Episode 8: Robots, Piers Full of Robots

As always, go listen to the the show! It really is better than reading the flat text, but for accessibility and archivability, here you go.

Worth noting up here only that if you’re in the Bay, Jonathan Hirsch and I are having a Containers party April 27 to celebrate completing the run. RSVP soon to come hang out with 200 of your closest new friends.

In the conclusion of this series, we peer into the future of human-robot combinations on the waterfront and in the rest of the supply chain. We’ll hear about the strange future of cyborg trucking and meet the friendly little helper bots in warehouses. The view of automation that sees only a battle between robots vs. humans is wrong. It’s humans all the way down.

This is the final episode of Containers, an audio documentary about global trade, technology, and the work that goes into moving stuff around the world. I’m your host, Alexis Madrigal.

We’ve come a long way together on this podcast. We started with the advent of containerization. From there, we traced this new system for handling cargo. We’ve learned how it has transformed everything, from how you get your toothbrush to the very structure of cities. We visited ships and considered the impacts of trade policy. Last week, we met longshoremen who lived through the automation of the dockwork they’d built careers doing.

And today, we’re going to look at the future of automation in all the different places where the goods of global trade get moved around: in ports, on ships, on trucks, in warehouses. What’s next?

The future is, more or less, robots. Way more robots! Way more sophisticated robots!

Are they gonna take all all the jobs? Will they become friendly whirring little work pets? How humans and robots are going to come together is the topic of this last episode.

Think about the simplest example: shipping. Rolls Royce, which is a leader in designing ship engines and navigation systems, have already announced a big push into self-steering, autonomous ships. But even if things just stayed on their present course without a radical change, the amount of human in the system has been dropping for decades. Bigger ships with smaller crews mean that the labor required to transport a ton of cargo across the ocean is a small fraction of what it would have been even 20 years ago. The biggest ships carry up to 18 times the cargo of a 1970 ship, and they have a smaller crew: sometimes there are only 15 people on board.

Whether or not we call this kind of thing “automation,” the human element, or the way a company might phrase it the “labor cost” of shipping keeps getting smaller and smaller.

On the shore side, things are much more complicated. There are now four shipping terminals that are at least partially automated in the the US: one each in Los Angeles, Long Beach, Norfolk, and Newark. At the facility in Los Angeles, both the operator, a company called TraPac, and the ILWU, the union, agree that the automation reduces longshore work by 40 to 50%.

TRAPAC VIDEO: This trailblazing automation project is the most technologically advanced container terminal in the United States.

For what it’s worth, I asked the local ILWU leadership many times to talk with me on the record about this stuff, but only managed to get several polite Nos. The terminal operators themselves are touchy about the topic, too. Both sides are itchy about tipping their hands for future contract negotiations.

But, luckily, we can piece together the basics of how the Los Angeles automated terminal works with some help from a port promotional video. A human operator brings over a box with a regular crane from the ship onto the shore.

Video: Then, fully automated shuttle carriers transport the container from the wharf to the waterside transfer area or the on-dock railyard…

These shuttles are not like Google’s self-driving cars, which have their own cameras and laser sensors to navigate the human world. Instead, the dock itself has magnets embedded in it, which simplify the task of positioning the automated transport carriers. There is a sort of map built right into the terminal that only the machines can use.

Video: From the waterside transfer area, the containers are collected by electrically powered, rail mounted, automated stacking cranes, known as ASCs.

These 39 huge machines drop the boxes into big stacks of containers, positioned perpendicular to the shore. Then, when a truck arrives in need of a particular box, the crane gets a box ready. The trucker backs up to the stack and then literally gets out and stands in a booth!

Video: The booth sends a signal to the waiting ASC letting it know it is now safe to deliver the container to the truck.

While this is new automated terminal is certainly as sophisticated as they come, the technology underpinning automation is not actually new. And that’s important for understanding how the hype around robots works. The first such terminal began installation in Rotterdam in 1990. Back then, port automation was seen as both inevitable and imminent.

As the rest of the world is going nuts for self-driving cars, automation of the nation’s ports has proceeded very slowly and does not seem either inevitable nor imminent right now. Why?

