This is a further contemplation of the nature of work in the future of near complete automation of the workforce. What jobs will people do when all jobs are automated? Previously, I wrote the following:
Many other jobs, I believe, never will be automated. Even if all jobs can be automated, many jobs are likely be more cost effective for humans to do. There will be jobs where humans will be cheaper than machines and they will get the job done more reliably and often more quickly. In particular, I am thinking of the dirty, dangerous, and heavy jobs at the lowest level.
Automation is very successful in eliminating the least dangerous and least strenuous jobs. We originally hoped that robotics would take jobs that put humans at risk of injury or reduced life expectancy. Certainly, robotics has made progress in this goal in many industrial tasks that involve highly predictable tasks in relatively stable environments. It is in the areas of less predictable environments and more constraining environments that robotics has not been able to replace humans, leaving humans to do this dangerous work, often with low compensation for the risks taken. In my opinion, dirty, dangerous, heavy jobs in unpredictable environments will always require human workers, that robots will never be able to do this work. As automation succeeds in replacing less dangerous jobs, such as office work and even IT or STEM-related jobs, the only remaining jobs for people will be the dirtiest, most exhausting, and most dangerous work.
Perhaps, we will get an economy that can support a universal basic income (or as I propose, a universal basic expense-account that prevents wealth accumulation). Even with that ideal, we will still need some people to do the work that machines will never be able to do. There will be people doing these jobs.
Recently, I watched a BBC documentary on saturation divers used in the North Sea oil fields. Saturation divers are people who are skilled at staying alive for weeks at a time in order to perform 12-hour shifts of manual labor in the depths of the sea.
There was a part of the show the described the advancement of robotics to do these tasks in the future, but that those robots are not yet as capable of men working in same environment. Perhaps there should be a documentary about the development of these competing robots, and the difficulties the robotics have in replicating human capabilities working at depths. Maybe there is one I haven’t seen yet, but I suspect it is not as interesting as the one about saturation divers simply because saturation divers are successful in performing their jobs while robots so far have been unsuccessful.
My hunch is that there will always be a need for saturation divers to do deep-sea manual skilled labor. Automation may be able to do some tasks, but it will never be able to do all the tasks being asked of the saturation divers. Humans are quicker to improvise to handle unexpected conditions. Also, there is a high confidence of returning a dive team after a successful multi-week dive consisting of daily 12-hour shifts of work: these same divers return for new dives several times a year.
I have an engineering background and I still have the can-do attitude that we can eventually build robotics to replace humans in this type of task. It’s been about 35 years since getting my engineering, yet I vividly recall the optimism I had back then that one of the high priorities for robotics was to eliminate the need for sending men into dangerous environments like that of the saturation divers. Even with reports that robotic devices are not yet as capable as men, my reflexive reaction is that the technology must be close, though. Then I reminded myself that I had the same optimism 30 years ago: the goal was as close then as it is now, or it was just as far away.
The BBC documentary series included other extreme work conditions at great heights, on live high-voltage transmission lines, etc. I didn’t watch those, but they seemed even more obviously challenging for robotics. Ascending a transmission tower and doing work on high power lines just seems far more difficult to automate than having something stand on the sea floor and perform tasks involving lifting, carrying, assembling or disassembling. Yet, we can’t yet build robots that will obviate the need to send men down to do that work.
On a personal experience, I have hired contractors to replace my roof or to do tree work: both involved men climbing into precarious conditions and doing tasks the required quick adjustments to actual conditions in order to get the job down without injury or damaging property (especially when it comes to trees near houses). Each time I see this type of work, my first instinct is that this is 2018: why aren’t these jobs automated yet? Then I think of what it will take to automate it.
I imagine a robot that that can take a tree down in an urban setting, requiring cutting the tree in sections starting from the top of the tree and working down to the trunk. We do have equipment that can clear trees in open settings, but in an urban setting there is no room for these equipment to operate. An automated tree worker would require some kind of human-sized robot that can ascend the tree on its own power, fasten itself at a stable location, and carefully wrapping ropes around the limb to cut so that it can fall in a controlled fashion, and then cutting it with the proper anticipation to brace itself when the limb begins to fall. I visualize various alternative designs that can do this. It seems feasible until I consider what happens when something goes wrong.
These are dangerous tasks, something is bound to go wrong whether it is human or machine doing the work. There is a lot of tree work in this area, and frequently enough I hear of news of someone getting killed or seriously injured. This risk is sufficient incentive to find a robotic alternative, but what happens if the robot gets in the same predicament? There would need to be another robot to deal with the new situation, or (more likely) we’ll have to use men to fix the problem or extract the robot.
I imagine the scenario of a human-sized robotic for roofing failing. The failure modes may be that it may become immobilized, stuck on the roof, or it may start making a mistake such as not nailing shingles correctly. Similarly a tree-working robot may similarly become immobilized, or it may misjudge the condition of a cut and start a different cut that will make the situation worse.
Both humans and robots may encounter the same scenarios of making mistakes or becoming injured.
