Dirty, Dull, and Dangerous — Why Automation is Not The Answer to Everything

By Lucas Whipple, Omni Venture Labs

Over the past decade, we’ve seen the advent of automation being applied to virtually every industry in some form. Some industries have been completely transformed by automation, but others haven’t seen the promised benefit.?I suspect the main reason robotics and automation have seen uneven results is that they have been applied indiscriminately to industries regardless of fit, and often don't focus on areas where automation is truly a differentiator.?

Often, the successful implementation of a robotic system boils down to what kind of environment the task at hand is in. For robots, handling completely unstructured environments is at best very hard, and at worst a fool’s errand. People, in contrast, deal with wildly changing unstructured environments every day without a passing thought. Things that robots tend to be very good at are generally things that people are not good at, and vice versa. Many excellent examples can be seen in this compilation of robots failing at tasks in the DARPA challenge 2015 that most people would find fairly straightforward, such as walking or opening doors.

Amazon has invested billions into automation, but they can’t seem to replace human pickers in their warehouses because even though the factory is heavily indexed and databased, humans are currently better and faster at picking and placing arbitrary objects. Robots are getting more and more integrated, but tasks like these — even though they sound simple — hide deep underlying complexity and pattern recognition tasks that are currently more suited to humans than to robots.

Self-driving cars are another fantastic example of humans thriving in unstructured environments where robots often struggle - Billions of dollars and some of the smartest engineers and researchers on the planet have just recently reached driving parity with the average US 16-year old with a learner's permit.

The above examples, while entertaining, are not intended to besmirch or denigrate the incredible achievements that have been made in robotics - especially humanoid robotics - in the recent past. It is meant to show clearly the kinds of environments that humans thrive in, but robots struggle to tackle.

Now that we have established the relative strengths and weaknesses of robots and humans, a more interesting question can be asked - what SHOULD robots do? A popular line of thinking is that efforts in robotic development should focus on tasks that fall into the 3 Ds: — dirty, dull, and/or dangerous.?

Opportunities

There are already numerous tasks that robots have been applied to that fall within the 3 Ds. Foundries, semiconductor manufacturing, cargo handling, and food handling - in these industries and many more, automation has improved or eliminated some very dirty and dangerous jobs. A great example is in meat processing facilities like this SCOTT Automation facility in Australia. Traditionally, meatpacking has been one of the more dangerous jobs for humans — not to mention dirty and potentially dehumanizing - that few people would elect to do given other opportunities. The company featured in this video developed specialized robots to make processing lambs almost entirely human-free, resulting in far fewer workplace injuries as well as improved yields and profit.?

Robots also work really well in other high-volume factory applications like this BMW paint shop, which limits human exposure to hazardous fumes and heavy machinery. Not to mention, painting car after car on an assembly line with the exact same color can get pretty dull.

Foundries can be dangerous and dirty places to work, and Kuka Robotics has long been working on robots that can work in these harsh environments and reduce the exposure of human workers to molten steel.

Another great application for robots is in cleanrooms - spaces that must maintain low concentrations of airborne particulates. Humans are constantly shedding skin cells, hair, and sebum, and the process of protecting against that every time someone needs to enter a cleanroom is a dull and repetitive task. Just look at how much effort FUJIFILM Dimatix has to go through to avoid introducing pollutants into their cleanroom. Robots by contrast have much less risk of contributing contaminants to an environment. Additionally, most of these cleanroom processes would be impossible to do well manually, furthering the case for robots in these applications.

The Human Element

It is worth noting very few of the examples above require significant interpretation or subjectivity on the part of the robots - the outputs of the processes can be quantified, and success can be objectively determined. The processes are generally repetitive, and do not generally require use of the incredible creative faculties humans bring to the table. I would argue that these tasks share another common element - they are dehumanizing. I believe one of the elements that should be considered when applying automation is the physical and mental toll a particular job would take were a person to do it instead.

Let us consider two approaches to the same problem - in this case, two injection molding factories.

Observe the Lego factory in Denmark. Here, robots make huge quantities of the same type of part - notice how people are heavily involved in maintenance and quality assurance work, but they never get involved with handling materials or processing sprues where they can contaminate things or injure themselves. This also allows the injection molding machines that Lego uses to operate 24/7 and yields an extremely efficient operation. This is the perfect marriage of robots and humans, each focusing on the tasks to which their unique strengths are best suited.

Contrast the jobs people are doing in the lego factory to this plastic molding factory in China where we see humans reaching into massive pressing machines, handling raw materials, and doing mundane, repetitive work in extremely hot environments. There are likely safety and product quality issues here. Given these concerns wouldn't it be best to automate many of these jobs?

This line of inquiry may seem straightforward and obvious, but beware - here be dragons. Philosophical and ethical discussions of automation inevitably become tangled in much thornier questions of poverty, labor exploitation, classism, human dignity, as well as ennui towards the market forces and economic systems that create conditions under which these topics must be discussed. A heuristic to determine if a given task "should" be automated would be a valuable first step in addressing some of these thornier concerns. Conveniently, we have spent the last several paragraphs doing just that.

Conclusions and Key Takeaways

• When considering automation, one of the key questions that should be asked is if the task to be automated is dirty, dangerous, or dull.
• For at least the near future, humans will be able to qualitatively assess, manage, and handle high-mix processes better than robots.
• The best (and most lucrative) applications for robots likely continue to be in industrial settings where highly repeatable, boring, and hazardous jobs are the norm.