What They Didn't Tell You About AI: It's Just "Normal" Technology

The study treats AI as a normal technology, not a superintelligence, and draws on past technological revolutions to project gradual implementation. Image: Freepik

We're old enough to know that tech hype is practically a structural feature of modern capitalism. In saturated markets, the relentless pursuit of differentiation is a survival condition in a venture capital arena hungry for disruptive narratives to justify their valuations.

In this sense, capitalism doesn't just sell products, but also dreams of the future. However, in the case of artificial intelligence, have market players gone too far? Constant versioning patterns point to an artificial intelligence that will emerge, the media says, like a kind of black hole.

From this "singularity," already knocking at our doors, "no sector will escape," say influencers, tech gurus, politicians, and even some academics seeking prominence. "AI will transform EVERYTHING," warns the AI Safety community. "Adapt or die," claim the FOMO hysterics (those panicking out of fear of missing out).

Now, a paper and book project, published in April by the Knight First Amendment Institute at Columbia University, brings a refreshing breeze to both the fog of a dystopian vision of an apocalyptic scenario without empirical basis and the sweet, cloying incense of the utopian vision.

For the authors of this public reflection essay, Arvind Narayanan and Sayash Kapoor, respectively a computer science professor and doctoral candidate at Princeton University, AI is "normal technology," just like other "transformative general-purpose technologies, such as electricity and the internet."

AI Innovation Is Fast, Adoption Is Slow

The study notes that another technological revolution — industrial electrification — took 40 years to be implemented. Image: karlyukav/Freepik

According to the study, the major problem with the current AI debate is that both sides place it in the position of an autonomous agent, capable of making decisions on its own and choosing humanity's destiny, attributing "full autonomy" that would leave it beyond human control.

When they speak of "normal technology," Narayanan and Kapoor are actually rejecting this technological determinism. Drawing from historical transformations, the researchers emphasize that it's human institutions — not the "will of the technology" — that shape AI's social impacts.

The assertion functions simultaneously as a description of the present (how AI really is today), a prediction of the future (a bet on its most likely trajectory), and a prescription for how we should treat it, including guidance on desirable policies and behaviors.

"We do not think that viewing AI as humanlike intelligence is currently accurate or useful for understanding its societal impacts," the authors state in the study. Similarly, this anthropomorphic perspective doesn't provide a basis for anticipating future developments or guiding decisions.

But perhaps one of the study's most important contributions is distinguishing between AI creation and adoption. While 40% of Americans have tried generative AI, they use it sparingly — just a few hours per month, not per day. In other words, people need to relearn how to work.

Drawing an analogy with the Industrial Revolution, the study recalls that electricity took 40 years to truly transform factories, because it wasn't just a matter of plugging into an outlet. Everything had to be redesigned: machine layouts, work processes, and company structures.

The Great Lesson of Electricity and AI's Future

During the Industrial Revolution, workers began supervising the machines. Image: Courtesy of United Artists

The parallel between the arrival of electricity during the Industrial Revolution and the introduction of AI isn't coincidental. True transformation only occurred when entrepreneurs realized they could place individual electric motors on each machine.

More than a century later, AI may be going through the same process. Today, many companies and individuals are simply "plugging" AI into existing processes, such as using ChatGPT to write emails or automate specific tasks. But this means making these tasks more expensive.

True transformation, like effective use of the technology, will only come with a complete redesign of business models, professions, work routines, and even social structures. Ironically, the belief that AI will create entirely new ways of producing, interacting, and living is what fuels the hype.

Just as Industrial Revolution workers began supervising machines, future jobs will tend to focus on configuring, monitoring, and supervising automated systems. Narayanan and Kapoor warn that without human supervision, AI can make frequent errors that could render it commercially unviable.

Despite some complacency regarding employment impacts and some excessively categorical predictions about human supremacy, the paper fulfills its purpose: it offers a balanced perspective between two extreme scenarios and suggests policies grounded in historical lessons rather than speculation.