Energy companies have been urged to embrace the opportunities available through generative artificial intelligence but remain sceptical about how and where it is deployed in the business.

Addressing the WA Energy Club’s May dinner, and in answer to a question on AI and the attraction of talent, Microsoft Director Mining, O&G, and Construction and Industrial, Malcolm De Silva said, “I don't believe in technology for technology's sake. But it certainly can make your employees’ lives easier. And that will obviously attract talent to your organisations.”

Mr De Silva was part of a panel including Microsoft Technology Strategist, Justin Rowling; Microsoft Senior Azure Data and AI Specialist, Richard Lee; and INPEX Analytics and Data Lead, Abhi Raguraman.

The panel fielded several questions relating to privacy, security, ownership of IP, and bias in data results and issues of outdated data.

On talent attraction, Mr Raguraman said that for new and recent graduates, use of generative AI was as standard as emails; a business without email would not attract talent.

Mr De Silva said: “If you want to attract the best and brightest, they're going to want to work on innovative projects. A lot of our clients are really embracing it because you are all competing with war on talent.”

As an indication of the proliferation of generative AI, in a presentation preceding the panel discussion, Mr Rowling said that when Chat GPT was launched,
it reached 100 million people worldwide within three months.

Microsoft has 200 data centres in 60 regions, with generative AI technology requiring more power for its operation. A data centre would soon open in Perth.

“The opportunity for the energy industry to assist by providing clean energy is enormous,” Mr Rowling said. “Microsoft intend to be carbon negative, water positive
and produce zero waste across their entire production chain by 2030, and by 2050 Microsoft will have undone all of its carbon emissions of its 50-year history.

“In practice, the technology to do that doesn’t exist today, so the technology industry is looking to the energy industry to innovate and come up with those new
technologies we need, because we know what got us to where we are today is not going to get us to where we need to go tomorrow.”

He said generative AI had democratised society’s access to technology and to artificial intelligence and broadened the scope of the questions the technology can be asked.

Mr Lee said generative AI’s multi-modal capacity – to deal with pictures, video, voice, and text – had elevated its usefulness. 

“It can create images, video, audio. This is where generative AI is right now, and these models have been created by the wealth of information on the internet.”

At a company level, Mr Lee said generative AI was excellent at retrieving enterprise information without the need for retraining a model.

Generative AI is good at understanding unstructured data such as emails, word documents, and powerpoints, and pulling out key parts of information.

“We can get models that are off the shelf and start using them,’ he said. “The models will give you the information based on the questions you ask."

Use cases in the energy and mining sectors included in health and safety, where generative AI can be instructed to act as an HSE supervisor, checking for correct
PPE and safety equipment and making recommendations on improvements and training.

Mr Raguraman said INPEX established its AI and data science practice in 2022.

The aim was to look at establishing AI and data science to help solve business problems and focus on areas including production and maintenance optimisation.

The company had to first increase its capability through employing data scientists, programers and coders.

The team worked through issues including reliability and safety and ensured they had representation from across the business as they determined the appetite for
AI and how it could be introduced.

On the question of ensuring AI does not act maliciously, the panellists said there were layers of protection that should be built in, including testing before the
model was pushed into production.

Mr De Silva Microsoft had been particular in how it named the off-the-shelf products.

“Copilot is essentially a catch-all branding name for our AI solution or generative AI solutions. The name copilot implies that there's a pilot making the decision.
While there's a level of responsibility on the software makers, there’s also an onus on the user to vet the output and review the decision,” he said.

On the issues of IP, he said businesses owned the IP of developments that came from the use of AI.

“There’s a misnomer that we would use your data to train our model. That's not true. Your data remains your data… it doesn't inform our model.  The retention of IP
and the data is with the organisation,” he said.

In relation to bias and outdated data, Mr Lee said models were initially trained on information available on the internet.

If enterprise data was involved, AI can be instructed to only search information between certain date ranges and prompted in other ways to reduce the chance
of bias and outdated information.

Mr Rowling said: “We in the technology industry use Chat GPT3 as the dawn of the era of generative AI.

“But it is version three of a service that has been around since 2021. There are organisations worldwide that have been using open AI and other services
prior to GPT3 coming out.

“It is a rapidly changing environment; GPT three was the big one that really made the whole thing explode in the market.”

Return to eNews