How ‘deep learning’ AI could change how the energy industry solves problems

Joshua Eckroth, Chief Architect, i2k Connect - Photo by ACCELERATE

The energy industry is sitting on mountains of data.

But unless the sector’s major players can find ways to easily unearth valuable insights from that treasure trove, the datasets are just unrealized potential.

The companies leading energy transition are already using AI to summarize and understand their existing data, to extract the insight they need to proactively solve problems they’ve encountered in their daily operations.

The adoption of AI has so much potential that at the Society of Petroleum Engineers’ ATCE conference in October, three companies agreed to explore developing an AI-powered large language model (LLM) specific to the energy industry.

One of their goals is to develop an AI model that can answer technical questions.

A LLM is a type of AI that uses deep learning and large datasets to recognize, summarize, and generate content.

Joshua Eckroth is the chief architect at i2k Connect Inc., one of the companies developing the LLM. i2k specializes in understanding industry domain knowledge within corporate documents.

Its solution can scan millions of documents, extract information, and then structure the information so AI can learn from it. The goal is to help users make better business decisions and run more efficient operations.

“We’re not asking people to take stuff out of SharePoint or wherever else it lives — that’s just where it’s going to be stored,” says Eckroth, who also holds a PhD in AI from Ohio State University. “Our solution crawls and analyzes documents in situ wherever they are stored, and adds metadata to them.”

Large language models can give users the answer to a specific question rather than telling them their keyword appears in 26 different files stored in SharePoint. The search for knowledge takes a few minutes rather than a few weeks, he says.


What’s the value of a large language model in the energy industry?

While i2k’s solution works mostly on documents, it has far-reaching potential for the energy industry.

The solution can already perform image understanding, which could pull out data points such as the depth of a well over time, Eckroth says.

It can also process a 12-hour internal video lecture series and add a query about stuck pipe, so that users are directed to the exact timestamp in the series that addresses their specific question.

“Equations we’re starting to handle now, too,” he says.

When it comes to training an employee to perform a new set of tasks or take on a new role, companies could use a LLM to accelerate knowledge acquisition by employees.

They could build models that understand the individuals’ competencies based on their resume, or their LinkedIn profile, Eckroth says. Departing employees or people moving on from the role in question could share their insights about it, which will be provided to people who need that knowledge.

The result is a LLM that can create and deliver content specific to the worker, a task, and based on what company documentation or training videos already exist.

“It’s a whole new paradigm of computing,” he says.

LLMs have plenty of potential, but also come with challenges and barriers.

Siloed data is one of those challenges, and many energy companies store their proprietary data onsite, while many LLMs have trained only on public data and have little knowledge about any specific company’s proprietary knowledge.

Deploying a large language model for internal use can be tricky, Eckroth says, because most companies do not yet have their own custom-trained models nor the expertise for building their own.

Ongoing feedback is another challenge, as a LLM won’t automatically recognize the introduction of new terms and the evolution of language, he says.

“There’s going to be an evolution of the data and the world you’re operating in,” Eckroth says. “Once you deploy it, you need a feedback mechanism. The things you used in training to see if it worked — keep those active after you push it out,” he says.


Adding AI to emerging cloud solutions

The emergence of AI as a critical tool for the energy industry comes on the heels of a surge in the adoption of cloud computing.
It’s a pivotal moment in the energy industry’s technological journey, and cloud technology provides a significant opportunity for energy companies to collaborate alongside stakeholders such as workers, communities, and investors to achieve a more sustainable energy future.

“We know these new technologies will play a fundamental role in not only facilitating but also accelerating the world’s energy transition goals,” says Pramod Jain, CEO of Computer Modelling Group (CMG).

Because of how complex and infrastructure-heavy the energy sector is, adopting cloud solutions can be particularly challenging in comparison to other industries.

Fortunately the payoff of adoption can be game-changing, and serve as a critical driver of AI and LLM. The advantages of cloud solutions for energy companies include:

  • A secure way to scale operations: A lot of investment has gone into developing cloud technology that ensures data is safely and securely stored, as well as accessed by authorized users. In turn, secure cloud solutions can lead to the deployment of new workflows that help to grow a business.
  • The opportunity to attract and retain new talent: To compete, energy sector companies increasingly require more skilled labor. Concurrently, younger generations who are looking for technologically-based careers with extensive growth opportunities are drawn to roles that use data — which of course includes cloud computing.
  • Better teamwork thanks to more widely available data: Cloud solutions help to “liberate data and make it available to everybody, from the geophysicist to the reservoir engineer.” Multi-disciplinary teams can make more cross-connections and ultimately better decisions when everyone involved has access to and can visualize the same data.


For these reasons and more, those in the energy sector are increasingly investing in cloud computing.

CMG, for example, recently acquired Bluware, a company specializing in cloud and interactive deep learning solutions for subsurface decision-making, including seismic interpretation.

Ultimately, the combined strength of cloud computing and AI solutions like Bluware, enables energy companies to scale production more effectively and efficiently, as well as better collaborate with workers, communities, and investors to achieve a more sustainable future.

Related Posts