Technology leaders such as Sundar Pichai (Alphabet CEO) and Andrew Ng (Cofounder – Google Brain and Coursera) have remarked that “AI will be as profound as Electricity,” highlighting its potential as a General Purpose Technology (GPT). GPTs are rare and transformative innovations that fundamentally reshape economies and societies. GPTs are innovations so foundational they ignite waves of complementary innovations – examples include the printing press, steam engine, electricity, computing and the internet.
In the 20th century, electricity transformed every aspect of industry and daily life – it was a GPT of energy and motion, enabling decentralized mechanical work, mass production, and modern urban living. Today, AI is poised to be a GPT of cognition and knowledge, enabling machines to perform intellectual tasks, automate decisions, and even generate content.
As an Operational Risk professional, I find this analogy both fascinating and thought-provoking. In this week’s article, I explore: Where the analogy holds, where it breaks, and what this means for evolving risk frameworks.
Where the Analogy Works: AI and Electricity as General Purpose Technologies (GPTs)
AI appears to be the next GPT. Why?
1. Pervasiveness & Ubiquity
Electricity permeated every corner of the economy, from manufacturing and mining to commerce and the home, displacing the steam engine as the primary sources of power.
AI is rapidly following in the footsteps of past transformative technologies, permeating nearly every industry. It is being used to diagnose diseases from medical images and to predict patient outcomes. In the financial sector, it helps detect fraudulent transactions and drives algorithmic trading. AI serves as the digital hand behind autonomous vehicles, and it is also playing a critical role in accelerating the discovery of new drugs and advanced materials.
2. Continuous Improvement:
The technology of electricity generation and transmission improved steadily. The transition from Direct Current (DC) to Alternating Current (AC) systems, along with continuous innovations in turbines, generators, and grid management, enabled the efficient distribution of electricity over long distances, at scale, and at progressively lower costs.
AI and machine learning systems are inherently designed to improve themselves over time by learning from more data and better algorithms, creating a self-reinforcing cycle of advancement. This is clearly reflected in the evolution of AI models (e.g. GPT-3 → GPT-4 → GPT-4o) each demonstrating enhanced capabilities driven by increases in computational power, data and advancements in model architecture consistent with data and Compute scaling laws.
3. Catalyzing innovation:
The ripple effects of electricity were immense. It not only enabled major productivity gains by transforming factory layouts but also sparked a wave of new inventions such as the telegraph, radio, household appliances, electric lighting, refrigeration, and modern transportation systems like electric trains and subways.
Similarly, AI’s potential to drive innovation may be its most transformative trait. It can dramatically accelerate scientific discovery and technological advancement across disciplines, while also giving rise to entirely new job categories and business models.
4. Societal Transformation:
Electrification fundamentally reshaped modern life and changed how people lived, worked, and connected.
Generative AI is now reshaping creative and knowledge work, with cognitive automation potentially triggering a labor market shift as significant as the industrial revolution.
Where the Analogy Breaks Down: Why AI is Not Like Electricity
This is where operational risk leaders must be especially vigilant. AI is not deterministic, not static, and not passive – which introduces new classes of risks.
1. Deterministic vs. Probabilistic
Electricity follows the laws of physics. AI models, especially GenAI (like GPT, Gemini, Claude), are probabilistic engines. They predict the next word, token, or classification. The same input can produce different outputs. AI can hallucinate, misclassify, or go rogue.
Risk Insight: Traditional pre-launch signoff, post-mortem audits and QA controls aren’t enough. We need dynamic and ongoing validation.
2. Static vs. Learning Systems
Electricity today works the same as it did 100 years ago. AI models evolve with new data. This model drift means performance can degrade or change without direct changes in code.
Risk Insight: Point-in-time reviews will not catch evolving risks. We need continuous assurance and monitoring.
3. No Autonomy vs. Decision-Making Agents
Electricity does not decide to shut down a hospital’s power or deny a mortgage loan. AI can recommend actions or even execute them autonomously.
Risk Insight: AI introduces autonomous decision-making risks. This requires governance structures akin to fiduciary responsibility frameworks (e.g. final lending authority often resides with a human, ethical use principles).
4. Fuel vs. Essence of Data
For electricity, coal, gas or solar is the fuel. For AI, data is the fuel. The quality, source, and integrity of the training data define the performance and fairness of the AI system.
Risk Insight: New threats like data poisoning, prompt injection, and adversarial attacks shift the risk focus upstream into data governance.
Electricity vs. AI: A Tale of Two Risk Frameworks
As foundational General-Purpose Technologies (GPTs), electricity and AI are critical to our modern life and economy. However, their risks are unique, requiring fundamentally different management strategies. This comparison breaks down the differences:
Final Thought: Prepare Your Risk Framework for the AI Revolution
The analogy of AI as the new electricity is compelling but it also carries the risk of fostering complacency. While electricity was tamed through engineering, regulation, and robust infrastructure, AI requires all that and more. Recognizing this, governments, global organizations, and business leaders are already engaging in conversations about AI risks, working to design safeguards that ensure trust, safety, and accountability without stifling innovation.
Managing AI requires risk frameworks that go beyond traditional control systems. Unlike electricity, AI systems can evolve, learn, and act with autonomy creating non-deterministic and context-sensitive risks. This means organizations must adopt continuous assurance, ethical oversight, and technical resilience as part of their risk operating models.
Do you see AI like a General Purpose Technology or Utility? Let me know your thoughts.
Risk Ascent | Exploring the Edges of Risk 
Leave a comment