In the case of mistakes, a quick human verbal command can correct the behavior or provide instructions for undoing or mitigating the mistakes. In the case of injury, we can send humans (trained in emergency rescue) to retrieve the injured without hesitation because we’ve already committed to having humans in dangerous conditions.
In contrast, if it were a robot getting into this condition, we’ve committed to keeping men out of similarly dangerous conditions. We would then have no choice but to employ a different robot to perform the mitigating activity (or somehow reconfigure the existing robot to do this). The retrieval of the disabled robot will require a different robot, probably bigger in order handle the more difficult task of extracting the disabled robot without causing property damage such as what would happen by simply causing the disabled robot to fall to the ground.
Generally, the robot for fixing a problem is a different robot from the one doing the original job. I assume the risk of failure is low. As a result, the primary job robot will be optimized for the job, it will not have the additional weight for doing tasks unlikely to occur during the duties of the job. Thus a new duty such as recovering from a failure will most likely require a different robot, one that may take a long time to deliver to the job site.
In contrast, with men on the job, if there is a problem, the same people doing the work can be quickly tasked to repair a failure or to at least stabilize the situation until suitable experts (such as emergency rescue) can arrive and generally they can arrive quickly because they are nearby.
I predict a future when all the available jobs will be difficult and dangerous jobs. Many of these jobs may be short jobs such as local jobs like home improvement, tree-work, building construction or renovation, etc. In a more automated environment, I suspect these simpler jobs will gradually disappear with urban planning and renovations for easier automation of such tasks. The hard jobs will be related to legacy structures. As these are replaced with newer more automation-friendly structures, these associated jobs will disappear.
A different story will exist with remote and isolated jobs, similar to the jobs described in the BBC series with saturation diving. These jobs exist today in many areas such as mining or environmental clean-up (such as at Fukushima nuclear meltdown site). Similar jobs will always exist and may even increase in number.
A common trope of automation apologists is that the future will produce jobs to replace the jobs automated. One example is the recent abundance of job opportunities in Information Technology (IT) work with jobs like computer programming and systems design. From a modern perspective, these jobs are preferable to the more tedious paper-based tasks that automation has replaced. The analogy is that the future will introduce similarly attractive alternative jobs.
From my perspective watching recent automation advances, I don’t share that optimism. There will not be an explosion of more intellectual jobs of our fanciful imaginations. The current automation revolution is automation of intelligence itself. For the various intellectual jobs we have today, getting machines to fill new job opportunities will be much faster than preparing humans. Older humans are very difficult to retrain. Younger humans have to go through a multi-decade education process to make them highly eligible to take on the job openings during their youth but long-since filled by automation.
Perhaps there will be a few human openings for the innovation, but I don’t think they will be anywhere near as numerous as present-day STEM or IT-related jobs. Even with modern automation of intellectual tasks, a single person can create and run an entire company that fills a niche in the economy.
In the future, I anticipate there will be extremely high unemployment of the high-intelligence type of workers. There will be a few exceptions, but the current numbers of people who work in these intellectually challenging jobs will not be sustained much longer. Maybe we will achieve a sustainable economy that supports a reasonable universal basic income, but I have difficulty believing that such a solution will satisfy the people who work in intellectually challenging fields. Nor will artificial environments such as virtual gaming be satisfying for long. They will want employment, but they will not be able to compete with modern automation for intelligence-prerequisite jobs. They will have to satisfy themselves with hobbies, competing with peers in economically irrelevant pursuits such as chess tournaments, or writing inconsequential blogs.
Yes, I get that there will be occasional spot opportunities like we’re currently experiencing with niches like machine-learning, but I expect these to be short-lived opportunities and they will employ a tiny fraction of who are currently employed in these mental-labor type jobs.
Meanwhile, there will remain challenging jobs like the one described above about saturation diving and their surface-based support teams. I think these jobs will perpetually evade automation. These jobs take a toll on their workers resulting in early retirement ages, early dismissal due to injury, or loss of life. As a result, there will be a steady stream of opportunities for replacements, although the eligible candidates will need to be young and very healthy, similar to the requirements for infantry-type military service.
Besides the remoteness and danger of these jobs, another characteristic of these jobs is the saturation aspect. Most of these jobs will not involve industrial 40 hour work-week standard consisting of 5 days of 8 hours each. These jobs will involve long tours of duty spanning many weeks or perhaps many months or years with 6-7 day workweeks and up to 12 hour days. The defining nature of these jobs is that while employed, the person basically is always employed or at least confined to a narrow environment next to a job-site. There may be some relaxation periods, but the typical recreational opportunities will only be available in the intervals between job-assignments.
A cycle of saturation-employment and saturation-recreational-periods will be the norm, but the relative times of each may vary far beyond what most people will find comfortable. Some people may experience far more unemployed times then employed times, while others will experience the opposite. In particular, the actual balance will be outside of the control of the worker. It will depend on the job and the physical limitations of human physiology or mental limitations of psychology.
