Discover more from Gad’s Newsletter
Your Holiday Gift May Be Picked by a Robot
When asked who they think picks and packs their orders at Amazon’s warehouses, my students’ common answer is: robots. So most of them are very surprised to learn that it’s actually humans (to a large extent).
Apparently, this is about to change.
Amazon just unveiled several new robots, most of them designed to do better than their predecessors in tasks such as sorting and moving packages in the warehouse (or across warehouses).
But the main surprise was Sparrow:
“... Sparrow’s robot arm dove into a bin full of random merchandise and plucked out specific items using a ‘hand’ made of small suction cups. It could identify and select merchandise buried below other items, adjusting its grip to handle different objects before depositing them in the appropriate sorting bin. Amazon says Sparrow can identify about 65% of the company’s product inventory and can tell if an item is damaged and discard it. It gets better as it learns.”
“Amazon’s newest robot, Sparrow, represents a huge leap forward in automation, the company says. Unlike those other robots — which can sort just a few dozen sizes and types of packages — Sparrow can recognize, select, and handle millions of individual products using computer vision and artificial intelligence.”
This has the potential to be a watershed moment in automation.
You may be wondering what’s changed now and why it’s taken so long to create robots that can do what humans do.
Over the last 20 years, automation has been lagging behind expectations. In fact, in some countries, the UK for example, we’ve seen reverse automation:
“The U.K. has become an exemplar of reverse automation. Automatic car wash sites plummeted up to 2018 as unregulated, untaxed, and often forced labour flooded the hand car wash market. The number of automated car washes halved from 9,000 in 2000 to less than 4,200 in 2015. The number of dedicated handwashing sites expanded to at least 20,000. This was not part of some broader global trend. Automatic car washes remain a growth industry.”
So in that sense, it's actually nice to see the US being one of the leaders in automation.
Similarly, it seems Amazon is heading in the right direction.
Why is Adoption Lagging?
Although I’ve written in detail about the reasons automation’s been lagging, let’s recap: First, robots were too expensive, so the return on investment was extremely uncertain. Second, robots were not actually very good at doing what they were supposed to (or more precisely, they were very good at what they were supposed to do, but humans didn’t always know exactly what they wanted them to do). So it’s “easy” to define and build a robot that stamps metal at an automotive plant, but it’s much harder when it comes to assembling iPhones, a task which requires making many adjustments and corrections as you assemble the parts.
This was the main reason behind the lag in adopting picking robots. For example, until now, Amazon has adopted many robots in its warehouses, but they were all primarily for sorting and moving packages. The actual picking was done by people.
Picking items from a box requires a robot with very good dexterity, strong precision, and knowledge of how much strength is necessary to hold an item. Just to illustrate how hard it is for a robot to do simple tasks that require this type of precision, take a look at this video showing the most advanced laundry-folding robot. I highly recommend devoting a minute of your time on this!
“Using machine vision, a neural network called BiManual Manipulation Network (BiMaMa-Net), and a pair of industrial robot arms, SpeedFolding can fold 30–40 randomly positioned garments per hour, usually finishing each within two minutes.”
So the most advanced technology in computer vision and robotics can barely beat an average human in a task we have engaged in since Ancient Rome (when this was primarily a man’s job).
But we finally have robots that can do, with some precision, what humans can do with extreme precision.
As I’ve mentioned before, historically, pandemics have been a boon for automation. A recent paper shows that during the last two decades, every pandemic has driven firms to invest in automation.
“The results show that pandemics lead to an increase in robot adoption over time, with some lag. In the second year, we estimate that about 0.35 more new robots are installed per 1000 employees and 0.7 more new robots in four years after a pandemic event.”
Because the pressure to rethink different aspects of a business usually allows firms to think more long term, and because wages have been increasing —as wages go up, the economics of robots starts to make more sense.
Amazon warehouse employees make $18 per hour, on average (approx. $3,000 per month or $36K per year). Once we account for benefits, using a common multiplier of 1.4, the cost of a worker reaches $50K (and that’s about to increase as Amazon announced they will raise the entry-level salary to $19 per hour or more). I assume that a robot costs as much as other, similar robots, around $50K (since most of the innovation is based on the software, rather than the hardware). Clearly, a robot needs an operator, it needs to be maintained, and have parts changed, but if salaries are going to increase and the cost of robots is about to decrease, it starts to make sense.
If we try the same math using a more top-down approach:
“Warehouse costs represent nearly $90B per year at Amazon today. Of this, I estimate about $20B is cost of warehouse employees (based on some fuzzy wage and employee estimates).”
You can start to see how every small improvement in productivity that comes from replacing humans with robots brings huge dividends, primarily as Amazon’s employees are becoming less productive:
But that brings the final and maybe main reason that, in my opinion, Amazon has some urgency in installing and deploying these robots.
A recently leaked memo at Amazon reported:
“‘If we continue business as usual, Amazon will deplete the available labor supply in the US network by 2024,’ the research, which hasn’t previously been reported, says.”
Amazon has been hiring extensively, as we’ve documented before:
But Amazon has also been relatively myopic in how they’ve addressed their workers. As previously discussed:
“David Niekerk, a former Amazon vice president who built the warehouse human resources operations, said that some problems stemmed from ideas the company had developed when it was much smaller. Mr. Bezos did not want an entrenched workforce, calling it ‘a march to mediocrity,’ Mr. Niekerk recalled, and saw low-skilled jobs as relatively short-term.”
Amazon is still hiring warehouse employees (even though it recently announced layoffs among their white-collar workers), and claims that these robots are going to create more jobs. But there’s no question that the new type of picking robots can have a massive impact.
The broader economic impact of this innovation cannot be underestimated:
“There are over 2,025,059 warehouse employees currently employed in the United States. The average age of an employed warehouse employee is 39 years old.”
If we look at data from the BLS, this is one of the fastest-growing areas of employment between 2011 and 2021, with an annual compound rate of 3.6%. So technology could make a significant dent, and over time, eliminate one of the most important jobs for people who only hold a high school degree (or less).
But don’t worry, the warehouse may still need employees who will now need to operate among robots. Amazon has thought about them as well:
“Amazon in 2016 was granted a patent for a system to transport humans, enclosed in a cage and on top of a robotic trolley, for work in areas full of other automated robots. (U.S. Patent and Trademark Office)”
… I really have nothing more to say…
Thanks for reading Gad’s Newsletter! Subscribe for free to receive new posts and support my work.