The New Connectivity of Industrialism

Matthew Putman, CEO Nanotronics

Connectivity is nearly always associated with the power of the internet, and the nearly instantaneous communication it has enabled. In recent years, this idea has expanded to communication with objects, with the now almost cliché term of “Internet of Things”.

Without any intention of diminishing the power of the information technology that has defined our era, there remains something that has been missing from valuable discourse on progress that is crucially needed for the world to see the pervasive abundance of a connected world. Most of this is a collision of discoveries and inventions that have only now been possible, so it is natural that they are not fully realized. While connectivity is generally thought of as accessibility to human interactions, the new universal connectivity will be between machines themselves, with humans playing the role of teacher, student and inventor. In the case of industry this will change the nature of manufacturing in more dramatic ways than we have seen since the second industrial revolution.

The modern technological connections that are known to us all, and even the future possibilities for a new paradigm, are very much due to products that have come out of the microelectronics industry. This is most apparent when looking at the increasing pace of performance in processors, and the adoption of energy efficient semiconductors for everything from LEDs to LIDAR for self-driving cars. Even recently these were not deemed financially viable, but have historically been more advanced than the software that utilizes them. This is something that we all feel intuitively when we boot up a computer and the operating system is essentially the same in 2017 as it was in 1995, even though the processors that run them have gone from single core processors that run at 33MHz to multi core processors that run at 5000MHz. That operating system, and many of our most ubiquitous programs such as Word, Power Point, and Excel are practically the same. With that almost unthinkably dramatic improvement those programs should be unrecognizably superior. There was at least one exception to that, which has been a catalyst for the change that we are, and will be seeing in a world that is driven more by the intelligence in the machine, rather than the machine itself. That is in game optimized AI. Since IBM’s Deep Blue beat the world chess champion Gary Kasporov in 1997, there has been a feeling that software could be different, but that difference has only recently become something that is even recognized in the daily lives of much of the planet. The new disconnect is not as much believing in the power of software, but understanding where the capabilities exist that are not being utilized and why those are the key to the future of education, accessibility and connectivity.

A factory engineer, assembler, inspector or IT professional would more than likely mention customers before they mention their own jobs if asked about why 2017 is more connected and advanced than it was in the past. An IPhone, an Amazon Echo, a Netflix recommendation and nearly everything we hold, or speak to, other than a human, have captivated and influenced us all. At the same time humans in the factories that make those technologies that enable those products to exist may not see the 8 hours of the day they spend at work using any of the advances that they admire and count on for modern entertainment, and sometimes even survival. This could be because of how invention has migrated to the computer in specific ways over the last 2 generations. The role of a computer scientist has been so elevated as the path to stability, and even wealth, that children are viewed as successful and prodigious for the ability to use technology to code programs on a computer, while they are looked at as lazy and distracted for actually taking advantage of the most sophisticated connected digital tools we have, such as online gaming. This has made a lot of sense, but the future presents a very different take on both the factory staff and the child. Boolean logic presents limitations that do not exploit the power of the hardware it runs on, even with the perception of software dominance. However, just in the last few years artificial intelligence algorithms are being implemented throughout consumer tech for search, voice recognition, and recommendation engines. The factories that make up the enabling technologies have not relied as heavily on AI as their customers have. Connecting machines through AI will improve yields, eliminate scrap, lead to faster product iterations by using data classifiers to make the first major change to process control in 70 years. Making companies more productive is not the end goal, or certainly not the most exciting way to engage students or the public as whole, but two important by-products come out of it: a new way of working, and a new generation of products that make life increasing better.

Computer Science is the career of choice for many, but the methodology of manufacturing has been the largest segment of the developed world economy for over 120 years. A factory is still a factory in much the same way as it was 120 years ago though, with one major change in the ability to avoid waste and make consistent quality products faster. This happened far before the computer revolution, 70 years ago just after World War 2. Often a blank slate for progress is equally as important as direct iterations on existing technologies and methods. This was true in when the United States and Europe participated in the rebuilding of the nearly completely destroyed post war Japan. The efforts in organizing manufacturing did not come from the usual people who build factories such as business leaders and production engineers, but from a mathematician, W. Edwards Deming. Building on work done at Bell Labs almost 30 years before, Deming put a theoretical concept to practice called Statistical Process Control (SPC). SPC was revolutionary because it was an acknowledgment that the process of building anything was a series of production nodes, each of which had variations through time, even throughout the day. SPC was an acknowledgment that tracking these changes was a better path to creating uniform products if the information being tracked was actionable. The point was not to see if a part was failing as it passed a node, but that it was statistically the same as the previous time that it did. SPC slowly took away the heavy reliance and common wisdom that products should only be checked for quality, and replaced it with products that relied on consistency, which in turn produced better quality. There remained a problem with this, and that problem has persisted even as computers fill factories now. Being a statistical sampling of data leaves details that larger quantities of data possess. Despite its useful power data can be seen as the new pollution. Big data alone creates backlogs that do the opposite of providing rapid process control. In contrast, real time AI detection and classification has the potential of eliminating the vast majority of data that is not needed. Outside of a factory Google aims to find answers to the searches that are important to the person asking the question, not just a list of everything that contains the words in the search bar. For AI to succeed in a factory, and replace SPC as the core to the way manufacturing works, meaningful information alone needs to be extracted from the mountains of data available. This information is powerful because it assigns causality and eventually makes real-time decisions to optimize a process.

This new outlook can be called Artificial Intelligence Process Control (AIPC). It will be a self-contained ecosystem of communication between machines themselves that generate solutions by optimizing for the final product rather than optimizing for individual parts of a process. This is as much like the way a body works as a self-contained protein factory and regulator of metabolic processes to optimize for life.

Though there may appeal a leap between programming, industrial practices, and innovation as almost separate ideas, they are connected themselves, and it is truly the new connectivity. Work becomes about optimizing for better and more interesting products. Traditional factory control is replaced by intelligence that is now only seen in consumer tech, and the values and skills we learn and teach are transformed from one where we build through programming alone, to connecting machines to do the work that we once thought of as the predominant path to success. The connection we make will be those that build a future through building things rather than a future where the things being built are the only things that connect us.