Will a robot steal my job?

It’s rather thrilling to imagine bands of robots skulking in corporate corridors, waiting for humans to pop to the coffee machine so they can nip in and take their jobs. Just think: you could come back from lunch to find an algorithm sitting in your chair, working far more efficiently and effectively than you ever managed, for a fraction of your take-home pay!

Or maybe not.  

Looking at what the robots can currently do, and what most jobs consist of, we’re facing a messier – and largely incomplete – handover.  

Right now, robots are brilliant at tasks: glorified spreadsheet calculations, data-mining, pattern-recognition, precision mechanical movements (fifteen years ago a robot surgeon cut out the lining of my left knee while the human surgeon sat at a computer in another room – I’ll post the video sometime when I’m feeling gruesome).  They plough through huge amounts of data almost instantaneously, they reproduce processes with precision and consistency.  They don’t get distracted, tired, careless or fed up.  They do what they are told to, and don’t get uppity about equal pay, corporate responsibility, work-life balance or any of those other human distractions from the capitalist dream of workers producing the most profitable results with the least investment in the shortest possible time.  Robots – unsurprisingly given the derivation of their name – rock at servitude-conditions task-fulfillment.

What robots are less good at are jobs: combinations of tasks with other less-bounded activities.  The variety in job-tasks alone would defeat most robots: it’s one thing to design a chatbot to claim for reimbursement of unfair parking fines or to report on a sports match; quite another to build something that can argue a case in court or explain how Lincoln City became the first non-league club in 103 years to reach the FA cup quarter-finals.  AlphaGo might have whipped even the best human champion at Go, but it could not even deal the cards for a game of Snap. 

There’s also the inherent complexity of superficially simple tasks, such as recognizing objects or even words.  We have all howled at GoogleTranslate’s errors, and felt a shiver of horror when an autonomous car failed to recognize a woman crossing the street because the heavy bags she had hung from her bicycle’s handlebars had no model in its training data set (yes, yes, I know people not machines were responsible for 93% of the 37,000-odd car crashes a year in the US alone, but still).  Today’s robots are remedial learners when it comes to inference, or common sense as we humans call it. Stray even slightly from examples in their training data set and they go all wobbly. 

But the real problem with robots taking over our jobs isn’t task-mastery, it’s doing all those other general activities that make jobs work – things like judgement, self-management, cooperation with others, adaptation of tasks or schedules to changes in circumstances.  Machines can’t self-manage efficiently or, so far, effectively.  They still need humans to set context, to adjust and calibrate, to control them.

So, are we wrong to worry about automation?  Surely if we can get robots to do the more routine or fiddly tasks, that will free us up for the interesting problem-solving, creative and connecting work that forms the best bit of most of our jobs?

Probably not.

Automative capitalism has a history of reducing jobs to tasks, cutting the workforce and letting the customer pick up the slack.  Look around your office at senior execs jabbing ineptly at keyboards to type their own emails, wasting twice the time it would have taken to dictate to a secretary on a non-value-adding task.  Check your fitness tracker movements next time you swipe and bag your own groceries (and stamp your feet in frustration, curse the idiot machine, spin around waving wildly for help) at an automated cashier.  We get stupidly cheap when it comes to paying for services – we look at the price tag and forget the cost of our own physical and mental labour.  We will pay for tasks rather than jobs, never mind if we get a worse outcome as a result.  And the whole point of substituting tasks for jobs is to employ fewer people at lower overall cost – not good for job-creation.

But more fundamentally, jobs don’t really work when you pull out the individual tasks.  Part of what makes a good lawyer is a deep knowledge of case law, built up through formative years of detailed research.  As we see with language learning, cutting to the chase of the decision point without doing the background research does not result in deep knowledge – Duolingo is addictive, but it won’t help you chatter away in Italian.  To do that, you have to dig down, learn the grammar and practise in live, interactive, repetitive situations.  Our brains become creative when they are exposed to multiple rounds of learning the hard way.  By doing the tasks that might seem routine but that allow us to develop the skills we need in a specific context to do the interesting work.  There’s also emerging evidence that humans don’t much enjoy just doing tasks rather than jobs: gig-work results in increased stress, only partially because of the weaker employment protections.  There’s something about humans that grows through experiencing the full learning cycle.

So, should we just stop automation?  Switch off the robots and go back to the Bronze Age? 

Even if we wanted to, we could not - and I for one would not welcome a world without modern dentistry, major motion pictures and year-round avocado toast.  Plus I still think there are big reasons to be optimistic about the future of jobs in the age of automation.  But we need to think much more seriously about how humans work, about how robots work, and about ways in which we can work together to bring out the best in both of us.  We need to get smarter about how we use robots, and how we adapt ourselves to work with them – maybe even for them – better.  

In future posts, I’ll be exploring some specific ideas: what new ways of working we might have to develop as humans; how to construct meaningful jobs; how to decide whether something is better done by a robot or a human; examples of people and robots working together and at cross-purposes; alternatives to jobs and alternative uses for robots apart from doing human work.

For now, go get that coffee in peace.  No robot is going to be taking your job this week…though in the frantic run-up to the holiday season, you might wish that one would.