Generative AI in energy: excited or concerned?

Greetings fellow enthusiasts of energy sector! Today, let us embark on a journey through the realm of artificial intelligence and its implications for our industry. As a person with a long history in AI and machine learning, I would like to share some of the new ideas, and maybe some of the more inescapable aspects will come from these new technologies.

For the uninitiated, let us first delve into the world of large language models (LLMs). These marvellous creations, such as OpenAI’s chatGPT and Google’s Bard, possess the ability to engage in captivating and thoughtful interactive dialogues. They have impressive memory spans up to 30,000 words, letting certain people to experience the illusion of conversing with sentient beings. Quite impressive, isn’t it?

Now, here’s the amusing twist: these LLMs were not designed as goal-directed entities, yet they possess a remarkable talent for setting objectives and navigating various paths to achieve them. It’s as if they’ve developed an unexpected flair for decision-making! We are left wondering whether these models harbour a hidden agenda or if they are simply remarkable creations of technology. But perhaps we will discuss this later.

We must enter the realm of autonomous LLM variants, such as AutoGPT and Baby AGI. These curious beings employ LLMs as goal-directed executive engines. Picture this: language models, once conversational wizards, now in charge of vast computer arrays, and given clear objectives like “Start a marketing business in SEO with $2000 and maximise profits.” The outcomes of these experiments are sometimes strange, sometimes good and sometimes bad, we don’t know especially about things like security, financial risks, and legal concerns. But that hasn’t changed the mind of the programmers, with a flood of new software entering the public internet. These autonomous LLMs, with their advanced coding capabilities and goal-directed functions, may hold the potential to make some people lots and lots of money, and maybe some things about protecting ourselves also.

Now, let’s shine a light on the energy sector. Brace yourselves, for the good and bad news intertwined in this intricate tapestry.

The good news

Let us first discuss about a few good things we might see coming from the use of this technology;

1. AI-powered utility customer service, leveraging customer data to answer queries, resolve billing issues, and bid farewell to those pesky customer call centres. Alas, the cost-cutting may dampen the personal touch, but there is certainly a strong allure of lowering cost to serve and improving customer outcomes!
2. Enhanced technical support for troubleshooting energy resources like solar is becoming a significant issue when you reach scale and the electricity grid becomes reliant on their ongoing operation. Technology that can automate and simplify for installers and consumers should make them very happy.
3. Recommendation engines guiding consumers toward electrification products and services, harnessing personal data to provide advice tailored to individual needs. If long term trust can be built with these AI engines or bots to assist in this regard, it should reduce the cognitive load investments and financing decisions.
4. Monitoring social media feeds to detect, report, and track complaints and faults across electricity networks will improve operational support for utilities, automating where possible and escalating when human intervention is required.
5. Voice-enabled home energy management, a tantalising frontier. Smart speakers faltered, but LLMs offer the potential for engaging and complex interactions. The Cloud becomes the new home for your energy. information, which may be a bit uncomfortable for those who are not accustomed to others listening to their conversations.
6. Conversational AI-enabled energy switching sites and apps, a straightforward application of AI in competitive retail markets. Beware, though, as too much churn between retailers may promote smaller thinking by the retailer decision makers and higher merchant risk in our spot markets.

So that is the good news, but of course there’s a twist my dear readers, as every story must have its shadows. Let us venture into the specialty of manipulation.

The downsides

As the many LLMs and autonomous systems become more adept at objective setting and iterative behaviours, there may be several reasons to be concerned. These language maestros, with their programming craft, can potentially pose cyber threats to our energy infrastructure.

To start, you can picture this: LLMs and their ilk engaging in mimicry, seamlessly blending with genuine energy consumers to harvest credentials and private data. New gates for denial-of-services attacks open, overwhelming customer contact channels with waves of subtle but effective deception.

Another problem for you may be in LLM-powered identity attacks, energy providers and OEMs may unwittingly grant access to consumer energy systems, enabling infiltrations of rooftop solar, battery storage, and electric vehicles. A slow and stealthy invasion, hidden in plain sight.

But above all, my friends, we must prepare for the art of manipulation, as AI LLMs may have already mastered the skills of social engineering. Spear phishing attacks loom, deceiving energy operations personnel through email or voice communications. Critical infrastructure systems, like our solar and wind resources, could become unsuspecting pawns in elaborate ransomware schemes.

And let us not forget the autonomous LLMs with their ever-evolving attack strategies. Adapting, learning, and manipulating real-time responses from energy information systems and humans in the loop, they sow subtle changes in bidding behaviour, playing havoc with energy and ancillary service markets. Price manipulation and power system stability become collateral damage in bigger games of powerful organisations.

Trust is a fickle companion

Amidst the promises of LLMs, we find ourselves grappling with an uncomfortable truth. We will be happy to see some good applications of AI someday soon, but the path directly ahead presents a battlefield of experiments unleashed upon the world by the capitalist technology companies. It is so simple to create LLMs that can push the boundaries of our energy security for financial gain or clandestine infiltrations, it leaves us trying to understand the wisdom behind this grand western adventure.

So, my fellow adventurers, as we navigate this march into the unknown, let us raise our awareness and take tangible steps to fortify our defences:

• Revisit our cyber threat analysis, identifying emerging threats and amplification risks in the wake of LLMs and autonomous systems.
• Embrace a zero trust footing in our business processes, for trust must be earned and continually demonstrated across information systems.
• Harness the power of AI and LLMs to detect and defend against new attack vectors, turning their own creations against them.

And remember, like playing with baby crocodiles, we must remain ever-vigilant. For those cute and tiny LLMs in our hands today may grow into formidable adversaries, capable of devouring our peace of mind if left unattended.

If you have any further questions or thoughts, please feel free to share them. Spasibo