What are the fallacies of public understanding of robotics?

The robot in Tesla's production workshop is working hard.

Editor's note: Shahin Farshchi, a partner at Lux Capital, an investment company, said in an IEEE dispatch on a prestigious technology website how he knew, as an investor, how to understand the technology of robotics that was advancing by leaps and bounds.

Faster processors, cheaper sensors, lots of open source code, ubiquitous links, and 3D print breakthroughs all contribute to the recent advances in robotics. I think that these forces will only grow stronger, and as more and more robots become part of our lives, we must face new challenges: how we accept and use robots.

Some observers fear that "interaction between people" will disappear, while others warn that they will not be able to find jobs since robots do things that shouldn't be done by machines, such as caring for the elderly. We rely on robots and automated equipment for a price, and these costs are borne by us, just as we apply nuclear, aircraft, and computer technologies.

As an investor looking for business opportunities in robotics, I've talked to a lot of people about this topic. I often hear their contradictory statements, so I often have to make my own conclusions. Below I will list five questions about robots, and I classify them as "misunderstanding" or "fact". I hope these issues will generate some thinking and discussion among engineers, policymakers, consumers and investors.

Robots will reduce people's job opportunities: misunderstanding

I almost say to every production and logistics management company, the use of robots can improve operational efficiency and improve the quality of life of employees, people are still a key part of the company. In fact, workers should treat robots as a tool, just as skilled craftsmen think of their precision instruments, thus giving employees a higher sense of achievement and more output.

Tesla uses robots to do work that is difficult or dangerous for the human being, and lets workers oversee robotic operations and ensure product quality. On the Tesla assembly line, the worker supervises the robot to complete the process of bonding, riveting and welding. These workers should be proud because they are part of a new era, using robots to reform traditional manufacturing industries, such as the auto industry.

Production and logistics management industry must rely on robots to survive: the fact

While total cost of ownership (TCO) is a common standard for purchasing fixed equipment, in the field of automated machinery, the "payback period" is usually used as the standard. Take handling as an example, with artificial words to pay 32,500 dollars a year to complete a special process. By convention, production and logistics companies purchase automated equipment for a one-year payback period. So it's tempting to use a 65,000 dollar robot that can do two processes. Moreover, the net present value, the promotion and the maintenance expense also counted in this 65,000 dollars.

The labor force in Asia is cheaper, but prices are also rising sharply. So Asian manufacturers such as Foxconn are actively looking for automated technologies to boost workers ' productivity to meet growing demand for their products. In the logistics industry, we also see the same trend. The competitiveness of logistics enterprises is to provide customers with cheaper high-quality courier services, and automation machine is the core equipment of logistics enterprises. Robots will be used to ensure the competitiveness of logistics enterprises.

The robot is still too slow: the fact

(Editor's note: The current algorithm for the development of robots is still "brute force algorithm", that is, all the possible results, and then from these possible results to search out the necessary information to complete the task, this algorithm is too slow. A breakthrough over the years has been to try to add cloud computing to robotics, speeding up the ability of robots to respond. ）

If you look at the DARPA Robot Challenge last year, you will immediately know that these robots are clumsy even when they are doing some very basic task. Although Moore's law has speeded up the ability to improve computer vision, the traditional search-based algorithm is still too slow to guide the robot through its task with decision trees.

If you've seen the "Robot PR2 Towel" video, you'll see that the easy task for us is a robot that needs algorithms to do even if it installs the most advanced sensors. With the improvement of the basic algorithm, there will be an important breakthrough, no longer use the "brute force algorithm" to achieve computer vision or complete other tasks. In particular, the field of pattern recognition has been a lot of research, plus when the computer vision algorithm shortened, it can be referred to the "cloud" robot to control the operation.

Robots are too expensive: misunderstanding

The use of household appliances is a typical example, many people buy home appliances are not hesitate at all. These appliances have benefited from decades of technological advances and millions of of usable spare parts, which are fragmented in their development and processing costs. We can also use this way to reduce the cost of developing a robot.

The problem, though, is that the hardware needed to make robots is too professional and expensive. The actuator, in particular, is the most expensive part of all robots and, unlike the processors and sensors on the camera, the actuator costs are not significantly reduced.

For example, Industrial perception, which has already been acquired by Google, has begun to study how to improve the performance of robots, allowing robots to identify and unload various small boxes in a container and achieve the same speed as humans. Although its robotic system uses sensors that are not expensive, it is relatively expensive to use actuators to power the robotic arm.

The good news is that I believe we will see a series of new innovations in the braking system that will greatly reduce the cost of making robots, just like home appliances. Notable areas include the mechanical joints that install special sensors that allow robots to control their movements precisely and safely. For example, such as the Carnegie Mellon derivative companies such as IAM Robotics, Redwood Robotics (has been acquired by Google), and Modbot, and so on, they are developing unique ways to streamline the engine, gears and sensors, thus greatly reducing the cost of the manipulator, The cost of a robotic arm usually determines the cost of making the whole robot.

Robots are difficult to operate: facts

Most automated robots require well-trained workers to operate, and require careful planning, calibration, and testing. These requirements are usually difficult for some enterprises to accept, some enterprises need mechanical equipment in a few days without error to operate. Although applications such as ROS (robotic operating system) and OPENCV (open source computer vision) simplify tasks and requirements, allowing robots to function and do useful things, they still need to be operated by robotics experts with doctorates.

If you want to start your PC from a command prompt, will the computer be as popular as it is now? It is also expected to be popular among developers, rather than among the masses that trigger the PC Revolution (followed by the internet revolution). The standard cad/cam/gerber/toolpath makers will be able to make automated machine tools so that small white users can operate. We also need to operate the robot more simply.