Science fiction abounds with stories of robot or computer takeovers, from 2001 A Space Odyssey’s HAL 9000 to Dr. Who’s Cyber Men—along with the flip side, the original Star Trek’s “Spock’s Brain” episode where the bad guys hooked Spock’s brain up to the system that controlled their civilization’s air conditioning and other processes.
Whether any of this could actually happen is the subject of much discussion today, as current artificial intelligence (AI) processes much more data and makes more decisions than ever before.
But can it think? Can it create?
No, says Meisong Yan, PE, a longtime reservoir engineer and owner of WE Resources. She believes that the capabilities of humans and machines are distinct, but compatible. For this relationship she uses the term “augmented intelligence.”
“By definition,” Yan points out, “augmented intelligence is a design pattern for a human-centered partnership model of people and artificial intelligence (AI) working together to enhance cognitive performance. This includes learning, decision making, and new experiences.
Instead of AI replacing humans, “It is designed to enhance, not replace, one or the other.”
For example: “In the oil and gas industry, years ago, the field of geo-statistics was developed to assist earth-model building while matching the available well logs. A dynamometer was commonly used to help humans diagnose a rod pump’s operation. Those repeatable analyses are now processed much more quickly by computers so geologists and engineers can correct abnormalities and optimize field results.”
As AI advances, can it replace humans? It’s no secret that it already has for some repetitive jobs, but it’s actually created others, such as data science, Yan notes. “With the current digitalization trend in the oilfield, people have started to worry about their job security. AI truly will replace certain repetitive jobs—for example: recording the data to generate a dynamometer and detecting certain abnormal signals.
“But,” she continued, “there are still plenty of jobs that AI couldn’t replace, and some are even created by AI. This is when the government should step in and provide necessary trainings to guide those workers to different jobs in which they could still thrive.”
Saeed Mubarak
Adding, “It’s just a system, it’s a tool,” is Saeed Mubarak. He is chairman of the SPE’s Digital Energy Technical Section, the founder of The Road to the Digital World LinkedIn group and a Texas A&M University SPE student chapter Industry Advisor. Mubarak has years of experience in the digital oilfield with Saudi Aramco. All opinions are strictly his own.
“There isn’t any intelligence other than in a human being,” he said. “Augmented intelligence is a much better expression than artificial intelligence.” He adds: “It’s good to say ‘artificial’ because it’s not real.”
And while we think of AI systems as coldly accurate and unbiased, Mubarak notes that they’re only as accurate and unbiased as the programmer, who tells it what information to use and what to ignore.
“It’s not intentional—sometimes it’s because of a limitation of knowledge. The smartest AI or data-driven model will be exactly the smartness of the programmer.”
Many people are enamored of the technology for its own sake, he observed, but there are two basic requirements for it to be of maximum effectiveness. “One of the required ingredients is a complete set of reliable data. If you don’t have enough data, it’s just like the six blind men touching the elephant. If somebody touches the trunk, they’ll say the elephant is more of a hose. And if they touch the tail, it’s a rope.”
If the data is limited, so are the conclusions. Back in the 1970s, the saying for this kind of thing was “Garbage in, garbage out,” which he said still applies.
At this point Mubarak made what some would consider a startling statement. “People say the oil industry is behind on technology. It is not.”
Modeling processes and reservoirs started before there was any such designation as a petroleum engineer or an SPE, he said. “They created models of oil, the facilities, and even simulation models for subsurface infrastructure, which is the reservoir.” Early pioneers based business decisions on these models created by hand and slide rule long before any intelligence was artificial.
“What we did in the old days, we used physically based models using the limited data, because we didn’t have full data of assets because no one can.” The oil industry observes its world through a series of widely-space 6-inch holes, and it’s hard to get a complete view of anything that way.
Over the last two decades, however, the industry’s processing power has leaped past slide rules and pocket calculators, making that view much better than it was.
