By Anne B. Miller
Editor’s Note: This article is excerpted from “The Digital Journey to Cognitive Manufacturing,” a paper by Anne B. Miller that was published in the proceedings of the 2017 Santa Fe Symposium on Jewelry Manufacturing Technology.
We are in the early days of a new era in computing that will profoundly change our world. Systems now learn. Unlike traditional computing that requires hard-coded programming rules, cognitive computing systems can process natural language and learn by experience—much like we do. These new systems can understand, reason, learn, and interact.
This is important because, for the first time in history, the volume of data we produce has outpaced our ability to make use of it. Eighty percent of the world’s data is dark—unstructured data that is invisible to systems. It’s buried in books, e-mails, journals, blogs, articles, tweets, videos, impulses from sensors, images, and sound. It is the type of data we encode in language and unstructured information. Cognitive capabilities can transform this data in new and meaningful ways.
Leaders of every industry are sprinting to become digital. But digital is the foundation, not the destination; the destination is cognitive. From self-correcting machines on the shop floor to Amazon Echoes in homes all across America, cognitive computing is changing the world as we know it.
Cognitive computing defines systems that learn at scale, reason with purpose, and interact with humans naturally. They apply human-like characteristics to convey and manipulate ideas. When combined with the inherent strengths of digital computing, they can help solve problems with higher accuracy, more resilience, and on a massive scale over large bodies of information.1 These systems are designed to adapt and make sense of the complexity and unpredictability of unstructured information.
Cognitive systems are not programmed; they learn. They interpret information, organize it, and offer explanations of what it means, along with the rationale for their conclusions. They do not offer definitive answers. They are designed to weigh information from multiple sources, reason, and then offer hypotheses for consideration.
There are four main characteristics to a cognitive system:
In the world of manufacturing, these advancements are so radical they are defined as the fourth industrial revolution. First was mechanization of work driven by water power and steam. Next came mass production achieved by division of labor and electrical energy. The third era was based on electronics and IT to further automate production; robots began to replace human workers. Now we enter Industry 4.0, which is defined by factories, machines, and parts capable of self-assessing, triggering actions, and exchanging information with each other, with the people with whom they manufacture, and with the people who maintain them.3
Cognitive manufacturing applies cognitive capabilities to digitize and optimize previously inaccessible areas of manufacturing processes. Systems have the capacity to monitor and evaluate manufacturing performance and then propose processes and operations improvements based on sensor and multifaceted data, optimized techniques, and advances in machine learning.4 This leads to “smart” equipment maintenance, factory operations, product design, quality, supply chain management, employee safety, and energy.
For example, consider a manufacturer that uses cognitive capabilities to reduce expenses associated with maintenance, parts and supply inventory, and production delays. They apply advanced modeling and custom analytics, real-time prediction of machinery health, and text analytics for enhanced asset health prediction. This has led to a 7 to 10 percent decrease in plant and equipment costs, reduced un-scheduled downtime, and improved parts and resource allocation via continuous updates of maintenance data.4
Modern manufacturing systems execute highly sophisticated IT-enabled operations and control infrastructure that tracks production metrics, quality metrics, and component status in real time. They exhibit the characteristics of “smart systems,” but decision-making processes and responses to the production alerts are performed by human operators based on their knowledge and reasoned judgment.5
Cognitive manufacturing is an evolutionary step in computer-enabled production system control. These systems perceive changes and know how to respond. They adapt the production to stay within target ranges of production cost, production rate, and sustainability indices such as energy and carbon footprint.5 They monitor and evaluate manufacturing performance and then propose improvements based on sensor and multifaceted data, optimization techniques, and advances in machine learning.6
Cognitive systems are fundamentally different from the traditional programmable systems, and there is often a learning curve as organizations figure this out and then determine how best to apply this new capability.5 What doesn’t change, however, is the need for good planning and a cognitive strategy. Early adopters in pioneering organizations describe three keys to success:7
Define the value. Not all business problems or processes are the best fit for cognitive solutions. Organizations should think through opportunities to find those that align with the unique capabilities provided by cognitive systems.