Short answer is: It costs so damn much. Each terminal that automates has to spend maybe $500 million dollars to implement the robotic system. And really, from a productivity perspective, there isn’t much of an improvement over using people. If you look at a list of the most productive ports in the world, it is dominated by the huge ports of Asia, which simply deploy lots of longshore labor.

If this has been the experience of the last 25 years, why even bother to try to eliminate the jobs that are left? Because after all, as we’ve already heard, so much automation already came with containerization. Most of the jobs that existed in 1965 are already gone.

One consequence of that situation, though, is that a relatively tiny number of longshoremen can shutdown the imports and exports along an entire coast. The docks are a chokepoint. Labor on the docks has a lot of leverage, which is one reason that the longshoremen make good money.

And that is where the robots come back in. For terminal owners, there is one huge advantage to machines: they don’t go on strike. In an unusually honest article in the industry bible, The Journal of Commerce, a port planner said that the reason to automate is that it “results in less exposure to interruption.” And by interruption, they mean labor action.

I mention all this detail because when we say the world is coming down to “humans vs. robots,” that elides the fact that it’s often humans vs. humans, using robots as a tool.

But it’s not just ports themselves that are automating. Similar and even more advanced technologies are coming to trucking and warehousing, with fascinating consequences and possibilities.

And when we come back, we’re gonna talk about the next big things in automation.

AD: Containers is brought to you by Flexport. Flexport is a freight forwarding company built around modern technology. They help over 2,500 companies run better global supply chains. Check them out at Flexport.com.

To understand what’s happening in port automation, we need to understand the broader context of, well, of robots, really. So we’re gonna dock the boats, go ashore, and look at the vehicles that take containers all around the country: the trucks. And the truckers who drive them.

Foster: There’s 1.7 million truckers in the US and if you look across all other driving jobs, taxis, driving, delivery, there’s another 1.7 million jobs.

This is Natalie Foster, a fellow with New America California and at the Aspen Institute, she focuses her studies on the future of work.

Foster: Now there are big technology companies on a massive arms race not only to figure out how you build and deploy, but now, how you actually move into regulating self-driving trucks. It’s a very scary proposition to imagine, overnight, 1.7 million drivers out of business.

Crazily, truck driver is the number one job in 29 states! And not just Idaho and Wyoming, but California, Texas, Pennsylvania. Big states. And why are there so many truck drivers in America now, relative to other professions? Well, the big losses in other careers have come from automation, jobs getting made obsolete, and globalization, jobs getting offshored. And up until now, truck driving jobs have been insulated from the changes that have affected factories or call centers or the ports, of course. In fact, truckers do well when there is more stuff moving around longer distances, so the big changes in logistics and trade have kept demand for truckers high. .

Foster: Critics will say that as technology has entered the workforce we’ve always figured out new ways to create jobs. The classic example at the turn of the century as we moved from the farms in to the factories. we couldn’t have imagined any other jobs besides the jobs that existed the farms. But we now know of course that farms are a very small percentage of the American work force.

And maybe! But that’s hard to just take on faith. trucking is the kind of decent job that exists in low-income, less urban, areas If truckers lose that work, it’s not immediately clear what else they might do.

Foster told me, though, that it’s probably not as simple as millions of jobs disappearing overnight. Trucking is a complex profession. And I wanted to learn more specifically about trucking and how trucking actually works as a job. Like, are the people in the cabs of trucks worried about self-driving vehicles?

Karen: In the work I’ve done with the industry and reading their publications and stuff. I think they are not as concerned about this happening in like 5 years or 10 years even as a lot of are. And part of that is because there is a lot more to truck driving than driving.

Karen Levy is an associate professor of information science at Cornell. She’s been studying how truckers work and the technologies that aid and surveil them for the better part of a decade. She broke it down like this: Sure, truckers drive but there is also all this other stuff they do!

Levy: If you’re driving a flatbed, you have to be able to secure the load. You don’t want stuff falling off the back. There is a lot of maintenance that has to happen. You need somebody who knows how to fix stuff. There are all kinds of weird contingencies on the road with weather, and all that. And some of that is addressable technologically, but some of that, you just need a person.

That’s made Levy skeptical that technology is simply going to put people out of work.

