Let’s talk about the changes that we have seen over the last few years in Artificial Intelligence.
Artificial Intelligence is not new and has been around for decades. However, we see a renaissance of this technology due to a few key factors.
The first key factor, and the most relevant for today’s topic, is the emergence of a new generation of edge controllers that are more capable and cost-effective than anything we have seen before. In some cases, for just a few hundred dollars, you can get a laptop class processor with gigs of memory and local storage space in an industrial enclosure capable of running complex machine-learning models.
We also see most cloud platforms being extended with native machine-learning capabilities based on open frameworks and libraries that enable pretty much anyone to develop an AI-based application with minimal effort or previous experience.
All these capabilities, when used correctly, have the potential to transform your operations drastically.
We have learned from the many conversations with our customers in this area that their challenge is not whether technology can help them solve a problem or not, but what are the right problems to solve that will deliver the most value to their business.
That’s why technology is not the most crucial aspect when selecting a partner, but industrial domain expertise and process knowledge are.
What about Machine-learning (ML)?
We truly believe that AI, and more specifically, Machine-learning has a tremendous potential to help with the challenges I have highlighted.
The key is to implement it so that it is open to continuous input and learning from the operators and not a black box implementation. The objective is to hide the complexity of data science and programming and present the results and events in the process context so that the operator can naturally interact with the models and continue to expand its precision and capabilities.
In effect, the machine-learning models are capturing the operator’s expertise in the context of the process and automating it, preserving the knowledge of the most experienced operators and allowing less experienced ones to learn from it.
By layering in other techniques like workflows on top of the detection, and classification done by machine-learning models, we can then implement fully autonomous optimization strategies that enable the agility required to react to volatile market conditions.
Is our product’s price up in the market? Let’s prioritize throughput while managing or minimizing emissions. Is our product’s price down? Let’s prioritize cost management and productive asset lifetime.
In today’s world, we need to be flexible in what our optimal process strategy is, but we also need to be able to change it to adapt to rapidly changing market conditions. Agility is what sets businesses apart.
Check back for my next blog about the role of edge computing.