplied, and independent within one’s own home for as long possible. Well, here you have all four: Google’s self-driving cars, Apple’s evolving iPhones as personal controller of your assistive devices, Amazon’s home delivery, and the in-home assistive robots many companies are trying to design.
The tech companies, it’s only partly an exaggeration to say, are firms whose business plans are based on old people. It’s the services they want and need as they (“we,” let me be honest) grow older, striving to maintain autonomy and independence for as long as possible. I might think (I do think) it’s unbelievably cool and not-old at all if one day I were to get a delivery by Amazon drone. But actually, after the (thrilling) first time, it’s just because I’m old, don’t really feel like leaving the house, and anyway am too infirm to do so. At least, not without a neuro-robotic “weak-limb support suit” for my legs, so I don’t lose my balance in the street and fall, and a Google car to get me to … the doctor’s office.
“Indirect” economic driver, however, because the Baby Boomers would not be paying out of their own pockets directly for many of the technologies that might be most important to them in retirement; government would wind up paying. The concomitant uncertainties, political and otherwise, that would affect what amount to “procurement” decisions within Medicare and such programs, make decisions to go into expensive and technologically thus far unproven research and development (particularly with respect to the most ambitious artificial intelligence robotics) economically uncertain, contingent propositions.
But now a familiar debate has broken out – around the employment effects that are likely to come from these new technologies. (The Economist magazine summarizes the debate and comments in this week’s edition.) On the one hand, innovative, disruptive technological change is nothing new. The result has always been short-to-medium term creative destruction, sometimes including the destruction of whole occupational categories – followed by longer term job growth enabled by the new technologies or the increased wealth they create.
In any case, over the long run, increases in the standard of living can only come through innovation and technological advance that allows greater economic output to be extracted from the same or smaller labor input. In a world of many elderly, retired Baby Boomers and historically smaller worker base bearing much of the cost of the elderly living and health care costs, that has to matter a great deal. Ben Miller and Robert D. Atkinson make the positive case for automation and robotics along these lines in a September 2013 report from the Information Technology and Innovation Foundation, Are Robots Taking Our Jobs, or Making Them?
On the other hand, maybe this time is different. That’s what MIT professors Erik Brynjolfsson and Andrew McAfee argued in their 2012 book, Race Against the Machine, and reprise in a more nuanced way in their new book, released last week, The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Machines. Maybe “brilliant machines” will replace many workers – permanently. Even without making sci-fi leaps of imagination for the capabilities of the short-to-medium term “artificial intelligence,” the emerging machines (as their designers intend them) aim to be not just “disembodied” AI computers, but instead genuine “robots,” possessed of mechanical capabilities for movement and mobility, and sensors, both of which are advancing rapidly technologies – not just AI computational abilities. Perhaps this combination – the AI robot able to enter and interact in ordinary human social spaces – does signify a break from our past experiences of innovation as (eventual) producer of net new jobs over time. Maybe significant new categories of work don’t emerge this time around – because as soon as one does, someone (or some thing) breaks it down into an algorithm, and then comes up with mechanical devices and sensors capable of executing the task – intelligently. As a Babbage column in the Economist put it several years ago, in this scenario, capital becomes labor.
What makes these machines special, then, is not just AI but, instead, AI embedded in a machine that can sense, decide, and then act, with mechanical movement and mobility, in the physical world – in the human social world. In that case, economically, changes could be unprecedented, with implications for both manual labor and even for the “knowledge workers” previously insulated from both industrial automation and global outsourcing. Unlike past technological revolutions, the result might not be whole new job categories gradually emerging (or at least ones that employ large numbers of workers).
It’s striking, however, that the argument over employment doesn’t quite directly concern the elderly. It doesn’t rationally engage aging Baby Boomers as consumers wanting these technologies and their services, or at least not Baby Boomers as both consumers and workers, with an interested foot in each camp. They’ll already be retired and so their own employment won’t be their issue. The employment issue in which, however, they will indirectly have an interest is the cost of human labor to care for them as elderly and infirm; that care is strongly labor intensive, in both unskilled and highly skilled (nursing) labor. The difficulties of machine interactions in ordinary human environments – versus, say, the highly controlled factory or assembly line floor – has meant that this sector has not so far been widely automated. But that is liable to significant change, if these technologies move forward – and if they are successful, they will have to become part of the calculation of the cost of care of the elderly, given that the technological shifts are not going to cheap, but the cost of both skilled and unskilled elder-care labor is only going to rise under current scenarios.
Which is to say, no matter where you stand on the automation-robotics-employment debate, if tech’s business plan is significantly about the growing ranks of the elderly as the target market, then to that extent, the employment debate is less important to the elderly as regards their own employment, but (at least if the technologies are successful) squarely in the cross-hairs of public policy on the costs of care for the elderly. It’s much more complicated than that, of course, and this is not to suggest that these are the only or even most significant factors. The business model for robotics is not simply about retirees; the market for self-driving cars, for example, if they come to work as hoped by Google and others, will be far wider than that; eventually it becomes about everyone. The point for now is only that the elderly form an important, though far from dominant, part of the markets for automation and robotics.
My own view is that Miller and Atkinson are mostly right about long-run job generation. The “this time is different” view seems to me overstated – as so often the case with AI, as Gary Marcus has noted. One should never rule out paradigm shifting advances, but so far as I can tell, the conceptual pathways as laid down for AI today are not going to lead – even over the long-run – to what sci-fi has already given us in imagination. Siri is not “Her” – as even Siri herself noted in a recent Tweet. For the future we can foresee, in the short-to-medium term, we’ll be more likely to have machines that (as ever) extend, but do not replace, human capabilities; in other cases, human capabilities will extend the machines. The foreseeable future, I suspect, remains the process (long underway) of tag-teaming humans and machines. Which is to say (mostly), same as it ever was.
The significant new job categories (I speculate) run toward skilled manual labor of a new kind. The “maker movement”; new US manufacturing trends toward highly automated, but still human-run and staffed factories; new high technology, but still human-controlled, energy exploitation such as fracking; complex and crucial robotic machines under the supervision of nurses whose whole new skill sets put them in a new job category we might call nursing technologist – these are the areas of work that point the way forward.
Quasi-manual labor – but highly skilled, highly value-added, and value-added because it services the machines. Or as economist Tyler Cowen put it in his 2013 Average Is Over, the next generation of workers will be defined by their relationships with the machine: You can do very well if you are able to use technology to leverage your own productivity. You can also do very well if you are able to use your human skills to leverage the productivity of the machine. If you can’t do either, though, you might gradually fall into a welfare-supported underclass – because the world of work, even apparently merely “manual” labor, is largely out of your reach.