Levy: So it could mean that you reorganize your labor force in some way so that it can still do those parts but not the driving, but the idea that 3.5 million people are going to be suddenly out of work totally is, like, too simple. And I think people in the industry recognize that.

There are real problems with the current model of trucking. Cracks in the system that don’t result from an overabundance of enthusiasm by people simply excited about the gee-whiz possibility of self-driving cars.

Basically: being a trucker in the just-in-time logistics world we have today is really, really hard. To make enough money, drivers often have to stay on the road for too long, regardless of what their bodies want or need.

Levy: Fatigue is kind of a push factor. We need something to address the problem of basically human to use this job has become in many ways unsustainable, if we’re going to be realistic about how much work they are doing. The fact they are not paid enough to do what they are doing. All of those are push factors.

Another way to think about the automation of goods being delivered over the roads might be that the task of driving across highways is fairly easy to automate but the job of trucker is not. And that’s something that’s true across the board. Here’s Foster again:

Foster: 49% of tasks that we do now, not singular jobs, but tasks across all kinds of work could be automated using off-the-shelf technology. Which means a huge shift in what sorts of jobs, how they get packaged up and how they get hired.

So, let’s try to imagine how that shift could go.automation we’ve learned on this show is usually about a new system of work.

In the case of trucking, the point of the system of work would be to move goods over the highways on wheeled vehicles. And one of the most intriguing possibilities is what I’d call… the Land Port model. I first heard it described by Alex Davies, the transportation editor for Wired Magazine.

Basically, self-driving trucks go bombing along the freeways, then they pull off into a Land Port.

Davies: Truck drivers will become like tugboat pilots. Once the truck pulls off the highway, into a designated autonomous trucking zone, the human comes in, climbs into the cab and does the rest of the driving. And meanwhile, trucks on the highway don’t have anybody in them at all. It totally seems nuts, but economically it makes sense. The technology wouldn’t be that hard to implement…

So the easy-to-automate task of highway driving would be matched to a new set of in-city tasks. Every trucker would become like a drayage driver, moving things in and out of a port. And I’d be willing to bet that the jobs would be dispatched a lot more like Uber than anything else.

To extend the vision of this system of work a little, perhaps truck stops would become large-scale mechanic shops. Self-driving trucks would detect they had problems and pull in to be tended to by their humans.

You end up with this strange-sounding future in which fleets of robotic carriers call for on-demand help from a labor pool of humans who can deal with social or mechanical contingencies that are, for now, beyond the reach of robotic control.

Right now, in the nation’s warehouses, fascinating cyborg systems like this have already developed. There are tons of robots in warehouses working in concert with humans already. For example, Amazon snatched up a company called Kiva Systems back in 2012 for $775 million.

Video: With the Kiva Mobile Fulfillment system, the operator stands still. Products, cases, items or orders come to them. Pallets, cases, and orders are stored on inventory pods, which are picked up and moved by a fleet of mobile robotic drive units. The distribution center is now completely dynamic, self-organizing, and self-adapting.

The robots that already came.

The orange Kiva robots sit low to the ground. They drive up underneath what looks like a shelf, and screw themselves into it, then they drive it over to the human who is picking the stuff to box it up, or whatever, and return to warehouse floor. Over time, the little pods move oft-picked products to easy to access locations and stuff rare purchases out of the way.

Different members of the fleet go plug themselves in or request maintenance or what have you. And the humans all stand around the edges of the warehouse, using their hands to pick products. Because, as it turns out, our hands are very very difficult to mimic. They’re versatile and strong, but also gentle. So that’s nice.

But there are other possible ways of structuring warehouse work that use different technology. Kiva robots use a system of tags built into the building — kinda like the container carriers at that automated terminal in Los Angeles we talked about — but a startup company called Fetch Robotics works quite differently. Its little robots are like tiny self-driving cars with considerably greater decision-making power baked into them.

I decided to go visit Fetch Robotics down in San Jose. They’re located out on one of those tree-lined boulevards of squat office parks, which seem to dominate the southern stretches of Silicon Valley. The companies that surround their offices have names that were clearly created playing technology word madlibs: Silitronics, Amax Technologies, Electromax, Advance Probe.

