At Augury, we use a rainbow of AI techniques: picking the right AI application for each specific purpose. “It’s all about using the right tool for the right job,” says James Newman, Head of Product and Portfolio Marketing at Augury. “Whether it’s Industrial AI, GenAI, or Causal, they all have particular strengths that can work to make your work easier and more impactful.”
It’s safe to say that Augury has always been in the Reliable AI business. And happily, we continue to create more tools that are reliable when applied correctly and for the proper use cases. We’ll use any technique to improve our models in real-time. It’s what we do.
Meanwhile, there’s a push to regulate AI, and many governments already have regulations – or will have them soon. In general, it’s about making emerging AI technologies transparent, explainable, safe, and based on FAIR data. These are all reasonable ideas everyone should aspire – in the name of creating Trustworthy AI.
Reliable AI – AI that works to produce the proper desired outcomes – is very much part of this vision. And as you all know, in an industry like manufacturing, you cannot afford mistakes in terms of both safety and the bottom line.
State Of The Art Industrial AI – And Beyond
Our so-called bread-and-butter AI is industry-renowned for its ability to predict when a machine will break down – in fact, it’s even guaranteed. Designed for purpose-built solutions, this AI will continue growing with more use cases, capabilities, and new ways of leveraging insight. And, we’ll continue to use our neural networks to do the heavy lifting in terms of in-depth modeling.
We will also keep experimenting with new and emerging forms of AI. For instance, our recent success with our Machine Health algorithms in applying Continuous Learning, which represents one of the more significant milestones in our quest toward overall expert-level AI, was primarily thanks to the rich – and accurate – training data created by Generative AI.
“Yes, GenAI has a reputation for hallucinating. But it would be best to remember it’s a tool, not an outcome. Gen AI’s accuracy is 100% based on the model it’s being executed against and the parameters around which it is being controlled.”
GenAI Is Your Friend If It’s Used Right
Yes, GenAI has a reputation for hallucinating. But it would be best to remember it’s a tool, not an outcome. GenAI’s accuracy is 100% based on the model it’s being executed against and the parameters around which it is being controlled. GenAI is not evil. It’s down to the people to control it.
The large language models (LLMs) of GenAI work by sucking up lots of data, learning the patterns, and then working to predict the following pattern – in an often-unknowable way. It’s about answering a question that resembles the answers people usually give. Hence, when it doesn’t have the data to fill in the blanks correctly, it starts hallucinating.
In the case of Continuous Learning, we ensured the LLMs we developed only had access to quality data – namely, the over 500 million hours of Machine Health data taken from over 100 types of machines and dozens of industries. In other words, reliable data begot reliable outcomes.
“By embedding a GenAI agent into our platform, we are now able to let our users engage with our AI very quickly and in the natural language they prefer – bringing our best-of-class AI to the front row, as it were.”
Complicated But Doable: GenAI As Reliable AI Assistant
It will take time to unfold all of GenAI’s potential and value, and it will take even more time for a high-risk industry like manufacturing that cannot afford to base its decisions on a hallucination. GenAI is still coming fast, but it will require a lot of work underneath it.
However, one other GenAI use case we will see in the short term is related to how the inner workings of all of Augury’s AIs have been largely hidden from our users. Yes, you are accurately told what machine needs fixing and how to do it within a specific time frame before it becomes a problem. However, the users could generally only dig deeper in a rather manual and cumbersome way.
By embedding a GenAI agent into our platform, we are now able to let our users engage with our AI very quickly and in the natural language they prefer – bringing our best-of-class AI to the front row, as it were.
An AI To Help Explain AI
Naturally, we’re not talking about just throwing ChatGPT at it, which would spark hallucinations and insufficient insights. You need to control the data that ChapGPT is dealing with carefully – and the same goes for any custom LLM we develop.
Either way, this AI agent will help users better understand what’s happening with the AI in the background. Those working on the factory floor can start querying the platform directly for supporting evidence: Why must I look at this machine? What is the metadata? Can I compare the metadata with the metadata of another machine? Has this happened before? Who fixed it and how?
This is where GenAI can shine: engaging with a trustworthy model to give you additional insight.
“It’s not just about considering how one thing impacts another but also how it affects many different factors and what changes you need to make to get your desired result.”
What Will Causal AI Mean For The Future Of The Factory Floor?
In many ways, Causal AI offers the perfect fit for fully transparent and safe AI. Because it’s all about learning cause-and-effect relationships between different data sets, these Causal AI models are very explainable thanks to their very construction.
As we aspire towards full Production Health, we can loop in more knowledge models based on domain expertise – combining data sets from maintenance, production, operations, quality, etc. As we bring in these other data sets, we’ll use Causal AI to examine all the cause-and-effect relationships between what we see on the machines and what the system produces.
In short, it’s about finding true causality rather than just the correlations offered by GenAI. It’s not just about considering how one thing impacts another but also how it affects many different factors and what changes you need to make to get your desired result.
“In short, Reliable AI is not about a single methodology.”
The Dance Of The AIs – As Choreographed By Humans
Causal AI will be huge for manufacturing. By allowing manufacturers to find relationships they may not have seen before, Causal AI will spark whole new ways of doing things to optimize processes.
And to help explain these intricacies, a GenAI agent may be looped in to help – but without losing the power of the Industrial AI that forms the backbone of highly accurate, and dare we say it, reliable AI insights. It’s this dance between different technologies that will define the future of manufacturing.
In short, Reliable AI is not about a single methodology. It’s about using the right combination of methods to provide reliable answers to people in reliability.
Stay tuned for more exciting Augury AI info! Reach out!
