September 14, 2023 AI in Manufacturing and Managing Risk

AI is a game changer for manufacturing companies, increasingly using it to streamline the way they do business and increase efficiency.

By Jade Davis

Let’s review how we got here and how we can use AI safely to harness its ability.

Historical Overview

The history of artificial intelligence dates back to our ancient past with hieroglyphs from ancient Egypt. Hieroglyphs are considered an early form of coding, which laid the foundation for modern AI. Today, natural language processing is a major subfield of AI, and it is used in everything from chatbots to virtual assistants. Egyptians also laid the foundation for AI (e.g., modern machine learning algorithms) through their use of math and geometry as skilled mathematicians to calculate area and volume. Today, modern mathematical models are used to analyze and make predictions based on data.

Ancient Egypt passed the totem pole to ancient Greek philosophers analyzing life and death, and inventors creating automatons (mechanical creations that moved independently of human intervention). The word automation was derived from ancient Greek and means “acting of one’s own will.” It was used to describe an automatic door opening and the automatic movement of heeled tripods.

This foundational work gave roots to the invention of the programmable digital computer, the Atanasoff Berry Computer (ABC) in 1942.

Other scholars trace the roots of AI back to 1942, when the American Science Fiction writer Isaac Asimov published his short story “Runaround.” The story is about a robot developed by engineers and revolves around the Three Laws of Robotics. Meanwhile, 3,000 miles away, English mathematician Alan Turing developed a code-breaking machine for the British government called the Bombe, to decipher the Enigma code used by the German Army in the Second World War. The Bombe, which was about 7 x 6 x 2 feet large and weighed about a ton, is generally considered another first working electro-mechanical computer. 

In the mid-1950s, AI research was formally founded by computer and cognitive scientist John McCarthy, who coined the term “artificial intelligence.”

From the 1950s forward, various scientists, programmers, logicians and theorists aided in solidifying the modern understanding of AI. More than 70 years later, we have witnessed innovations and historical advancements that have catapulted AI from being an unattainable fantasy fiction dream to a tangible reality.

Manufacturing Industry

Manufacturing companies have utilized machine learning of AI to advance the industry. The benefits of AI have produced significant results over the past few decades to prove naysayers wrong. For instance, it is common knowledge that AI optimizes operations and increases efficiency and productivity. Machine learning has proven its ability to predict machine failure and maintenance and manage supply chains, risk, sales volume and quality of products. Most recently, digital twins have been proven in the field and the court of law.

Risks and Mitigation

The cost-benefit of AI makes its use unavoidable. Therefore, it is important to understand risks to proactively avoid and mitigate them.

Cybersecurity

Most crucial is cybersecurity – and we expect this risk to remain at the top for the foreseeable future. One of the primary cybersecurity risks associated with implementing industrial AI in manufacturing and production processes is external network connections to the AI model. Typically, the AI model is hosted in the cloud, and manufacturers establish external network connections from their facility to the AI model. This external network connection creates a potential vulnerability for cyberattacks. Further, in 2022, researchers found that 83% of manufacturing clients still had undocumented or uncontrolled external connections to their operational technology (OT) process environments. Because ransomware groups target external connections, it is imminently important for businesses to tighten their security.

The most effective security control for reducing the cyber risks associated with remote access remains multifactor authentication (MFA). Additional steps to mitigate include limiting the number of different remote access vendors and solutions in an environment; requiring remote connections to request access with authorization versus remaining always active and unauthorized; and achieving the capability to rapidly disconnect external connections. The question is not “what if?” but “when?” Therefore, any mitigation preparation that allows incident response to be successful quickly is best.

Data Privacy

Data privacy is an additional concern and risk if not managed. Businesses should also ensure that data is anonymized as much as possible before it is sent to industrial AI systems. In addition, vendors, solutions and products must be vetted to ensure data privacy concerns are minimized. Now referred to as data difficulties, collecting, sorting, linking, storing and properly using data has become increasingly difficult as the amount of unstructured data being collected from sources such as the web, social media, mobile devices, sensors and the Internet of Things has increased. As a result, it’s easy to fall prey to pitfalls such as inadvertently using or revealing sensitive information hidden among anonymized data.

Physical Safety

Physical safety is another risk in the manufacturing industry. Accidents and injuries are possibilities if operators of heavy equipment, vehicles or other machinery do not recognize when systems should be overruled or are slow to override them because the operator’s attention is elsewhere. It is also important to triple-check override scripts, for example, to ensure the judgment of people building the scripts is not faulty. Accident reports are now showing scripting errors, lapses in data management and misjudgments in model-training data from human input as a contributing factor.

Many in the industry are now using deep machine learning to ensure audits are confirmed by machines double-checking code. Further rigorous safeguards are mandatory. Without them, disgruntled employees or external foes may be able to corrupt algorithms or use an AI application in malfeasant ways.

Conclusion

Businesses must develop acceptable use policies when integrating AI into their operations to mitigate the risks delineated above. Processes need to be created for monitoring algorithms, compiling high-quality data and explaining the findings of AI algorithms.

Manufacturing leaders must embrace AI, make it a part of their company culture, establish standards to determine acceptable AI technologies, secure systems and create a culture of trust and security.

Jade Davis, Of Counsel Attorney at Hall Booth Smith in Tampa, focuses her practice on data privacy, cyber security and construction.

Subscribe to Industry Today