Automation failures requires burst-mode human labor

My last post rambled around the notion of a different labor model emerging in the modern world of automation.   Automation is particularly good at replacing human jobs that involve regular and routine activities even if those activities require specialized training and intelligence.   Automation can and will replace any human activity that can be described as routine actions.   This is occurring even for those activities requiring intelligence to find patterns in diverse data sources as well as experience from training or actual practice.   Automation will eventually replace nearly every job that can be described as steady work with a reliable income.   In short, automation makes income an anachronism.

Steady employment and income is a recent development in human history, replacing a far more common situation of steady idleness punctuated occasionally by generally short duration work opportunities to solve some immediate problem.   An example is the burst of labor needed for harvest that must be completed within a narrow window of time when the crops are ripe but not spoiled and where the work is permitted by sunlight and weather.   This one of time of the year would be the period with the peak employment for a population where the rest of the year the same population will largely be idle.   During the idleness, there is no employment, no income.   People will be surviving on what they earned during the harvest.   That earning was often in the form of social acceptance and reputation to merit access to resources to get through the rest of the year.   This was a non-income based economy.

The examples of burst labor from historic periods will not return.   Agriculture for example has been largely automated, taking into account the seasonal demands, and preparing the logistics to perform the actions at the exact times needed.   Given the predictability of the agricultural demands, the automation could also include preventive maintenance as well as automated repair maintenance for all but the most unexpected failures.

There will be a return to a period where most people will adapt to an economy of only burst labor requirements.   For the most part, these burst requirements will come in some form of emergency.   The emergency can be some disaster causing very wide spread damage that renders automation irrelevant to help.    That emergency may be some natural disaster (earthquake, major storms or flooding, etc).    In addition, I expect to see emergencies created by automation itself.

I am hypothesizing a perfection of automation that is fully compatible with the best possible actions given access to the most complete and accurate measurements.   The automation would work far more reliably than human operators to make the best possible choice given the observations, and take into account not only short terms objectives but long term strategy to avoid problems.

Even with this idealized automation, I’m confident that automation will create unexpected problems or disasters where the recovery will be entirely left to human labor.    In these scenarios the workers will be suddenly thrust into a very intense demand for their labor (whether physical or mental).   Given the near total reliance on automation on supplying resources, the recovery would need to be completed in the shortest possible period of time.   People will need to work very hard, with very high stress, and with very little time for rest and no time for leisure.

An easily recognizable example is the immediate aftermath of an earthquake with people trapped in collapsed structures in an area cut off from electricity and fuel supplies.   The number of people saved will depend on very intense efforts of people with little to no help from machinery and especially from automation.

Future disasters from automation will be less visibly apparent but equally urgent.   In a widely automated economy, a failure of that automation can easily place large populations at risk of losing their lives unless humans quickly act to restore the automation’s delivery of the essential benefits.

The recent grounding of the Boeing 737 Max line of aircraft in an example of a failure of automation that requires rapid recovery efforts by humans.

I am assuming that the cause of the grounding is related to MCAS automation to augment the new flight characteristics of the new engines.   This theory (not yet proven) is that the automation itself placed the aircraft into an unrecoverable situation before the pilots could stop it.   If this is true, then the automation itself caused crashes, and will continue to cause crashes in the remaining aircraft unless this problem is solve.

Another aspect of automation is that new plane itself (along with the flight controls) relied heavily on computer simulations of extensive data about aircraft characteristics and environmental conditions learned from decades of experience in actual flights.   This type of automation failed to identify the flaws leading to the grounding of the aircraft model.  Those flaws may be in the equipment itself (including the MCAS) or it may have been in the setting of training requirements to prepare pilots for the new characteristics of the unusual engine thrust relative to the wing and fuselage.    Computer based engineering and simulation testing were essential to deliver this aircraft to the market.

This is an emergency because of the urgency to return this plane back into service.  I acknowledge the arguments that the plane may have been prematurely placed into service due to greed.    However, from the perspective of modern needs for better fuel efficiency, there was an urgency to get a plane like this into service.   While fuel savings can translate to greed in the sense of greater profits for airlines, this aircraft answered the urgent need for reduced carbon footprint of air travel.

There is substantial support behind the idea that global warming is a clear and immediate threat to humanity within the next decade or so.   Assuming this time window to solve global warming is true, we need to push into service precisely this type of technology that reduces carbon emissions.   If we are going to solve global warming in time to prevent global disaster, we need to make many more bold innovations that may bring even larger risks of locally catastrophic failures.