To repeat an earlier point, this vision of the future assumes some form of economy that can support some form of universal coverage of basic life expenses. I optimistically assume that the work is optional from the perspective of assuring a reasonable living standard. At the same time, I pessimistically assume that the available work will generally not improve the living standard over not working at all. Certainly, the living standard during the saturation assignment likely will be lower than not working at all.
Despite that, I expect that there will be a significant proportion of the population that will seek out these opportunities, and most will devote their best efforts in excelling at the task.
A counter-argument may be that of experience of history of communist or socialist regimes (the most recent example being played out currently in Venezuela) where something similar to a universal benefit system inevitable devolves into a culture of poverty. I suspect something like that will happen with a UBI scheme: the defining level of acceptable living standards will gradually decline. At the very least, once we have full automation, I don’t see much chance that the standard of living will increase beyond that the initial conditions.
These cautionary examples of past failures of socialism type governments may be relevant for the majority of the population. However, I think even within these degraded economies there remains a sub-population that is gainfully employed in specific dirty jobs of sanitation, policing, military service, heavy manufacturing or mining, etc. From a macro-economic standard, perhaps these jobs are under-performing in terms of driving economic growth, but the jobs themselves remain, and the most essential jobs are filled with people who dedicate their energy into their jobs.
I predict a post-scarcity economy with a UBI will have frequent civil unrest of people who are bored or unsatisfied with declining living standards. I also predict that that same economy will employ a select few people who are very well suited for specific essential manual jobs that inherently resist any attempt at automation. The declining UBI living standard may in fact encourage people to devote the majority of their life-goals on doing the best possible job at these tasks. Given the option of inescapable declined living standards, some people will prefer to find reward in a job accomplishment (or merely surviving the experience) as a form of compensation.
I suspect these populations existed in the oft-cited failures of socialist type governments. I see evidence of this in the most extreme example of North Korea that is able to present a credible military threat backed by technology despite its relatively low economic conditions. That threat comes from motivated workers in saturation-type jobs even if an oppressive government is required to keep the majority of population from objecting to their economic conditions.
My point is that humans are generally interested in being productive in some capacity. Many may find a way to satisfy themselves with a culture of poverty resulting in long periods of idleness, but inevitably (I believe) there will always be a small minority within the population who will volunteer for saturation-type jobs even if they come with risk or at least requirements of long periods of work-specific isolation.
In this blog, I described an alternative form of government based on data and urgency, a government I label as a dedomenocracy. This has some parallels with past socialist or communist ideals, but I proposed a constrained model of a default condition of libertarianism or anarchy punctuated by authoritarianism limited to the most urgent issues. From the perspective of gathering the optimum data for automated algorithmic governance, the ideal is to have full liberty to collect data about people’s natural behavior in the current reality. The ideal is to leave people alone so that we can gather good data. When urgent conditions do require intervention, that intervention may be severe and strict in order to resolve the issue as quickly as possible in order to release this rule back to the default of liberty.
This model of government is consistent with the trends of automation that make automation of government a possibility or even an inevitability. I discuss the concept not as an ideal to strive for, but instead as an inevitable end-state to prepare for. Due to the near-complete automation of the entire economy, the liberty will be a liberty of leisure instead of a liberty for pursuing new enterprises.
Within this model, there will remain some jobs that will always resist automation and they will be similar to the jobs described above. People will fill those jobs, and for the most part, their work will be known only to themselves and their supporting teams. The broader society may never recognize these contributions, let along celebrate them. Yet, the jobs will be filled and people filling them will do their assignments to their best ability, often with successful achievement of their objectives.
While I see most IT or STEM related fields eventually being fully automated with artificial intelligence (such as machine learning), I see some jobs within the field falling into the saturation employment category that resists automation. We will eventually automate most of the innovative work of IT or STEM. Most of the STEM or IT related jobs today will eventually disappear, probably very soon.
One area that may remain is what I describe as dedomenology that I distinguished from data science. Data science strives to improve the various technologies that improve our data capabilities in the areas of acquisition, storage, retrieval, analysis, presentation, and decision-making. In contrast, I describe as dedomenology the scrutiny of the individual datum at each stage from the original observation, to the successive refinement, and the final consequence. Dedomenology scrutinizes the data and the processes that transform that data for potential flaws such as data corruption (including omission), data misinterpretation (not measuring what we want to measure), or algorithms having biases that negate trustworthy observations.
In my experience, this type of work requires long devoted time that defies a standard work hour schedule. This is not physically dangerous work, although it may be physiologically harmful from the stress involved in confronting a vast amount of data from a variety of sources and processing-stages in order to address some urgent complaint. In any case this kind of work has a saturation aspect, requiring very long periods of work stretching over many days regardless of the concepts of standard working hours such as a 40 hour workweek. When something needs to be tackled, it will employ the dedomenology continuously until there is some level of completion. In the event there is nothing that needs to be tackled, the dedomenologist may be unemployed. I don’t see the the latter being very likely. Perhaps a different person will have the task for the next assignment, but there will be an endless stream of assignments that someone will need to dive into the depths of the data ocean and staying there for a long time until the assignment is over.