We walk across a classic startup open floor plan, engineers tapping away, the sound of robotic arms filtering in from a workshop. As we sit in a conference room, I can spy a robot that looks like a really tall Roomba with a basket attached to it rambling around behind CEO Melonee Wise.

She says that warehouses need robots like hers, basically because not enough people want to work in them.

Wise: A lot of people don’t like to talk about it like this, the fact is that there are 600,000 jobs that are going unfilled in the United States and that gap is getting bigger and bigger. The turnover rate for any manufacturing or warehouse job is about 25%. And so, there is a need for automation because people aren’t showing up to do the work. That’s how we got interested in the space. Why we got interested in the space.

So what they focus on is helping humans get things from Point A to Point B within a warehouse or factory, so the companies can make more efficient use of the people they do have.

For any person within a warehouse or manufacturing facility, their work process breaks down into tasks and one major part of that work process, that one task, is walking from a pack area to a picking area. And that typically can be a 3–5 minute walk and they do it maybe 5 times an hour and then next thing you know, 25 minutes of their 60 minutes is walking.

So, the picking area is what you’d think of as a warehouse. Like, the racks of stuff. And these can be vast vast spaces in the millions of square feet. And the packing area is where stuff gets shipped out. So, in the most basic deployment of their system, they just put robots in the picking area and humans grab stuff off shelves and then tell the robot to deliver it to the packing area. And that alone lets people get 20–30% more work done per hour.

They call their core product, the little robot, Freight.

Melonie: Freight is a relatively small circularish robot. It’s about 14” tall and it is 22” wide or long, and 21” wide. And it is white and gray and looks friendly.


And what it is, basically, is a little tiny self-driving vehicle.

Melonie: Freight is an autonomous unit. So it has an on-board computer and it has all the sensors necessary to work in an environment. It has a 25 meter planar laser scanner, so it can see 25m out.. We also have a time of flight 3D camera, that allows us to see the volume in front of the robot.

So they take this little machine, which can look around and make decisions about what paths to take, and they drop one into a huge facility.

Melonie: The way our robot actually works is that when we bring a robot to facility, we make a map of that facility using its laser. And once it has that map, that map is distributed to all the other robots and then we can tell the robot where to go within that map. And it basically uses the features in the environment. Shelves, poles, anything you would see in warehouse, to reference where it is going and how to get there. And it does so in a collision free fashion.

They send the robots high-level tasks to over a wireless network and the robots coordinate how to do that work. A lot of what they do is following people around. And so they’ve had to train the robots to recognize human legs. They do this by feeding it lots of data, by which I mean images of legs.

Wise: We have a dataset of a lot of different people’s legs. Very thin legs to very big legs to legs with really reflective slacks to legs with jeggings, things like that. And all of that data is fed into something that gives us a leg classifer. This is a leg or this is not a leg.

The reason they need all that data is that, as with self-driving trucks or cars, there are always gonna be these unusual things that happen. On highways, maybe it is a person or debris where those things shouldn’t be. Or in an example that Google famously gave a couple years ago: one of their cars once encountered a person in a wheelchair chasing a wild turkey down the street. Their cars had never seen that exact situation before. Engineers call these circumstances edge cases. And they are why Google’s self-driving vehicles or Uber’s self-driving trucks are not autonomously driving the roads today [I should say, parenthetically here, that they are of course driving around on the roads, just that they always have human-minders aboard]. For most workaday situations, they perform great, but they need more training data to teach the cars as many edge cases as possible. And the same is true for Freight, as they found out taking the bot out to trade conferences.

Wise: The way our Robot docks with the autonomous charging dock is that it looks for the shape of a charge dock. And it is a relatively unique shape. Very unique. But when we’re at conference there are typically all these curtains everywhere for the different booths and things like that. And the dock was near a set of curtains and the robot started trying to dock with the curtains over and over and over again. And we were like, HUH. I think we need to fix that.

We drive over to their training warehouse and I watch as these robots simulate warehouse work. They pack freeweights on top of them to test their durability under various loads. The most fun thing is to try to get in their way because the robots try to find paths around you.

It is the mega-happy version of human-machine labor symbiosis. Friendly little helpers following our jeggings around, helping us work better. And I think a lot of automation will, in fact, work like this.

