Ever faster processors, cheaper sensors, abundant open-source code, ubiquitous connectivity, and the advent of 3D printing are some of the forces behind the recent proliferation of robots. As I see things, these forces will only get stronger, and as more robots become part of our lives—in homes, offices, factories, hospitals, and many other places—we'll inevitably face challenges involving our adoption and use of robots.
Some observers are voicing their fears about a decline in human-human interaction, while others warn of an irreversible and senseless loss of jobs, with robots taking over tasks that, they argue, should not be performed by machines (such as caring for the elderly). Trade-offs will certainly be part of our growing reliance on robotics and automation. And it will be up to us to manage these trade-offs, just as we have with other technologies such as electricity, the automobile, aviation, nuclear power, computers, and the Internet.
As a VC looking for investment opportunities in robotics, I talk to lots of different people about their views on the future of this industry. Many times what I hear from these people are totally contradictory, so I often have to come up with my own conclusions. Below I present a list of what I consider are five pressing issues concerning robotics—and I identify each as a myth or a fact. My hope is that they can provoke some thought and debate among engineers, policy makers, consumers, and investors. Let me know what you think.
Robots are intended to eliminate jobs: MYTH
Almost every major manufacturing and logistics company I’ve spoken to looks to robotics as a means to improve the efficiency of its operations and the quality of life of its existing workers. So human workers continue to be a key part of the business when it comes to robotics. In fact, workers should view robots as how skilled craftsmen view their precision tools: enhancing output while creating greater job satisfaction. Tesla Motors is just one example of using robots [pictured above] to do all the limb-threatening and back-breaking tasks while workers oversee their operation and ensure the quality of their output. At Tesla's assembly lines, robots glue, rivet, and weld parts together, under the watchful eye of humans. These workers can pride themselves with being part of a new era in manufacturing where robots help to reinvent and reinvigorate existing industries like the automotive sector.
Manufacturing and logistics must adopt robots to survive: FACT
Although total cost of ownership is a popular yardstick used for purchasing capital equipment, payback time is more commonly used for automating basic (typically arduous) worker tasks. If we use pick-and-place as an example, one estimate says that each work "cell" (or equivalent of one worker) costs $32.5k/year (not including benefits) for a single shift in the United States. Historically, manufacturing and logistics companies have adopted automation equipment with one-year payback. Therefore, a $65k installed robot, operating on two shifts, is highly desirable. However, the net present value of down time, upgrades, and maintenance must be calculated into the $65k number. Labor costs are lower in Asia, where each work "cell" is closer to $20k, though that number is expected to increase sharply. It's not surprising, then, that Asian manufacturers, including Flextronics and Foxconn, are actively seeking automation technology to increase productivity of their existing workforce and meet increasing demand. And we are already seeing a similar trend in the logistics sector. Deregulation and technology brought about the notion of the third-party logistics provider, also known as the 3PL, which offers transportation, warehousing, and pick-and-place to suppliers in a wide variety of product categories and marketplaces. 3PLs compete on their ability to deliver higher-quality services as lower cost to their clients. With technology being responsible for the 3PL's coming into existence, they continue to spend vast sums on automation in each of their hundreds of warehouses to improve the quality and variety of their logistics offerings—in other words, robotics will allow them to remain competitive and survive.
Autonomous robots are still too slow: FACT
If you were watching the DARPA Robotics Challenge last year, you would be quick to notice that the robots are awkward and slow at completing even the most basic tasks. Although Moore's Law has sped up capabilities like machine vision, traditional search-based algorithms for navigating through decision trees are just too slow in runtime. As the famous video of PR2 folding towels showed, tasks that we take for granted as humans are time-intensive for algorithms—even with the help of the most advanced sensors—to navigate. Significant advances can be made with fundamental algorithmic improvements that take the "brute force" out of machine vision and other tasks. In particular, lots of research are going into pattern recognition, in addition to handing control over to the "cloud" when machine vision algorithms come short.
Robots are too expensive: MYTH
Modern household appliances are examples of dedicated pieces of hardware that people buy without thinking twice. These devices benefit from decades of incremental improvement, and millions of units in the field over which development and tooling costs are spread out. The same could apply to robots. The problem, though, is that robots still require specialized—and costly—hardware. In particular, actuators are among the most expensive parts in almost every robot, and unlike processors and sensors like cameras, the cost of actuators is not coming down at a significant pace. Consider, for example, Willow Garage spinout Industrial Perception. Before being acquired by Google, the startup managed to get close to human speed in identifying and unloading assorted boxes from a container. But although its robotic system used inexpensive sensors like a few cameras and Microsoft Kinect devices, it still required a relatively expensive robot arm powered by conventional actuators. The good news is I believe we're about to see a lot of innovation in actuation systems—and that will finally bring the cost of robots down, as it happened to appliances. One promising area involves the special sensor-equipped joints that allow robots to control their motions in a precise and safe way. Groups such as CMU-spinout IAM Robotics, Redwood Robotics (acquired by Google), and Modbot are using unique approaches to reduce the number and simplify the motors, gears, and sensors needed, hence dramatically reducing the cost of the arms, which in most cases dominate the cost of the entire robot.
Robots are difficult to use: FACT
Rethink Robotics’ Baxter is one example of a robot designed to be affordable and simple to program. However, those qualities come at the expense of the speed and precision associated with traditional industrial robots. Most autonomous robots require highly-trained workers and painstaking programming, calibrating, and testing. These requirements are often unacceptable for businesses that expect capital equipment to be 100 percent operational within days, if not hours of shipment, without hiccups. Although efforts like the Robot Operating System (ROS) and Open Source Computer Vision (OpenCV) are trying to simplify tasks and requirements to have robots up and running and doing useful things, they're still mostly used by experienced roboticists with PhD degrees. Would the personal computer have been as popular as it is today if they booted to a command prompt? They probably would amongst developers, but not among the masses that drove the PC, and later the Internet, revolutions. Standard CAD/CAM/Gerber/toolpath generators made automated machine tools a no-brainer. We need the equivalent for robotics.
Shahin Farshchi is a partner at Lux Capital where he invests in hardware and robotics companies. He is based in Palo Alto, Calif.
Shahin Farshchi, an IEEE Member, is a partner at Lux Capital, where he invests in transportation, robotics, AI, and space startups. Follow him on Twitter: @farshchi.