“Three things have changed in this domain. First is computing power, the second thing is faster communication, and finally, storage capacity—hard discs, servers, and the Cloud.”
Technology simply for its own sake is not useful, he cautions. “Whatever the technology is, it has to serve a purpose. We don’t just go for technology because it’s fancy, because it’s the song of the day and everybody is riding the wave.”
Lone Star Analysis uses AI and ML to diagnose issues and create predictive models for preventing future problems, in ESPs and a wide range of other industrial applications. They began in 2006 doing this kind of analysis for the Department of Defense, and soon realized that their processes could be used in industry. They entered the oil and gas markets about six years ago, according to Davey Brooks, who is the company’s VP of Automated Intelligent Analysis Solutions.
Davey Brooks
“A lot of the challenges that we were addressing in aerospace and Department of Defense and other areas, were problems that had commercial applicability,” he said. They were analyzing processes that involved physics, chemistry, and other sciences that are consistent in every industry, so they realized it would be simple to take those general principles and hone them for individual areas like oil and gas.
“Learning from some of the engagements we’ve had outside of the commercial space, we’ve looked at how we could apply those capabilities to making organizations more effective and more profitable.
“That was where we came up with productizing things like our ESP solution from an energy perspective, vehicle CBM, and the rotary screw compressor for manufacturing.”
Physics and math associated with any pump or compressor that moves a liquid are the same no matter whether it’s pushing oil or water or air. “If you use that as the starting point in solving the problems associated with a specific asset that has very unique characteristics, it becomes much easier because you’re taking the commonality and saying, we don’t have to learn how all that other [industry-specific] stuff works, we’re going to apply hose physics principles to any variety of assets.”
Brooks feels this sets Lone Star apart from traditional AI programs that start any installation with a learning curve about the individual location or pump. “You don’t need to have the (AI) model train itself, because you know the way that asset should be operating. So let’s pull the sensor data from the asset itself and feed that into the model and say, okay, it should be operating one way, but here’s the way it’s actually operating. so here’s what that trend starts to look like, and then, what is the probability that it’s either going to fail or that it’s going to start to operate outside of an optimal range.”
The goal is to extend the asset’s life and have it run more optimally during that time.
The AI can apply known principles to a prediction of how it will operate, then suggest remedies.
Because an ESP only gives a small amount of data at a given time, it takes longer for a purely field-based model to gather enough to form a prediction—which is to Mubarak’s point about complete and accurate data. Bringing existing knowledge to the table speeds the learning curve, Brooks said.
Recent downturns, including especially the COVID-19 bust, have given clients opportunity to re-evaluate and upgrade their automation systems. This has given rise, Brooks said, to a new level of “We’ve never done it this way before.”
Before 2010 that statement revolved around sending pumpers to sites every day whether anything was wrong or not. The advent of SCADA systems and remote monitoring did indeed make producers more efficient, but advanced analytics pushes that boundary further ahead—and not everyone is ready to make the next leap.
Brooks was recently on a call with a prospective client. While the prospect was excited about alternatives that speed up the processes of ML and AI, it came down to, “How do you step away from the traditional sense of doing it because that’s the way we’ve always done it.” The company then challenged Brooks and Lone Star to use their model to solve issues the prospect’s existing systems couldn’t fix.
When discussing companies’ existing automation systems, he said, “They don’t want to hear that their baby is ugly. I think people have kind of stepped away from that.” They’re realizing that basic automation has not provided the value they need, and are now ready to take the next step.
Are some jobs lost? Yes.
Are the benefits huge? Yes.
Is efficiency vital to the survival of individual companies, if not the industry as a whole? Quite likely.
Yan called for government programs to retrain workers.
Mubarak declared that AI is still dependent on human governance.
Brooks said AI paves the way for his company to help clients across the industrial spectrum.
These issues call for thought and discussion—two things AI can inform, but cannot do.
_____________________________________________________________________________________________________
Paul Wiseman is a freelance writer in the oil and gas sector. His email is fittoprint414@gmail.com.