CERAWeek by S&P Global — an annual global energy conference focusing on the industry's biggest goals and challenges — has officially begun. The conference is aimed at bring world leaders and experts in the energy sector together to discuss themes and challenges presented in the modern day. One of the top themes is artificial intelligence and how it can permeate within the sector to solve problems.
Head of Energy and Natural Resources at Palantir, Matt Babin, (PLTR) joins Yahoo Finance to discuss the energy transition, AI, and more covered at CERAWeek.
In terms of concrete examples of what AI can do in the sector, Babin states: "Let's say that I'm a petroleum engineer working on an asset. I'm trying to answer the question of 'is my asset producing as well as it can?' But what does 'well' mean? Does it mean as many barrels of oil as possible? Does it mean as little water? Does it mean lower carbon? Does it mean lower cost of maintenance? It could mean any of those things and those all pull different data sets at different times from different tools. You can use a large language model to help answer that question and that enables you to then solve different objective functions at different times."
Watch the video above to hear Babin explain why the energy sector is a big one for Palantir.
For more expert insight and the latest market action, click here to watch this full episode of Yahoo Finance Live.
Editor's note: This article was written by Nicholas Jacobino
Video Transcript
[MUSIC PLAYING]
JULIE HYMAN: You're watching Yahoo Finance. I'm Julie Hyman. And of course, AI has infused everything, even the energy industry. I'm here at the CERAWeek by S&P Global Conference in Houston, Texas, where the industry gathers to talk about all of the pertinent issues. And of course, AI is on that list.
And so my next guest, perfect to talk about this issue, that's Matt Babin. He's head of Energy and Natural Resources at Palantir. Matt, thanks for being with us. Appreciate it.
- Thank you for having me.
JULIE HYMAN: So I was talking off camera with someone earlier today, who said he felt that the energy industry was perhaps not as far along when it comes to AI as perhaps some other industries. Would you agree with that assessment? Why? And how-- or how do you get them up to speed?
MATT BABIN: Yeah. I think I'd disagree with that assessment actually. Maybe we take contrarian takes to a lot of things at Palantir. But I think the energy industry has been using artificial intelligence and machine learning for over a decade. They've just been using it in a different way than we're speaking of it right now. If you think of things like deterministic models and reservoir simulations, high-performance computing centers, the energy industry has been using that technology for a long time.
What I think is interesting is the proliferation of speed of development of what we're seeing now in large language models in gen AI pieces is causing the energy industry to think of using technology differently, right? I think the energy industry moves sometimes on longer cycles, right. A boom or bust of the commodity takes three or four years. Procurement can take a year. Information security reviews can take six months.
In the last year, we've had GPT-4, Turbo, Vision, Aramco announcing meta brain. That's four huge large language models all in the space of one procurement cycle. And so where I think the energy industry can move faster is sort of experimenting and adopting some of these new technologies in an accelerated fashion.
JULIE HYMAN: How? You know, and how is gen AI specifically going to be pertinent to them or helpful to them?
MATT BABIN: Yeah. I think it'll be different for different companies. And I think that's probably the first big trap is that some people are looking at AI as a tool and then finding a problem. And I think the companies that are doing this successfully are finding a particular problem they want to solve, and then looking to technology as a lever to do that.
So perhaps that's that you want to reduce your power consumption in the field, or you want to better track your emissions profile and mitigation. You want to have fewer suppliers or fewer supply chain interruptions. Those are all great places where you can use AI and machine learning and bound models to solve problems. But that all starts with finding the right problem rather than just saying you're going to use a technology.
JULIE HYMAN: Now in terms of who you guys are working with, I know you've worked with BP for quite some time as a client, Kinder Morgan another that you're working with. So can you just give us a couple-- for me, the challenge with AI is always wrapping my head around concrete examples of what it is doing right now. Can you give us a couple of examples?
MATT BABIN: Sure. There's two classes of problems that I think are the best fit for this technology right now, and both involve bottlenecks. So how are you using technology to get through a bottleneck? The first one is on context. So let's say that I'm a petroleum engineer working on an asset, and I'm trying to answer the question of, is my asset producing as well as it can?
Well, what does well mean there? Does it mean as many barrels of oil as possible? Does it mean as little water? Does it mean lower carbon? Does it mean lower cost of maintenance? It could mean any of those things and those all pull different data sets at different times from different tools. You can use a large language model to help answer that question, right. And that enables you to then solve different objective functions at different times.
The second type of problem is one around capacity. So let's say that my job is approving some element of a process. I'm approving invoices. I'm approving utility bills that I get for my assets in the field. I'm evaluating pig runs in a midstream company for integrity issues, right. Those are all things where, again, AI and a large language model can help me triage those more appropriately. I have a limited amount of bandwidth and capacity. I want to put my smartest brains on the hardest problems, not on procedural problems. And how do I use AI as a triage tool to get through those more fast-- more quickly?
JULIE HYMAN: Interesting. So you've talked about what it can do for energy companies. What is the energy Opportunity For Palantir? How big a business is it now, and how big do you want it to be?
MATT BABIN: Yeah. We've been in energy for over a decade, although a lot of people don't know that from the beginning. You know, I think our technology, building on what we built for the government at the very beginning sort of a real focus on privacy on data protection-- in the government context, that looks like civil liberties. For commercial entities, that looks like data protection and security. We're most successful in highly regulated industries, right, where you're never going to turn to a black box model and say, I'm going to drill this well because a model told me to. I'm going to approve this patient for treatment because a model told me to.
So if you look at our largest commercial industries of heavy manufacturing, energy and oil and gas, utilities, and health care, those are all heavily regulated industries where decisions really matter. The energy industry is one of our three or four biggest sort of planks of our commercial business. But I think it's very small compared to what it will be a year or two years, five years from now.
JULIE HYMAN: And when it comes to these energy companies, you know, from what you described helping them with these large language models, helping them with AI are they collecting the data, and then you're helping them on that side of it? Or are you also helping with some of the data collection at this point?
MATT BABIN: We never do data collection. We never own data. We are never a data owner, data broker, data seller. Our customers, we provide them software. And they use that software to interrogate, analyze, and make decisions on their own data. They own that data.
They own what we call the ontology, which we think is sort of the most important point of-- you know, a large language model knows nothing about your business. Large language models are fascinating pieces of technology. But if you are Exxon or BP or Kinder, they don't know your business and they don't know what you're trying to do. The ontology is what binds that large language model to those nouns and verbs of your business, but customers own all of their own data.
