'Artificial Intelligence' has been misnamed

By Mark P. Mills and Julio M. Ottino 
Real Clear Wire

This summer marked the 70th anniversary of the meeting when the term “artificial intelligence” was launched. But it’s been less than four years since the launch of generally useful AI. ChatGPT's arrival in fall of 2022 triggered a tsunami of news stories and commentaries amazed by AI’s parlor tricks, along with myriad narratives about impending AI harms, even catastrophes.

Having witnessed one of history’s shortest journeys from awe to shock, Silicon Valley now has a problem. The problem can be traced to the naming act itself. Calling one of the most consequential technological advances in history “artificial intelligence” has not only animated anxieties but also inspired hyperbole including the idea of an inevitable super-intelligence.

The idea that computing machines might appear intelligent originates in the 1840s with mathematician Ada Lovelace. She assisted Charles Babbage, builder of the world’s first (mechanical) computer. When electronic computers arrived a century later, MIT mathematician Norbert Wiener founded the field of cybernetics to describe how biological and (electronic) machines learn to control behavior using feedback loops—the essential feature that differentiates AI software.

In 1950, Alan Turing—father of the imitation game now called the Turing Test—published a paper titled “Can a machine think?” A few years later, Dartmouth math professor John McCarthy chose the term "artificial intelligence" to promote a Summer 1956 workshop exploring how a machine might “simulate” human intelligence. McCarthy could hardly have foreseen the anxieties that the term AI would animate compared to cybernetics.

Conflating the two words—artificial and intelligence—has trapped debates around questions that are interesting but hardly practical. Is it truly intelligent? Can it replace human intelligence? Will it become conscious? Will it want to hurt us? 

Invoking super-intelligence accelerates anxieties. The name framed the debate before it began, leading to political backlash from bans on data centers to proposals for an AI moratorium. Meanwhile, questions that matter are given short shrift: How reliable are these systems, and how can reliability be reliably measured? Should and can their use be governed differently than other technologies? Who bears costs? Who captures benefits?

Meanwhile, regardless of the urgency to address many real-world challenges to establish appropriate guardrails for a radically new technology, too many of the leading AI practitioners and innovators are instead throwing around, and hyping, even more fraught terms such as “artificial general intelligence” and “super-intelligence” in manifestos and visions of AI’s future.

It’s likely too late to unwind the term AI as a profoundly misleading term-of-art, but it’s useful to understand why the name is a problem. Names are not neutral. They are among the first tools by which understanding is constructed. In effect, names are models or compressed theories that create expectations.

The term AI not only compresses an entire range of technological ecosystems into two words but also misleads. We don’t call a car an artificial horse, nor an airplane an artificial bird. The differences from nature are readily discerned.

Consider another computer-related term, “the cloud.” That name evokes something weightless, benign, and far away. But what it describes is a network of thousands of increasingly enormous warehouse-scale data centers consuming power and water. For two decades that word contributed to the invisibility of the physical reality. AI’s naming problem runs the other way. While “cloud” conceals too much, “artificial intelligence” promises too much and is too ambitious.

Too ambitious because it invokes the most mysterious human capability: intelligence. And too broad because AI tools encompass so many distinct ecosystems: language models, prediction systems, vision systems, scientific discovery tools, autonomous virtual agents, physical robotics, optimization algorithms, and cybersecurity.

Investors and professional communities, though, use metrics, not just words, in order to consider appropriate actions. Investors evaluate expected value: Even a tiny chance AI could transform any of medicine, science, education, manufacturing, or pharmaceuticals justifies massive investments. Meanwhile, communities react to expected utility: Whether jobs disappear, whether infrastructure costs and disruptions land on them, or whether promised benefits ever arrive. A name that inflates both risks and payoffs exaggerates the shock and the awe.

If, instead, we spoke about "expert assistants," "scientific accelerators," "autonomous decision advice," or "natural language knowledge infrastructure," or used well-established specific terms such as “self-driving car,” or even “super-apps,” people might evaluate such tools differently because each implies different benefits, risks, and governance questions.

The taxonomy of where AI venture investments are taking place, and what organizations are in fact doing, already makes clear that such hyper-specialization is the real future. Investors may understand the distinctions, but reality is being trampled by hysteria, further amplified by the hypertrophied term, super-intelligence.

Before we decide how to regulate AI, we should first ask whether its misnaming has led to misunderstanding. We cannot govern what we cannot conceptualize, and we cannot conceptualize what we cannot name. It may be too late to break the world’s addiction to the name, artificial intelligence. But the first step to recovery is to admit the addiction, and recognize that it's unhealthy.

Mark P. Mills is executive director of the National Center for Energy Analytics, a faculty fellow at Northwestern University’s engineering school, and author of “The Cloud Revolution.”
Julio M. Ottino is a professor at and former dean of the McCormick school of engineering at Northwestern University, a member of the National Academy of Sciences, and author of “The Nexus.”