The Future of Robots

When the first industrial robot, Unimate, appeared on a General Motors assembly line in 1961, it was a modern marvel. The job it performed, transporting die-castings and welding them onto car bodies was not only onerous, but dangerous to human workers, who faced both the risk of injury and being exposed to toxic fumes.

Over the last 50 years, robotic machinery has been vastly improved. Due to more sensitive motors and actuators, they’ve become incredibly precise, which enables them to work with small components, often with far more accuracy than a human can achieve. That’s what’s has allowed robots to move from making Buicks to smartphones.

Yet the future lies not in greater precision and accuracy, but the ability for robots to collaborate effectively with humans. Rethink Robotics is one of the companies at the forefront of this revolution, so I talked to Jim Lawton, the company’s Chief Product and Marketing Officer, to learn more about what we can expect the future of robotics to look like.

From Work-In-Progress to Just-In-Time

For a long time, manufacturing facilities were set up to produce in large batches. Once a product was engineered and designed, factories would churn out thousands of identical items without stopping. This was an effective way to produce goods cheaply, but it also resulted in a huge amount of work-in-progress inventory.

Yet today, the vast majority of manufacturers have switched to lean methods, where the idea is not to push out as much product as possible and then let it sit in a warehouse waiting for customer orders, but to produce products “just in time” to satisfy demand. So rather than trying to push supply down the sales channel, factories allow demand to pull it through.

The benefits to just-in-time methods are obvious. It saves millions in inventory costs and allows for better customization. Manufacturers no longer have to predict exactly how many different models and colors they need to make, but can adjust their production levels according to customer orders.

Still, that kind of flexibility is a problem for traditional robots. In order to work with precision, they need to be set up specifically to perform a particular task. If you have to stop production for half the day to retool and reprogram robots every time you need to make a slightly different variation of a product, the efficiency gains you are supposed to get from just-in-time disappear.

The Rise of the Collaborative Robots

Rethink Robotics was co-founded by robotics pioneer Rodney Brooks in 2008 with an initial investment coming from Amazon CEO Jeff Bezos, but remained in stealth mode until 2012. His vision was to create a profoundly different kind of robot, one that could safely work with humans and adapt to new tasks with ease.

The difference between the company’s robots—it has two models, Baxter and Sawyer— and the descendants of Unimate is not more advanced hardware, but intelligent software. They are designed for “unstructured” environments, meaning unlike conventional robots they know where they are in relation to other objects and can make adjustments when things shift.

For example, let’s say you wanted a robot to place items on a conveyer belt. With a traditional robot, you would have to make sure the items are precisely positioned. But Baxter or Sawyer can simply be shown what to do and will sense where the boxes are, get them and put them where they need to go.

If humans get in their way, they will avoid them or wait for them to move out of the path. If the conveyer belt has stopped and there is already an item on it, the robot will wait until it gets moving again before it puts another item on it. If the robot needs to perform another task it can be trained in minutes, not hours.

Another interesting aspect is that the robots’ displays are designed to communicate through expressions. They show smile if things are going well, frowns when they are not, raise an eyebrow when they are confused. Their eyes show where they plan to move so its coworkers can anticipate their motions. They’re more like co-workers than a factory machines.

Remaining Challenges

Today Rethink Robotics is a thriving enterprise. Lawton told me that while a few years ago it was getting orders in ones and tens, now its customers are buying hundreds at a time. He predicts that soon it will not be uncommon to receive orders for thousands of the company’s collaborative robots.

Yet still there are problems to be solved. First, although today’s robots are physically agile, which allows them to move around a factory floor safely, they still have trouble with tasks that require high levels of manual dexterity, such as threading a needle or unwrapping a package. Those are still jobs that are best performed by humans.

Another challenge is that today’s machine learning algorithms have limited ability to understand broader context. For example, if an algorithm is trained to recognize dogs, it can do that with a high level of accuracy. However, if it is shown a picture of a family playing frisbee with their dog on the beach, it will be clueless about anything else besides the dog.

Lawton believes that being able to do that type of unsupervised learning will be key to the future of robots. It’s not like in the early years with Unimate, where the greatest challenges were to develop robots that were more advanced mechanically. Today, the biggest barrier to advancement is being able to make robots that are more intelligent.

The Road Ahead

As machine learning algorithms improve and other advances, such as neuromorphic chips, enable smarter robots, we’ll be seeing them do jobs that only humans can do now, except in many cases, they will be able to do it better. They will be able to access vast databases, recognize people, machines and parts, be able to derive insights and act on them.

That means that robots will soon be able to not only do grunt work, but also diagnose problems and devise solutions. They’ll be able to do routine, highly repetitive work like quality assurance, inspecting components and finished products. What’s more, the cost of all that capability will likely come down dramatically as technology advances and the cost of components declines.

Clearly, this raises concerns about robots creating unemployment and income inequality, but Lawton hasn’t seen that effect come to pass. “In the customers we’ve sold them to, factory managers are generally using the robots to enable their human workers to do higher level tasks, rather than to replace workers,” he told me.

One thing is clear. Robots will become an increasingly important part of the economy and will drive productivity for decades to come. The challenge for humans is to improve on skills that robots will not be able to do for a long time, if ever, like asking insightful questions, perform non-routine tasks and, most importantly, work well with other humans.

Wait! Before you go…

Choose how you want the latest innovation content delivered to you:

• Daily — RSS Feed — Email — Twitter — Facebook — Linkedin Today
• Weekly — Email Newsletter — Free Magazine — Linkedin Group

Greg Satell is a popular speaker and consultant. His first book, Mapping Innovation: A Playbook for Navigating a Disruptive Age, is coming out in 2017. Follow his blog at Digital Tonto or on Twitter @Digital Tonto.