Tesla is a cautionary tale of too much automation, too soon

When Tesla CEO Elon Musk announced plans to ramp up production of his company’s ambitious new Model 3 electric sedan last year, he warned that months of “manufacturing hell” lay ahead. But after nearly a year of ups and downs, the automaker finally hit a crucial target at the end of June: it made 5,000 Model 3s in one week.

The Model 3 is supposed to be more affordably priced than other Tesla models, once the company can make them in high enough volumes. And it’s believed that a mass-market electric vehicle could finally help Tesla turn a profit. But first the company has to prove to buyers and shareholders it can build them fast enough, more reliably, and in ever greater numbers.

Tesla’s solution? Not more automation — more humans. For all the talk of robots putting people out of work, not every workplace is ready to be wholly disrupted just yet. And while more automation is certainly coming, where you’ll find it today depends on the job.

“What [robots] are not very good at is anything that’s closer to what we can do with our hands,” says Angela Schoellig, an assistant professor at the University of Toronto who specializes in robotics.

She pointed to the fine manipulation and complex movements that typically elude most of the robots in use today.

“We clearly, clearly cannot replicate what the human hand can do in any way.”

Troubles with the fluff bot

Musk learned this the hard way. According to The New York Times, Tesla originally designed an assembly line for the Model 3 that boasted more than 1,000 robots and other assembly machines. Not all have worked as well as expected — to the point where the company built a whole new production line under a massive tent outside its factory — and now Tesla reportedly plans to hire about 400 employees a week to keep production on pace.

“There are some things that are very well suited to manual operation and some things that are very well suited to automated operation, and the two should not be confused,” Musk said during the company’s first-quarter earnings call, adding that “the vast majority” of Tesla’s production system is still automated. But, he admitted, “we did go too far on the automation front and automated some pretty silly things.”

Tesla CEO Elon Musk told shareholers in May that the company had gone ‘too far’ with automation and wound up automating some ‘pretty silly things.’ (Joseph White/Reuters)

By way of example, Musk described a robot designed to place noise-dampening fibreglass fluff on top of a battery pack. “Machines are not good at picking up pieces of fluff,” he said. “Human hands are way better at doing that.”

Part of the challenge, says Schoellig, is that the types of robots used in warehouses, factories and assembly lines today are generally pretty dumb. Most have to be manually programmed to do what you need them to do. And while some companies like Tesla have tried to imbue their robots with greater intelligence and awareness, more complexity only adds to the likelihood that the robot won’t always perform a task as precisely or accurately as you might expect said Schoellig — which is exactly what Tesla found.

A suction cup for hands

Beyond Tesla, much of the automation currently used in factories and warehouses is designed to replace more basic tasks. In a warehouse belonging to the U.K. grocery chain Ocado, more than 1,000 wheeled robots shuttle groceries around a pre-defined grid. In Amazon’s warehouses, squat-wheeled robots ferry goods between aisles, while humans do the packing.

Even in Tesla’s factory, robots previously programmed to bolt and wire car seats now merely move the seats into place. Humans finish the job.

“It’s still the case that if you want to build robots that are very reliable and robust and you don’t have to spend half of your time repairing them, you still try to make them as mechanically simple as possible,” said Schoellig.

She pointed to Amazon’s annual Picking Challenge, where teams compete to see who can build the robot most adept at identifying and sorting products (because, eventually, Amazon wants to automate that too). Last year’s winner? A robotic arm with a suction cup on the end. It’s a simple concept, about as far as you can get from a functioning human hand, but can still pick things up.

Of course, simply throwing more humans at the problem isn’t a long-term solution, either. Tesla employees interviewed by Reuters sounded wary that the company could keep up its current pace the way it currently operates. That’s to say nothing of ongoing concerns about health and safety that current and former employees have raised.

But Musk is convinced that they can go even further and hit 6,000 models per week in July. It’s still a far cry from last year’s plan to build 20,000 Model 3 sedans a month, but that’s where the robots can help — just maybe not doing as much as Tesla once thought for now.


Leave a Reply