A short history of the names given to the field of thinking machines, framed by a satirical 2026 executive order in which President Trump renames “artificial intelligence” to “super intelligence.” The video traces how the field has cycled through more than a dozen names over 70 years — and how each renaming reflects either a scientific schism, a public backlash, or a marketing reset.
Norbert Wiener, already famous from his 1948 book Cybernetics, looked at the biological brain with an engineer’s eye. He described a neurological patient whose hand oscillates violently when reaching for a cigarette — cerebellar tremor, the brain overcorrecting each movement — and pointed out the same pattern in a badly designed thermostat that sends room temperature into oscillations. The study of such self-correcting systems is control theory; Wiener named its application to machines and animals alike “cybernetics,” from the Greek for steersman. By the early 1950s it was the recognized name for the study of thinking machines.
But the field’s fate was tied to Wiener’s difficult personality. A child prodigy pushed harshly by an abusive father-figure relationship, he carried depression and a distrust of authority. After working on WWII anti-aircraft guns, he refused further government and corporate work out of fear his science would be misused (biographers say paranoia), and he cut ties abruptly with close collaborators. His “brand” was not for everyone.
In 1955, John McCarthy, a 27-year-old assistant professor at Dartmouth, wrote a grant proposal whose title contained a phrase nobody had used before: “artificial intelligence.” His premise — that every feature of intelligence could be precisely described and simulated by a machine — was less biological and more logical and symbolic than cybernetics. He asked for $13,500 and got about half, enough to run the summer 1956 workshop with attendees including Allen Newell and Herbert Simon, who demonstrated the Logic Theory Machine: it proved propositional logic theorems by applying hard-coded rewrite rules backward from the theorem to the axioms. Here intelligence meant choosing the right sequence of rules — no neurons, no feedback loops.
Years later McCarthy admitted the terminology was also political: he wanted to escape cybernetics’ concentration on analog feedback (which he considered misguided) and to avoid accepting Wiener as a guru or arguing with him.
In July 1958 the US Navy demonstrated Frank Rosenblatt’s Perceptron, running on a five-ton, room-sized IBM 704. The New York Times reported the machine learning to distinguish left-side from right-side squares after fifty indifferent trials — a binary classifier. The architecture had three layers: sensory units storing ones and zeros, association units firing above a threshold, and response units producing ±1. Only one set of weights (the V values) was learnable; the sensory-to-association connections were fixed, and Rosenblatt couldn’t make them trainable. That was the perceptron’s Achilles heel. The Navy nonetheless imagined it would one day walk, talk, see, write, reproduce itself, and be conscious of its existence. Rosenblatt refused the “AI” label — his book said the program investigated the neurodynamic principles underlying natural intelligence — and named his field “neurodynamics,” much closer in spirit to Wiener.
Marvin Minsky, a Bronx high school classmate of Rosenblatt’s (one grade apart), co-published a 1969 book highlighting the limits of single-layer learned weights and doubting more layers would help. Neural network funding dried up, Rosenblatt died in a 1971 sailing accident on his 43rd birthday, and the 1973 British Lighthill review concluded the whole field had failed its grandiose aims. Both the UK government and the US military cut funding. The result: the first AI winter.
Money returned in the 1980s from industry, which bought expert systems encoding human expertise as handwritten rules — a natural fit for McCarthy’s symbolic approach. The neural network camp returned too: the 1986 book Parallel Distributed Processing, co-authored by a young Geoffrey Hinton, showed that multi-layer networks could learn far more than the perceptron, with backpropagation adjusting weights in every layer. The Achilles heel was gone; the movement became known as connectionism — knowledge living in connection strengths, not symbols.
By the late 1980s, expert systems proved expensive and brittle, the market collapsed, and the second AI winter arrived. “Artificial intelligence” carried most of the disillusionment; the New York Times noted that computer scientists avoided the term for fear of seeming “wild-eyed dreamers.” The work continued under other names: machine learning, knowledge-based systems, computational intelligence, informatics.
Hinton, Yann LeCun, and Yoshua Bengio — today’s Turing Award winners — weathered the winter on Canadian funding and rebranded their struggling field “deep learning” in the mid-2000s, referring to networks many layers deep. With enough data and compute, the field renaissance’d in the mid-2010s, and “AI” regained popularity — ironically now attached to the neural network camp that originally opposed McCarthy’s symbolic focus, culminating in the 2020s boom.
Which brings the story to 2026: the proposed rebrand to “super intelligence” is the first driven by success rather than failure. AGI seems (to many) achieved, and “artificial” frames the technology as against our nature — it versus us — while “superintelligence” sounds like a positive national achievement. If history is any guide, though, it won’t be the last name either.