Aug 062016

The world continues to be abuzz with talk about the mixed promise and peril of artificial intelligence, both narrow and general. One of the specific concerns is that the robots are coming for our jobs. Fears of technological unemployment are not new, but this time it’s different. This time the machines are coming for every job.

A key question folks are asking is: which jobs will the robots take first?

The conventional wisdom seems to be that robots will take so-called “low skill” jobs first, and that the “high skill” professions will be safe for somewhat longer (although not much – a decade or two perhaps). But this thinking might actually be backwards. The reason why is that professional work is often high-stakes work. And who do you want doing the job when your property or future or even your life is on the line? I think the answer is clearly machines.

Brain surgeons and anesthesiologists get paid big bucks not just because their skills are so rare and therefore in short supply, but also because people’s lives are on the line. Lawyers are paid handsomely for similar reasons: people’s futures and livelihoods hang in the balance. Corporate executives make a good deal of money because of the high-stakes decisions they make that affect their company’s shareholders and employees. And so on.

The following matrix maps occupations into four quadrants based on a rough assessment of the skill (as measured by educational barriers to entry) and stakes (as measured by the risks to life and property from incompetence) involved:

TU matrix
















Cashiers, for example, have already been partly replaced at supermarkets. But the stakes are low, so how much do consumers actually care whether it is a human or a machine doing the job? The answer, I suspect, is not much. The capacity already exists to replace more jobs like this with machines, but the incentive to do so is largely on the supply side.

But for high-stakes work the answer might be very different.

To start, take the example of self-driving cars. Public opinion here already seems fairly clear: these are going to put a lot of people who drive for a living out of work. And not just because it will make taxi fare and other services cheaper, but – very importantly – because once we can trust the machine to with the job, we will overwhelmingly prefer that a machine do the job. Consider: in 2025 once self-driving cars are a mature commercial technology, which would you prefer drive your spouse or child home at 2am – one of their drunk friends, a human taxi driver at the end of a 12-hour shift, or a self-driving Tesla? This is already a major line of Tesla’s PR and marketing: autopilot saves lives.

Driving is not a “high skill” job, at least in the sense that the educational barriers to entry for the job are low. I suspect consumers would choose the machine alternative in other similarly “low skill” work where lives and property are at stake as well. Would you rather a person or a robot cook your food? (I trust I don’t need to conjure stomach-turning imagery here). Would you rather be ticketed by a completely unbiased and impersonal robotic police officer, or a human one? (If you don’t happen to be an American with brown or black skin, take a moment to imagine how your answer might be different if you were). From a safety point of view, would you rather human construction workers or robotic ones had built your home? And so it goes down the list.

OK, now what about jobs in the high-high quadrant? These professional jobs are the ones that are ostensibly safer from AI takeover, at least for a while longer than their “low skill” counterparts. But is that really the case? An IBM Watson AI system in Japan just successfully diagnosed a rare genetic disease that had a patient’s human doctors stumped. And beyond just diagnosis, I know that if I were given the choice between a robot’s unerring precision and a human brain surgeon’s shaky hands, I would take the robot every time. The same, I’m afraid, would go for my lawyer or tax accountant (if I had one).

So, contrary to conventional wisdom, once AI are viable alternatives to humans in high-stakes occupations, I suspect there will be swift and overwhelming demand for the machine option – even if those options remain expensive. And if artists, musicians, writers, and scholars like myself remain employable somewhat longer, it may actually have little to do with the inability of machines to match our skills, and much more to do with the fact that the stakes to life and property are so much lower.

Bottom line: if you’re worried about automation putting you out of work, the stakes of your job may matter more than the skill required to do it. And – unintuitively – low-stakes jobs may well be less vulnerable than high-stakes ones. At least for a little while.

Leave a Reply