As we adapt to greater automation to support the human population without destroying the habitability of the planet, we will need to act so quickly that it is inevitable that we will see more catastrophes in the future.   Given the larger global catastrophes either of environmental collapse or more likely economic collapse, we do not have the option of abandoning whatever effort led to some local catastrophe.  Nor do we have the option of slowing down the innovation to bring new solutions into practice.   Instead we have no choice but to quickly find the cause of the newly introduced failure mode and then quickly fix it with a combination of improved automation and improved training of operators such as a separate pilot training specifically for the 737 Max that may be inherently more difficult to pilot compared to less fuel-efficient models.

There is also an economic urgency.   The entire industry has already adapted to the economics offered by these more cost-effective planes.   A prolonged grounding of the aircraft, or worse a permanent disqualification of the aircraft would lead to large scale failures that will propagate throughout the entire global economy.   Entire airlines and aircraft companies face potential failure from the loss of this particular source of efficiency.   Also risking failure are the many other businesses that have adapted their planning around more cost effective access to certain markets.

For this discussion, I am assuming that there is an urgency to get this plane back into service, while also restoring confidence that the recent events will never recur.   The particular example of the 737 Max merely illustrates the modern automation model’s need for labor that is quite different from the industrial model of labor.

An economy of automated engineering and automated operation presents a new burst requirement on human labor.    We need a labor model that can rapidly dedicate people to solve emergency problems caused by automation and yet that automation is essential for continued operation of the economy or for avoiding some future disaster.

In the 737 Max grounding example, we are told that the grounding may last a long time.   I have no doubt that there is a lot of work to be done, and there is no denying the clock time required to get do this work.    Instead my concern is that the constraints of the employment/income labor force are the root cause for stretching out the schedule to complete the necessary changes.   The work ahead needs to be performed by existing employees re-assigned from their other duties to work instead of solving this problem that is an emergency for their industry.   The work to be done within the manufacturers and the federal agencies will be preformed as a regular income-based work load.   To the workers, their workdays will remain roughly the same workstation, only their duties (and the stress) will change.

Having no direct experience in this problem, I can only imagine that there may be some increase in overtime hours especially among the specialists.   What is really needed is a pulse of augmented staff who can accelerate the work.    Such a pulse of relevant workforce is probably impossible in the current environment.   At best, they may be able to recall some retirees or suspend any pending retirements.

What is really needed is a massive training program to rapidly prepare new specialists to tackle this problem.  Those being trained may come from other occupations and will need to return to those occupations once the urgency passes.   In the current income-based employment model, the trainees will need to at least have a continuation of their income (if not some added incentive bonus) during the training, and they will need some guarantee that that they can return to their old jobs when the current work is complete at some indefinite time in the future.

The income-based employment model is inadequate to deliver burst burst labor of the specialized skills needed to address unexpected emergencies that will be increasingly common with advanced automation.    We will attempt to cope by tolerating the drawn out recovery time, but clearly we need a better model to shorten the recovery times.   We need a way to quickly staff a short term requirement with people who are quickly trained to do the work completely and yet expeditiously.   Meanwhile, we need to do this without taking reassigning people who are already productively employed on other priorities.

Eventually we will need a model to pull in otherwise idle labor, to impose very intense training on them, and then to subject them to strenuous work to complete the task in time to ward off catastrophic consequences of the absence of some essential automation.

The employment/income model for labor is incapable of providing the labor needed to address automation emergencies.   I imagine a more ancient model to reemerge where people will be conscripted to some new task, required to take very intense training followed by immediate work where both the training and work will consume most of their time, far more than the standard 40 hour work week.   This conscription will need to be selective in the terms of selecting those who have appropriate aptitudes to successfully perform the work.    It may be voluntary to some extent, but once signed on they will not have the option to leave as long as their skills are needed.   The duration of the effort may involve some compensation but it will not be income in the sense that they will spend the money immediately.   The compensation may be contingent on the success of the entire project, leaving no compensation at all for a failure to recover from the original emergency.   As a result, any compensation may occur after the completion of the job or even spread out in periodic rewards spread over a long period after the task is completed.

This is not an employee/income model for labor that we will need to recover from future emergencies, especially those resulting from automation failures.   The staffing essential to respond to emergencies are likely idle outside of the time of the emergency.   Those who are productive in other areas are not really eligible for emergency assignments or even if they were best qualified, their vacancies will need to be filled by temporary staff who will need to vacate the positions to accommodate the returning staff at the end of the emergency.   In either case, we need to employ people who are otherwise idle, subject them to intense and stressful conditions of rapid training and working at adequate levels to assure the successful completion of their roles.

The future automation economy relies on an idle pool of labor who have the capabilities to perform emergency training and duties, but they will be idle for long periods both before and after the emergency.    We need some means to sustain a pool of idle labor and keep them in good mental and physical conditions to act in emergency scenarios.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s