Even though organizations have always used data, its value has become more apparent than ever. Artificial intelligence (AI) and machine learning can analyze large datasets to extract insights and enable data-driven decision-making in any industry, from manufacturing to professional sports to risk management for damage prevention.
Irth has experimented with AI and machine learning for nearly a decade, making us an early innovator in our industry. Data is suddenly so important for others because technology has finally become good enough and more affordable.
At the Irth User Summit, panelists in our session “The Impact of Data-Driven Learning & Large Data Models” discussed how AI and machine learning technology are here to stay and how these advances impact risk management for utility and energy companies.
Here are some of the main takeaways:
ChatGPT and other generative AI tools have been getting a lot of mainstream hype, but they’re just the tip of the iceberg. Tools that use natural language processing (NLP) make it possible to create prototype applications with low or no code. Risk management involves working with lots of disparate types of data—Irth alone has over 130 million tickets in its system—and now organizations can develop proofs of concept for applications that scrub, structure, and interrelate that data without having to invest nearly as much time, effort, and money as they would have in the past.
A few years ago, the largest language models had a couple million internal parameters. This sounds like a lot until you learn that today, the largest language models have 180 billion internal parameters. They’ve become much more powerful in a short time, and the possible applications are nearly endless.
Many risk management organizations aren’t yet able to fully leverage large language models. These models need to be able to access large amounts of data to be properly deployed, which is very difficult when the data isn’t normalized and centralized. Organizations need to make a concerted effort to pull data out of their silos and put it all in the same place so that AI tools can get to work.
For example, many organizations in the risk management industry are currently trying to recreate what a one call center does by using available data to determine where lines are buried. That data needs to be accessible and normalized for that project to be successful. Thankfully, many more people in the industry see value in participating and investing in projects that require de-siloing data, which means the future of AI-powered risk management is bright.
Risk management involves accounting for all sorts of externalities that could cause issues at dig sites. In the past, you gathered as much information about these externalities as you could, and then you crossed your fingers. But now you can put that information to much better use.
Let’s say you have a toe of a slope that hasn’t critically failed but is shifting at a rate of about one foot per year. That would mean that the location you performed a few years ago would be off by a few feet—which could really throw a wrench into things. It’s now possible to detect that shift, measure it with satellite radar, and then feed LIDAR data sets into machine learning models that would output actionable suggestions.
Because these models are built to analyze large amounts of data, you can now make data-driven decisions that account for every possible variable, including operators, line locators, excavators, climate-related issues, geophysical shifts, and more. And the more (and more diverse) data these models include, the greater their efficacy will be.
Until recently, most people had no idea what AI was other than a common trope in science fiction. Those who did know what AI was tended to be skeptical, which was fair, considering how many firms made spurious claims about AI being able to solve any and all problems with the click of a button.
The new wave of AI hype in 2023 meant that more people became aware of what it can do and what it can’t do (yet). As a result, people across communities and industries have become more willing to build and train models to tackle big problems while maintaining realistic expectations. Conversely, AI firms have gotten better at listening to experts in specific domains.
For example, in the first half of 2023, there was a revolution in the meteorological community where they started to use massive AI models to make weather forecasts at scale. And in the risk management community, Irth and its partners have begun exploring how to detect methane events using a mix of satellite images and algorithms. An expert in physics and signal processing can tell you why you might see methane in a satellite image, and then an AI component can clean up the map to make it useful.
With increasingly better collaboration between AI and risk management experts, adoption will only become more widespread–and successful.
Artificial intelligence research is at an all-time high, and as a result, innovation should continue at a rapid clip. Downstream of that research are perpetually growing libraries of open-source code and algorithms that will allow organizations across industries, including risk management, to quickly set up, develop, and deploy test models without having to do anything from scratch.
There will also be more experts who can help synthesize and stratify these AI innovations so that domain technologists can apply them to specific problems—such as choosing optimal dig sites. And with cloud infrastructure getting cheaper, risk management organizations can afford to experiment and fail, which wasn’t the case just five years ago. This is key because experimentation and failure are essential to innovation.
Anyone in risk management and damage prevention knows the impact that GIS had on the industry. AI is poised to have just as much of an impact, if not more so. It’s the next evolutionary stage of this industry, and Irth is prepared to help your organization take full advantage.
To learn more from Irth’s risk panel, check out the full video here.