Friday, February 7, 2014

Man vs Mechanization

Predictions that automation will supplant human workers are nothing new.  Indeed, in many industries machines have already taken over jobs previously performed by low skilled workers.  Yet, overall the inevitable rise of the machines has fallen rather short of most futurists predictions.  Still, over a long enough time line it's difficult to imagine how automation doesn't supplant most low skill jobs. After all, machines in general only grow cheaper while human labor often rises in cost.

Given the machines eventual workforce dominance it's an interesting exercise to determine which jobs are reasonably safe from robotic theft.  To this end let's define what machines do well and where they fall short.
  • Machines perform routine tasks quickly and cheaply.
  • Machines work without fatigue or distraction.
  • The cost of a machine is heavily front loaded and ends when a machine is decommissioned.
  • Machines do not react to novelty with adaptive behavior. 
  • Machines are generally either fully operational or non productive.
  • Machines do not exceed their designed purpose.
Points one through three are for the most part advantages machines have over people while points four to six are deficits. Machines are generally more productive, at less cost and require no payment when they're not working (such as retirement benefits or vacation time).  These characteristics make them generally preferable workers when suited for the task. Points four through six roughly define the tasks ill suited to machine labor.

Machines do not react to novelty with adaptive behavior.  Roughly speaking this means tasks which deal with highly variable inputs are often ill suited to automation. In customer service for example workers must engage with a customer who may have any number of complaints (both reasonable and unreasonable) and attempt to satisfy the individual.  Contrarily, the only input of importance for a cashier is barcodes and currency, two largely uniform inputs a machine can easily work with. Thus self check out lanes in stores make a great deal of sense while self customer service counters do not.  Interestingly self return counters in stores would make sense (after all it's just a reverse checkout) if not for the likelihood of deception on the part of customers. The essence of this point is that we have yet to create machines which make decisions wisely.  A task must be routine and approachable algorithmically for machine labor to be efficient.  Any job which consists primarily of judgement and decisions is likely safe from a machine takeover for the foreseeable future. This includes governance, law enforcement, any profession requiring extensive human interaction (note that just because a profession currently involves human interaction does not mean it requires it),  law, insurance, any profession involving diagnosis (both health and maintenance), inspection personnel, and management.

Machines are generally either fully operational or non productive. There is the possibility of performance degradation over time, but machines are a far more binary system than people.  When a part of a machine fails it's very likely to stop completely. People are far more resilient and predictable when it comes to component failures.  If a robot doctor (an independent robot surgeon, not a tool used by a human surgeon) had a component seize during a surgery the results would likely be life ending. However, a  human doctor with a hand cramp simply pauses, stretches out his hand for a moment and continues with the procedure.  Situations like these make machine labor unsuitable in scenarios where failure is unacceptable.  If there were a profession who's duties included "push this button once an hour or everyone on earth dies" we'd likely build a machine to push the button, but you can also be sure at least one human would be there as well to make sure it was done.  Professions made safe by this point are: many healthcare professionals, large construction project jobs, pilots, and any profession which puts lives or large amounts of money at risk if done incorrectly.

Machines do not exceed their designed purpose.  Essentially this amounts to machines are not very creative.  They do not invent, innovate, create art or entertainment.  Any profession which relies primarily on the imagination of the human mind is for the moment unlikely to be taken by a robot.  This includes artists, actors advertisers, academics, inventors, engineers, programmers, design, everything hollywood, and essentially any scientific pursuit.

It's important to note a key difference between machine automation that replaces a job and machine automation that assists a job.  Arguably on a plane the autopilot is a machine that does most of the work.  However, it is unlikely that we'll eliminate the profession of pilot any time soon.  Similarly, many surgeons are using machines which reduce the invasiveness of surgeries and speed recovery times.  But machines which conduct major surgeries without human input are probably not just around the corner.

Undoubtedly there are a number of professions not explicitly mentioned which are equally safe from the automation revolution.  However, much of manufacturing has already been supplanted by robotic workers and the trend continues.  The next wave is likely to be the elimination of many service industry jobs as cheap and easy to use interfaces replace cashiers and restaurant workers.  If half your job involves handing someone a receipt be on the look out for R2-D2 coming to take your job.  Though as dire as that sounds, keep in mind even the self check out lanes at grocery stores still have a real person there making sure things go smoothly.

That's it for this week. Until next time stay safe and rationale. 

No comments:

Post a Comment