We’ve been using artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), for many years. In fact, we use them every day for tasks such as data preprocessing, image analysis, online chatbots and driving assistance in our vehicles. All these applications use training sets and responses.
DL is a subset of ML, which uses artificial neural networks (ANN) to learn from large datasets. Generative AI (gen AI) is a type of AI that uses DL techniques to generate new content, such as images, music and text.
ChatGPT is a genAI that recently celebrated its first birthday, and its impact has raised the awareness of the capabilities and risks associated with technology, particularly tools with limited understanding of how it produces results. Meanwhile, managing risk is important to any business, particularly when considering new technologies.
At the recent IoT North event I attended in Calgary, much of the first day’s conversation focused on the application and impact of the Internet of Things (IoT) in various industry sectors. Day two zeroed in on the impact AI will have on our future, reinforcing the relationship that already exists between these technologies. I believe this relationship will only increase. Below, I discuss why as well as some of the challenges that must be overcome to make this future happen.
Let’s start with the basics of how the Industrial IoT (IIoT) is different from IoT with regards to requirements for security, interoperability, scalability, precision and data resolution, programmability, low latency, high reliability and resilience, and serviceability to maintain required availability targets.
The concept of an Internet of operations technology provides an IIoT infrastructure of interconnected entities, people, systems and information resources. This is combined with services that process and react to information from the physical and virtual worlds with real-time and operations technology (OT) requirements. Some of these unique requirements include support of cyclic and report-by-exception updates, area classification electrical requirements, and more rigorously available targets to support multiple years of uninterrupted service.
Though OT has been around for quite some time, in part because of its broadness and potential impact, it’s still in its infancy. Therefore, many of the concepts discussed at the event in Calgary are equally important and applicable for AI implementations.
One challenge many organizations have a hard time avoiding is the concept of “buying shiny things” because they’re new, exciting, and everyone else is doing it. The geek in us wants to play with the new toys, so combining OT with AI is tempting. However, like any project, these efforts must begin with advancing the organization's overall goals.
The next challenge discussed by multiple panels was the need to avoid silos. The more you create silos, particularly with new technology projects such as OT or AI, the higher the risk of failure. Multiple presenters stated that roughly 85% of new technology projects fail due to either silos or lack of clearly defined objectives and outcomes.
It’s also difficult to identify blind spots because you can’t know what you don’t know. However, by avoiding silos, you get a greater range and diversity of viewpoints, and reduce risk while increasing the chance to spot blind spots. Gen AI targets the role of knowledge workers during the creation process, and it looks across datasets for new patterns to help break down silos.
Traditional ML and DL tools are critical for safely connecting and managing orders of magnitude more data available from the OT environment by preprocessing at the edge device, or for identifying the health and integrity of an asset by combining embedded diagnostics with process data.
Yes, new technology is fun. It offers potential huge benefits, too. However, new technology requires diligence and the same discipline as traditional projects to achieve success.