But like so many of the big topics that we’ve addressed on this show — trade, technology generally, automation narrowly — there will be slices of the population that will lose their jobs. Some companies will use the technologies to drive down human wages or break unions or make work conditions worse. This is just the reality of what happens when so much power is concentrated in companies and so little among workers.

It’s early January 2017. After a long 2016, it’s time for Port Director Chris Lytle’s annual address to the maritime industry of Oakland. The event is being held in Jack London Square, at Scott’s Seafood, the kind of classic place where they serve sourdough bread in a basket with little gold-wrapped pads of butter. People from all the terminals and major shipping lines and retailers and exporters and port commissioners are gathered around big round tables. There is the din of the tradeshow.

2016 was a banner year for the port. The steady and rarely excitable Lytle gave a rousing review of all the challenges that the Port had met. One of the terminal operators had filed for bankruptcy. There was a labor dispute hangover. The global market for shipping is miserable. There’s more pressure from east coast ports. And yet, still, improbably, Oakland had a good year.

Lytle: Guess what, the numbers are in, it’s an all time record for our cargo handled at the Port of Oakland.

Even Oakland’s mayor, Libby Schaaf, showed up for the annual event.

Schaaf: I just wanna say how happy I am to be here every time I get to be with the port family. As many of you know, I served as the public affairs director for this organization and I can’t say how much respect and fondness for all of the many people — and you’re all still here! It’s been 10 years! What’s up?

The Port is still there. They’re all still out there, despite all the other changes in Oakland and the Bay and the world. The ships and the sailors, the captains and the longshoremen, the Target shoppers and the warehouse workers. The truckers and the activists.

I remember the bus driver who took me out of the main terminal there on one of my last reporting days. Her name was Greta and her father had been a longshoremen, too. And I was telling her about this podcast and she said exactly what I’ve been thinking for the last 8 months and what is certainly the right sentence to end this series.

Greta: It’s a whole different world down here.

You’re right, Greta. And I love it.

That’s it for this episode and for this series. If you want to keep up with what’s going on with my reporting, I’m starting back up writing at The Atlantic magazine in May and working on a book about all the stuff we’ve been talking about in the podcast. Sign up for my newsletter, 5it, and you’ll get regular updates on everything.

A big thank you to the director of audio at Fusion, Mandana Mofidi, for all she’s done to get this series on the Internet.

Containers has been produced and edited by the sagacious Jonathan Hirsch. It’s crazy, but we have not yet met each other, even though I’m pretty sure he is my long lost brother.

If you’d like to witness our first meeting, come on down to the Containers Closing Party, Thursday April 27, at 7pm, at Oakland Hot Plate, 348 13th Street in downtown Oakland. There’s an Eventbrite invitation pinned to my Twitter page. Just RSVP. All listeners are welcome. Jonathan and I will do a little song and dance and then we’ll all hang out.

A whole series of thank yous to a lot of people, some of whom know they helped and many who do not. To the 99% Invisible crew, who inspired me to do audio. Thanks Roman, thanks Avery, thanks Sam. To Anna Sussman from Snap Judgment, who inspired me to make the podcast better. To the Journal of Commerce’s Bill Mongelluzzo, who clearly knows everything about west coast ports. To Rachel Slade for her reporting on the El Faro. To Lu Olkowski for her reporting on southern California ports. If you need more waterfront, go check out her podcast, Cargoland. To Liam O’Donoghue, for his podcast East Bay Yesterday, if you need more Oakland stories, he’s got you.To the Port of Oakland’s Chris Lytle, John Driscoll, Mike Zampa, Kyle Brunelle, and Marilyn Sandifur — thanks so much for sharing your work and time. To Brian Nelson, Michael Vawter, and Frank Silva for the generous donation of their knowledge of the waterfront.

Thanks, too, to Ryan Petersen at Flexport for making this crazy project financially viable.

Thank you to Gabby Miller and Ruth Gebreysus and Jamal Jellyfish and all the rest of the Oakland crew. To Robin Sloan, for being my sounding board and media inventor model. And finally, to Sarah Rich, my wife, for being the best first listener while also putting up with all the late nights and busted schedules.