Artificial intelligence (AI) has been making big headlines, especially since San Francisco-based OpenAI launched its ChatGPT chatbot and virtual assistant in November 2022, partly because AI has the potential to revolutionize many, if not most industries around the world — including Canada’s oil and gas industry.

In this, the first of a five-part series exploring the potential impact of AI on the Canadian oilpatch, some basics about AI will be provided, the oil and gas advantage explored, and major benefits and potential impediments to AI’s adoption will be discussed.

The following four parts of the AI and the digital oilfield series will explore potential applications, introduce a few upstream service providers, and discuss how to overcome two major impediments to AI’s adoption: a lack of in-house skills and talent; and the threat of cyberattacks.

Artificial Intelligence 101

The precise definition of AI has been heavily contested, but in my mind Columbia Engineering, a department of the Ivy League university in New York, has provided as good a definition as any:

“Artificial Intelligence is the field of developing computers and robots that are capable of behaving in ways that both mimic and go beyond human capabilities. AI-enabled programs can analyze and contextualize data to provide information or automatically trigger actions without human interference.”

To do so, according to the San Francisco-based AI Accelerator Institute, AI must be able to accomplish four very human tasks:

• Learning, which allows AI systems to assimilate data and enhance their functions autonomously, without direct human coding.
• Reasoning and decision-making to allow these systems to employ logical rules, probabilistic models, and algorithms to derive conclusions and make decisions based on inference.
• Problem-solving, which allows AI systems to process data, manipulate it, and apply it to devise solutions for specific issues.
• Perception using real or simulated sensory organs to allow systems to interpret data to identify objects and comprehend their physical relationships (e.g., distance) to these entities.

The seven major branches of AI to achieve these four tasks, again according to the AI Accelerator Institute, are as follows:

• Computer vision – techniques that assist computers in seeing and understanding digital images and videos.
• Fuzzy logic – a technique that helps to solve issues or statements that can either be true or false.
• Expert systems – a program specializing in a singular task, just like a human expert.
• Robotics – programmed machines that can automatically carry out complex series of actions.
• Machine learning – the ability of machines to automatically learn from data and algorithms.
• Neural networks/deep learning – at the heart of deep learning algorithms, neural networks are inspired by the human brain, and they copy how biological neurons signal to each other.
• Natural language processing – allows computers to understand both text and spoken words like humans do.

“Machine learning (ML) is where most of the work is presently being done in AI, and as a result has become its key driver,” Russ Erickson, VP of investment and partnerships at Edmonton-based Amii [Alberta Machine Intelligence Institute] Research, told DOB Energy. “Despite ML being a subset of AI, the two terms are now often used interchangeably.”

It should be noted that Columbia Engineering classifies machine learning algorithms into three types:

• Supervised learning – where machines are trained with labeled data to predict an outcome.
• Unsupervised learning – where machines are trained with unlabeled data, with the model extracting information from the input to identify features and patterns, so it can generate an outcome.
• Reinforcement learning – where machines learn through trial and error, using feedback to form actions.

The oil and gas advantage

The oil and gas industry is well-poised to adopt artificial intelligence and machine learning technologies for a number of reasons, according to Nick Robbins, executive advisor, energy at AltaML, a leading AI-powered solutions company based in Edmonton.

The first reason is the maturity of datasets in the industry, especially in respect to: subsurface and reservoir characterization; drilling and completions data architecture; and production and operations data.

“Relative to other industries where sensors, instrumentation, and reporting architectures can be rather nascent, the oil and gas industry has been innovating in data capture, synthesis and interpretation for the better part of a century — and collectively we build upon that existing foundation of data every day,” he said.

In addition, the industry benefits from a culture of innovation and scale superior to other industries, according to Robbins.

“There are multiple ways the industry builds structured opportunities to gather data and actionable insights that can be leveraged for AI,” he added. These include communities of practice, citizen science programs, and informal opportunities to share best practices along the entire value chain.

