Physical products have increased in availability and decreased in real prices for decades. This has been achieved through a two-step process: the application of technology, and the offshoring of manufacturing to counties with low labor costs, in particular China. As the world deals with the impacts of the COVID-19 pandemic, it has become apparent that offshoring manufacturing brings with it a consequence of making our supply chains more fragile. This post will explore how and why manufacturing was offshored in the first place, and how new technology can change this trend through local smart automation.
We rely on products made in factories for almost every aspect of our daily lives. With a growing global population, and rising quality of life for most people on the planet, the total amount of value generated by manufacturing has exploded in the past 50 years, from a base of $2.8 trillion in 1970 to almost $15 trillion today. The growth is only accelerating, with an increase of 2.5x since the turn of the millennium.
As the total volume of products we make and use has grown, so has the way we make them. The United States of America was once the world’s largest manufacturer, with many of the innovations we rely on for high quality products developed here. The first moving assembly line began operation on October 7, 1913 at the Highland Park Ford Plant, and allowed the Model T to be produced in just 93 minutes - faster than the paint of the day could dry.
While manufacturing in America has continued to grow, it is the growth of manufacturing in China that has led the world since the turn of the millennium. From a start of less than half the US’s output in 2000, China’s manufacturing growth allowed it to overtake the US in 2010, and today stands at more than 70% higher.
Seen another way, from an equal footing in 2010, the total manufacturing output of China compared to the US looks like this:
So why has this happened? There are many individual reasons, supported by various governmental policies. But the underlying trend was simply the movement of manufacturing to China from high income countries starting around the year 2000, with a resulting increase in the percentage of China’s population employed in making products for the rest of the world.
Automation has been the counter-balancing technology trend to the movement of manufacturing to lower cost labor. Both the capabilities of machines and robots, and the amount of capital deployed into them, have increased over the past decade. The total number of industrial robots in service has more than doubled in the past decade.
While the total investment into automation has increased, it is far from evenly distributed across countries. Both the total number of robots, and the capabilities of these machines, varies widely.
The capabilities of automation installed today are also not evenly distributed. For example, the total number of typically more advanced cobots - collaborative robots, designed to perform tasks alongside humans - is still a fraction of the total number of traditional industrial robots.
Looking at the details of applying automation to manufacturing from a finalised product design sheds some light here. Far from being a streamlined and simple process, it typically involves multiple parties, contractors, technologies and methodologies.
A typical automation project involves multiple iterations of each of these steps, especially if starting from a design that is not completely mature. The final step - line setup - requires specialised programmer knowledge, with weeks or months spent at the (overseas) production line itself. The programming required is very deterministic, with the goal of performing the same physical actions repeatedly and with high accuracy. Even with the advancement of software tools to do this, it remains a laborious task and prone to error.
Successful projects can take four months to implement, and more typically take six months to a year. Unsurprisingly therefore, automation projects tend to be reserved for high volume production, where the time and costs involved can be amortized over a very large number of units.
The advances in automation hardware has not yet been matched by software. Since a line typically takes months to set up, a small change in product design or operating parameters of the equipment requires time consuming debugging and reprogramming.
New innovations in artificial intelligence have allowed dramatic improvements in areas such as self driving cars, digital assistants, and online shopping. For example, Google’s word accuracy for speech recognition recently crossed the threshold for human accuracy with advancements in machine learning.
Similar improvements are now achievable in the physical world. For example, OpenAI recently demonstrated a robot hand solving a Rubik’s cube, a problem that was simply not feasible to solve without AI.
Key to these increasing capabilities are advancements in simulation environments, reinforcement learning, and domain randomization. The final point is a particularly exciting development, and means that vast training data sets can be generated by altering the input parameters into simulated environments. We will explore this in future posts.
The next few years will see these technologies applied to manufacturing, bringing benefits to product designers and consumers alike. Instead of a long and disconnected supply chain, designers will be able to get instant feedback, transparent pricing, and faster turnaround times than are possible today. Most significantly, more and more products will be able to be manufactured locally, bringing economic, societal and environmental benefits.
Launchpad is developing the software, hardware and AI tools to dramatically speed up the manufacturing of low and medium volume products, made here in America. We’re still early in our development, and will be launching a private beta early next year. Please get in touch with us at firstname.lastname@example.org if we can help you, or sign up to our newsletter to stay up-to-date with our developments.