The unique cognitive business value should be defined upfront and should be mapped to the organization’s goals. A cognitive vision and roadmap should be developed and reviewed on a regular basis. The benefits of these systems evolve over time as they get smarter and drive more value. Cognitive transformation is a journey, with value increasing over time. Organizations should deploy solutions to a subset of trusted users who understand this before deploying to a larger group.7
Prepare the foundation. Cognitive solutions are trained, not programmed. This requires a commitment of time and resources from human experts to define the question/answer pairs for the system to learn and to supervise the learning. Organizations should assess their talent base and close any skill gaps. Besides the right expertise and technical skills, the team should be willing to experiment, learn, and explore along with the system.7
A quality body of knowledge, which may include structured and unstructured data from multiple databases and other data sources, is critical. Organizations should define the solution expectations and requirements and then define the “sufficient observation space” necessary to meet them. This may require increased partnering and changes to policy to get the data to support the decisions of the solution expectations and requirements.7
Understanding implications on existing dependent processes and policies is imperative. The way users interact with cognitive systems is entirely different from the way they interact with traditional systems. These systems may disrupt existing processes or fundamentally transform work so new policies may be required.7
Manage the change. Change management investment is often the first area of cuts in attempts to reduce IT system implementation costs. But these are not your typical systems, and so change management is even more essential.
Organizations that have begun their cognitive journey typically follow four major steps.7 The first is to chart the course for the cognitive journey. This includes identifying candidate-use cases across the organization and functional areas. These use cases should identify the computing solutions and processes to be disrupted. It should include the business case and key metrics to track value. A cognitive roadmap should be developed that includes a clear change-management strategy.
The second step is to experiment to validate the cognitive strategy. This step tests and validates cognitive-use cases through prototyping. It allows users to see what the end state could look like, and it tests the underlying business case hypothesis.
Next is to develop the solution and train the team. This is where the real work begins and investments are required. The focus is developing the solution around the priority-use case and training both the system and the users.
The fourth step is deploying the solution. But this is just a milestone celebration along the journey. Once the solution is deployed, even greater learning begins for the system and users. It is important that tracking business benefits and accuracy levels be an ongoing activity.
As companies begin to adopt cognitive manufacturing, they are applying the general keys to success and steps to cognitive described here, but tailored for the specific needs and opportunities of a manufacturing environment.4
Build a foundation of data. To enable cognitive manufacturing, equipment and assets need to be instrumented and outfitted with sensors to collect data. In addition, the team should gather already existing data. This data should be on processes and operations targeted for improvement. It should include data across systems—both structured and unstructured. The team should connect systems and sensors to bring in real-time data for more accurate insights.
Visualize the patterns. Next, they should quickly build up dashboards and use simple analytics to see and determine patterns. They should supplement this data with external sources of data and analyze variables that impact the process and operations targeted for improvement.
Advance to analytics. Now the team can apply purpose-driven analytics to gain new insights from the data. They should develop advanced models, process a combination of variables, and utilize the prediction engine to generate the best recommendations that drive the most business results.
Infuse with cognitive. When dealing with vast amounts of data, cognitive capabilities bring light and clarity. The team should start to bring in cognitive components, machine learning, and cognitive processing to act, resolve, and deliver better results.
The era of cognitive computing is here, and adoption is expected to be explosive. Cognitive systems augment our capacity to understand what is happening in the complex world around us.8 They are the only way to get value from all the volume, variety, and velocity of data—the great new natural resource of our time.8
It is incumbent for business leaders to learn how digital transformation leads to the cognitive journey and to take the first steps. Create a cognitive roadmap. Try free solutions. Experiment. As Yogi Berra said, “It’s hard to make predictions, especially about the future.9 The best course available is action—learn by doing and becoming a cognitive pioneer in your industry.
To read the complete paper, visit santafesymposium.org/papers.
1. Rob High, “The Era of Cognitive Systems: An Inside Look at IBM Watson and How It Works,” IBM Redbooks (2012).
2. Bob Vavra, “Cognitive computing delivers answers, asks new questions,” Control Engineering (February 12, 2016).
5. J. Rhett Mayor and Elizabeth Hoegeman, “Cognitive Manufacturing,” National Academy Press OpenBook; (abstract referenced).
6. Steven J. Skerlos, “Advancing Sustainable Manufacturing with the Use of Cognitive Agents,” The Bridge (National Academy of Engineering, Winter 2013)
7. Dr. Sandipan Sarker and Dave Zaharchuk, “Your cognitive future. How next-gen computing changes the way we live and work. Part II: Kick-starting your cognitive journey,” IBM Institute for Business Value (2015)
8. Ginni Rometty, “The Natural Side of AI,” The Wall Street Journal (October 18, 2016)