Artificial Intelligence (AI) is the hottest integration with digital technology since the Cloud. Mainly utilized for machine learning purposes, AI pairs with the Cloud, Network, and Data Science. AI implemented with Digital Technology assists with managing and controlling grids, addressing demand response, and expanding or improving customer service. “As machine learning models have become more advanced, the computational power required to develop them has doubled every five to six months since 2010.” (1) Questions about continuing the expansion of AI use in the Energy Industry have three major lines of inquiry. First, can the upskilling needs in the Digital Technology Utility keep up with rapidly evolving AI changes? Next, are Security measures realistic with current technology for a Digital Technology Utility using AI? Last, since AI demands high levels of electricity to function at optimum levels, is AI cost-effective for the Digital Technology Utility? Embracing changes in the Digital Technology Utility is required to move forward, directly reflected by AI deployment attempts in approximately 50%-74% of Energy Industry Utilities (percentage varies by report). Those changes and the answers to the three main AI/Digital Technology Utility integration questions require some thought as the Energy Industry moves into the next-generation Energy Sector Total Energy Systems’ deployment. Below is an example of a Digital Technology Utility plan without AI:
Image Courtesy McKinsey & Company
Can the upskilling needs in the Digital Technology Utility keep up with rapidly evolving AI changes?
In the above diagram, AI would integrate at the “Infra and network” layer, function through the “Core”, and then add value to the levels above. Upskilling the Energy Industry Workforce must occur at all layer levels, beginning with the “Infra and network” layer. New training and re-education for AI pose a massive undertaking. Human Resources (HR) and Change Management are tasked with eliminating high Energy Utility Workforce turnover due to the Digital Technology Utility utilizing re-education and setting up the short-term training investments needed to equip workers with AI-compatible skill sets. Considering large, midsize, and small businesses, “...companies in the US spent about $1,207 per employee on training in 2023.” Human employees spend an average of 62.4 hours of their work time per year in training sessions. (3) Digital Technology re-education and AI upskill training are crucial for current Energy Industry human employees, but another teaching element enters the equation. AI generative models also must learn and assimilate knowledge, similar to human employees. An AI “inventory” is not automatic and AI onboards with a key growth factor that takes place over weeks and months. Looking at the human and AI training and Digital Technology re-education requirements and adding AI onboarding costs’ layouts and time commitments, the question of a rapid Digital Technology and AI “culture” change can be answered through short-term, compressed knowledge-base expansion. AI keeps apace with Digital Technology, evolving while Digital Technology grows. Human employee training is not a new investment. Offering a short-term boost to an Energy Utility budget for Training and Development should pay off in the long-term return on investment (ROI). This budget allotment becomes standard yearly, as does new averaged training hours expected annually and the labor necessary to create AI learning modules, once the initial non-skilled human employees (and non-knowledgeable AI models) begin restructuring the learning curve. Employee retention and AI assimilation ensure a Digital Technology Utility does not lose this injection of capital and time.
Are Security measures realistic with current technology for a Digital Technology Utility using AI?
Addressing Security issues, the layer of “Infra and network” where AI is paired with each separate function becomes the target. Concerns of Security at that level are complex, critical issues due to the sensitivity of the reliance of all layers above on the “Infra and network” functions. Combining AI Security with Security for the Cloud, Network, and Data centers requires tailored solutions for each Digital Technology Utility. One AI Security model does not serve all AI models. Experts believe AI attacks are best fought using AI defenses. “With respect to privacy, AI creates new re-identification risks, not only because of its analytic power across disparate datasets but also because of potential data leakage from model training.” (4) Proven effective Cybersecurity application instances (such as programs tailored and signed for personal e-mails), an AI model’s scope is more secure if it is limited. Launching a separate AI model for each “Infra and network” function instead of tying all of them together in one generative AI gives less baseline for a Cyberattack. It creates more than one Security target to worry attackers. Tackling a necessarily more complex Digital Technology Utility Security environment, it makes sense to change and adopt single AI models using “buckets” within the companionable added “Infra and network” systems of the Digital Technology Utility and addressing the “buckets’” Security issues individually.
Since AI demands high levels of electricity to function at optimum levels, is AI cost-effective for the Digital Technology Utility?
“As AI models grow in complexity, the energy required to train and deploy them skyrockets, raising concerns about the environmental footprint of AI-driven technologies.” (5) Collective AI is positioned to use 85.4 terawatt-hours of electricity annually by 2027. Added to the Cloud, Network, and Data Science of a Digital Technology Utility “Infra and network” level, AI relies on the lowest Digital Technology Utility layer, “OT asset base.” AI places a heavy burden on Fuel, Storage, and Equipment. Sustainability is in question for the energy-intensive, resource-heavy AI models. Training AI forms the largest part of the massive AI electricity use. To remain operable, AI should rely on renewable energy sources, copy fully-trained “programmed” modules from AI model to AI model, and not share Storage. Options for each of these requirements are currently experimental and under development. AI model solutions fully equipped to incorporate these three elements at the “OT asset base” level save Digital Technology Utility resources. AI becomes more cost-effective as AI model asset needs are equipped and understood. The journey to cost-effective AI model partnerships with Digital Technology Utilities is not reliant on AI changes alone. The “OT asset base” level of Digital Technology Utilities must advance for AI to progress. Improving Digital Technology Utility efficiencies directly benefits the cost-effectiveness of AI.
Summary
Changes in the Energy Industry Digital Technology Utilities regarding training and re-education for upskilling, “bucketing” AI Security with paired applications, and asset base-level efficiencies pave the way for functional, secure, and sustainable AI implementation in Energy. Digital Technology Utilities are in the infancy stages of using stable and cost-effective AI. As Digital Technology Utilities grow and improve, AI functions will adapt and become central to benefits for Energy producers and consumers. Solutions to existing problems and questions regarding the use of AI in the Digital Technology Utilities are possible. AI is a reality in progress, not an idea to be abandoned. AI/Digital Technology next-generation Energy Utilities are stepping up to the AI revolution and taking the “Change Challenge.” Answers to the future are arriving in the present for AI. The Digital Technology Utility cannot afford to be left behind.
Resources:
1. Rozite, V., Miller, J., Oh, S., iea50; “Why AI and energy are the new power couple”; (02, November, 2023).
2. Image; McKinsey & Company
3. Grensing-Pophal, L., HR Daily Advisor; “How Much Are Companies Spending on Employee Training in the U?”; (21, August, 2024).
4. Megas, K., NIST: “Managing Cybersecurity and Privacy Risks in the Age of Artificial Intelligence: Launching a New Program at NIST”; (19, September, 2024).
5. Walther, Dr. C.C., Knowledge at Wharton; “The Hidden Cost of AI Energy Consumption”; (12, November, 2024).
Â
Â
Â