“Finally, the oil and gas industry’s relentless focus on measuring value allows teams to evaluate the relative impact of their investments in AI/ML initiatives,” Robbins said. “This involves not only demonstrating the value but also the relevance and scalability of AI projects across their entire portfolio.”

Major benefits

The primary application of ML and AI is to optimize existing processes by finding those use cases with abundant and accessible data and a clear value proposition, according to Bruce Duong, senior manager of recovery technologies at provincial Crown corporation Alberta Innovates. “Fundamentally, wherever there is data being collected, there is an opportunity to use that data to improve existing practices,” he said.

“The use of machine learning is an added layer on top of that existing data, made possible with innovations in techniques and computing power that allows data analysis and forecasting at a rate much faster than has been possible in the past,” Duong added. “Deploying ML and AI reduces friction in business processes — i.e. not a substitution to human intervention, but as a tool to improve existing workflows.”

“And historically, beyond recovering physical resources (molecules of oil and gas), the industry has been recovering, managing, analyzing, and relying on data to improve economic and environmental performance.”

In terms of the upstream, “ML and AI are enabling tools that help find and get resources out of the ground — i.e. the basic laws of thermodynamics still apply, but with digital tools we can get closer to optimal performance.”

“In that context, opportunities exist where there is a dataset or use case, for example, artificial lift, where there is a delta between the optimal operation of a pump (based on rated throughput, temperature, rpm etc.) and its actual performance (volume of oil, temperature, production fluid) — and we can build a model based on collected data, to give a prediction of improved performance with adjusted operations (pump speed etc.),” he said.

According to Duong, AI/ML use cases can be divided into two broad categories — ‘increasing value’ and ‘reducing cost’.

“To increase value, consider improving resource recovery (better exploration, drilling, production), operational efficiencies, transportation optimization etc.,” he said. “To reduce cost, consider real-time environmental monitoring (emissions, GHGs, water), predictive maintenance, detection of leaks or anomalies, improved health and safety.”

Major impediments

On the other hand, Erickson, Robbins and Duong made mention of several potential impediments to adopting AI and ML by oil and gas — and other — companies.

“Trust and acceptance from human users are a significant challenge,” Duong said. “As in the adoption of any novel technology, there is a learning and understanding curve that is required to make sure everyone is on the same page of what a technology [such as AI/ML] can and cannot do, so that there is comfort with deploying these tools for specific processes.”

And this is magnified in the case of AI and ML, making “explainability of ML and AI workflows crucial for human operators and users to ‘turn over’ specific decision making and forecasting processes to an AI,” he added.

In addition, the initial investment by oil and gas companies to adopt AI-related solutions even on a limited scale can require “a large initial investment in terms of time and money,” according to Erickson, especially since data consistency is imperative for making good decisions. “Otherwise, it’s a case of garbage in, garbage out,” he said.

Organizational capacities also pose a significant challenge within the oil and gas sector, Robbins said.

“To be frank, it’s very difficult to obtain and retain AI and data science talent, while the oil and gas industry has been burdened by perceptions of social license and brand equity, affecting an exceptional pool of young, talented data scientists. Seasoned data scientists find that pursuing a multi-disciplinary, multinational, and industry-agnostic career in a tech company or consultancy far more appealing. These roles not only offer more variety, but often provide higher compensation and value propositions.”

This makes it even more difficult to scale up individual or niche AI projects “from a specific dataset, asset, or team to an enterprise level,” according to Duong, “as those incremental improvements need to be deployed across the operations of a company to have a significant impact.”

Finally, “there are ongoing concerns around data privacy and cybersecurity — as these vast amounts of data are being collected, distributed, stored, analyzed — there are multiple touchpoints where bad actors could access that data and information,” Duong said. “There have been notable examples of operation-critical data breaches that have had significant negative impacts.

“As more processes become dependent on ML/AI based workflows, sufficient safeguards and security are going to be necessary to ensure the whole data chain from acquisition to automation is protected.”