Jerry Kaplan of Stanford, author of Humans Need Not Apply, had an interesting post in The Atlantic this morning, suggesting that male-dominated fields may be more susceptible to robot replacement than female-dominated fields.  I think it’s wrong.

Here’s why: I’ve been closely monitoring news on robotics and artificial intelligence for a couple months now, and so far from what I’ve seen the areas where people are plunging in and launching startups don’t seem to line up with this basic idea that physical work replaceable, psychological/caring work irreplaceable. Just yesterday in my news roundup, I included two links to new startups working different approaches to monitoring senior citizen’s health care with A.I.  And I myself I recently interviewed a researcher whose whole thesis is on how robotic assistants might be used to aid disabled or elderly people. So far, I’ve seen nary a sign of a house-painting robot.

Because, it turns out, navigating autonomously through an undefined space is a really, really difficult problem. (It was one of the big challenges for Whitney Crooks, the researcher I interviewed.) So far most of the most successful robots that do navigate on their own — Amazon’s warehouse bots, or the Roomba — are practically two dimensional, stuck on the floor of a single room. The recent DARPA challenge — in which many of the bots were in fact being remotely controlled by human operators — saw the vast majority of the field fall down go boom, repeatedly, when confronted with problems like climbing stairs, walking through uneven terrain or pulling themselves through confined spaces.

That’s partly because vision is itself an extremely difficult problem — more than half of our own brain matter is devoted to solving it. Houses come in all shapes and sizes; they have gables and cornices and windows and frames. A house-painting robot that’s 96% accurate would still be one that might paint a stripe through your rose bushes.

That’s not to say that the house-painting problem is unsolvable — as with chess, a combo of man plus machine might be far superior to either alone, and even if you only replace three of the four guys on a painting crew with one robot, that’s still 75% of the painting jobs.   

But the thing is, the examples Kaplan uses of emotional, psychological labor…I’m not sure they are. He suggests, for example, that you might need a human nurse to watch over a dementia patient and know when to give them a sweater.

But while handing out a warm sweater might be what you’d do if you had a carer nearby, that’s not the only way to solve the Chilled Granny problem — you could also just raise the ambient temp in the room to something a little more toasty, and that could be done remotely. Further, while noticing Granny is cold is certainly something a nurse could do, it’s also something a slightly souped-up Fitbit or AppleWatch could do, by monitoring changes in pulse rate, skin clamminess, or micro-vibrations that might indicate shivering.

In fact, with home security services like ADT plunging full speed into the Internet of Things, it’s easy to imagine some future health tech monitoring service, with an operator or two manning a room of screens, coupled with a field team to respond to emergencies and make house calls. (One of the studies i linked to yesterday is already headed in this direction.) That kind of job wouldn’t necessarily replace nurses or home carers one for one — but it might help prevent or delay people from needing to move into nursing homes. Many people are already willing to install monitors in their cars for a break on their car insurance; how many seniors would be willing to strap on a wristband and install a video camera or two in their homes if it meant they could stay in them?

All of which is to say that what we think of as caring is to a large degree, sensitivity — using our senses to observe and draw conclusions about the states of others. And to the extent that something can be reduced to a sensory problem, an issue of monitoring and recording data points, it becomes an information problem. We are already very good at information problem, and getting better everyday. It may well be that to the extent that job replacement involves manipulating the physical world, it will be difficult. To the extent that job replacement involves manipulating information, it will be easy.