How Robots Help with Prosaic as Well as Complex Tasks
Advances in technology are disrupting virtually every segment of everyday life. While large language manipulation and reusable rockets grab headlines, it’s in more mundane tasks that robotics is more likely to make people’s lives easier and more productive. Robotics can get rid of some of the world’s worst jobs and boost productivity tremendously.
For example, Andrew “AJ” Meyer, whom I interviewed recently, has founded a company called Pickle Robot Company that builds robots specifically designed to load and unload trucks. Have you ever watched workers on a loading dock? Although they have forklifts and some other labor-saving devices, most of what they do hasn’t changed since the 15th century, when maritime trade became a major industry. It’s backbreaking work, it pays poorly, and workers wear out or get injured quickly. If ever there was an activity begging to be disrupted, this is it.
Pickle is just one of many examples of robotics, and other advanced technologies, changing the way the work is done in old-line industries that make up an essential, possibly dominant, part of the supply chain. I focus on this company because I was able to get Meyer to spend quality, unrushed time with me explaining his product, the underlying technology, and how it is changing the nature of a particular kind of work, but this story can be told in practically every industry and in every country. Because of the concurrent high-speed processing, connectivity, and artificial intelligence revolutions, old ways of producing goods and services are yielding to new ones at a pace practically unprecedented in human history.
Who should do repetitive backbreaking work, man or machine?
Pickle’s flagship robot, the Dill (get it?), mimics the movements of an efficient, well-trained longshoreman. Its muscles do not tire, it does not get hurt on the job and sue the employer, it does not go on strike or demand raises, and it doesn’t drink too much and show up late the next day. It can load as well as unload. It perfectly matches the popular image of a robot. All that’s missing is a pair of googly eyes.
Meyer is the leader of a small group of MIT grads, all fairly young, who founded Pickle in 2018. Even in today’s superheated atmosphere, to go from a concept to a working product in three years, when the first Pickle working model was built, is fast. What is the special sauce? As with many innovations, it’s the ability to repurpose ideas from existing technologies. (Glider + motor = airplane; computer + telephone = internet.) In a fascinating passage from the interview, Meyer explains how, ironically, the effort to build self-driving cars led indirectly to Pickle robots:
If you looked at Pickle from a certain angle, specifically its engineering diagrams and software architecture, you might not know that it wasn’t a self-driving car. It has almost all the same algorithms and the same types of sensors. Instead of a steering wheel and other car parts, it has the motors that comprise the arm. Instead of navigating a map, it’s trying to pick up boxes – but mathematically these things are very similar.
I couldn’t have guessed that. It’s a significant insight.
Why now, when the need has existed for decades?
The need to lift heavy objects and move them around robotically has been obvious for a long time. Why is this need being fulfilled now (by Pickle and its older and better-known competitors, Honeywell and Boston Dynamics), instead of decades ago?
While the basic design lived in people’s heads and appeared in old science-fiction movies, the intellectual ecosystem needed to build the robots wasn’t there. The knowledge we have now that we didn’t have then includes:
neural networks, which are collections of algorithms that mimic the design of a human brain to recognize relationships among vast quantities of data,
a technique called back-propagation for training the networks, and
improved algorithms for constrained optimization (a concept familiar to investors who use optimizers).
The material ecosystem was also lacking: we didn’t have good enough sensors or the graphics processing units (GPUs) that make neural-network computing fast enough to be practical. We got better sensors from the smartphone industry and GPUs from video games.
This set of handicaps, and the manner in which they were overcome, reminds me of an influential 2019 blog post by progress theorist Jason Crawford, “Why did we wait so long for the bicycle?” Crawford began by asking,
The bicycle, as we know it today, was not invented until the late 1800s. Yet it was a simple mechanical invention. It would seem to require no brilliant inventive insight, and certainly no scientific background. Why, then, wasn’t it invented much earlier?
One answer lies in materials science. We didn’t have the right kinds of metals, or a supply of rubber, or ways of making bicycle chains. Once these resources became easily procured, we got bicycles. Likewise, once we had neural-network computing and the other resources I just described, we got usable longshoreman-robots. In fact, we got several competing designs. Honeywell uses a vehicle-sized robot that grabs and sorts large numbers of boxes at once. Boston Dynamics’ Stretch warehousing robot is closer in concept to Pickle’s robot.
But Boston Dynamics seems interested in building general-purpose technologies such as the Atlas robot that famously runs, jumps, negotiates complex parkour courses, and otherwise mimics human mobility – seemingly just for fun. Applications will abound, perhaps first in medicine, but we don’t know what those will be yet. Pickle’s robots are focused on one purpose: warehousing.
“Human beings will always have a place in this”
I asked Meyer what happens when judgment is required. For example, suppose Dill encounters a messy truck with cargo all over the place. Can it organize the cargo? Will it ask for help? Or will it just go haywire like a deranged robot in the movies?
In another scenario, suppose the robot makes a mistake. Will it “know” that it has made a mistake? Will it correct the problem?
It will ask for help, Meyer explained:
It can resolve many of the mistakes that it makes on its own, but not all of them. And when there’s a set of mistakes that it can’t resolve on its own, its only choice is to rely on a human being to come resolve it. Human beings will always have a place in this.
Reimagining the worst jobs in the world
Meyer observed that, of all the jobs in a warehouse, loading and unloading boxes is the worst: body-destroying, soul-destroying, injury-producing. Those are the jobs that robotics should target first, because that’s where the greatest human benefit is. But warehouse robots will not eliminate warehouse workers. The workers will be able to leverage their effort using the robots. Meyer notes:
I've heard CEOs say that the only humans who will be replaced by robots are the ones who don’t leverage the robotics to increase their productivity. Instead of a person being able to unload 600 or 700 packages an hour out of a truck, that same person can now unload 6,000 or 7,000 packages an hour because they’re supervising 10 robots.
The possible (and obvious) downside: fewer human beings will be needed in that warehouse. It may not be a 90% reduction in the workforce because increased productivity will lower prices and stimulate demand – but the workers who have been replaced may not benefit immediately, or ever. A sanguine answer that might come out of a business school is that being put out of work by a robot frees the worker to do something more productive; after all, he too can be trained to operate robots at another business. Even if he is second best at being a robot-supervisor (a reasonable supposition for the worker who is let go), his life will be easier and better paid than if he had to lift boxes for the rest of his pain-wracked and exhausting life.
In many different aspects of the industrial sector, advanced technology is eliminating this backbreaking work and leveraging the productivity of workers. Interestingly, the most dangerous industry for workers is commercial fishing. Other hazardous industries include farming, forestry, mining, and oil drilling. I look forward to robots and other high-tech innovations making these activities safer, easier, and more profitable in the near future.
Source of data if not referenced otherwise: Quent Capital Research, MSCI BARRA, Bloomberg.