The Relentlessness and Multiple Lives of AI
A very short note on the history of the term 'artificial intelligence'.
Posted on: February 14, 2025
It’s hard not to be both perplexed and impressed by the power the term “Artificial Intelligence” (AI) holds on human imagination and the phoenix-like acrobatics it displayed again and again ever since the 1950s, when it was coined. Michael Jordan grumbled about the term rather accurately in a recent talk, and in slightly more detail in this 2019 essay. This later essay makes for an interesting read, even if just because of how the world went in a completely different direction. Jordan’s frustration is likely shared by many, but the collective intellectual frustration with AI seems to do little to dislodge its hold on our life. On the contrary, following on the “all publicity is good publicity”, AI is just getting bigger.
A while ago I thought to write a bit about the term AI. But the constant flood that is our current reality kept its relentless flow, and I just didn’t get to it. I did have a pretty interesting email thread about this back then, including with David Mimno, Anna Shechtman, and Gili Vidan. I will be honest that I looked for something interesting, maybe even nefarious in its introduction. I didn’t find it, unless one considers conventional academic marketing and territorial marking to be particularly exciting. There wasn’t even a genius grant writing spin.
Stephanie Dick penned a wonderful essay on the topic. It’s a crisp discussion of the evolution of what AI means, and the remarkable survival (and revival) of the term itself, even when its content morphed dramatically over the years. The part about expert systems is particularly intriguing. Yes, expert systems failed. They are largely treated as a lesson in vacuous hype. But, the expert system approach enthroned knowledge in contrast to logical processes (of the earlier methods). Today, this seems almost prophetic, because of how the entanglement of knowledge, linguistic skill, and “reasoning”(?) seems so fundamental to the success of contemporary LLMs. Another interesting tidbit that Dick accurately points out is how AI took over the ML community not through its willing adoption by ML researchers and practitioners, but from the outside – through funding initiatives and externalized communication. There’s much more to be discussed regarding the impact of how AI ate ML, because this is not just a terminology quibble, it created research pressures that significantly shape the field.
I am including two quotes that I noted to myself along the way, both focused on the origin story of the term AI. The connection to the little attention McCarthy’s book about automata received makes for a particularly interesting lesson about communication.
Margaret Boden’s 2-volume history of cognitive science says this about the Dartmouth project:
The two younger men, Minsky and McCarthy, would later be great names in cognitive science, and especially in AI (Chapters 7.i.e, 10, 12.iii). Indeed, McCarthy was already developing the ideas for his seminal paper on ‘Programs with Common Sense’ (1959), which he’d deliver in London two years later (see below, and 10.i.f). And Minsky distributed drafts of his influential ‘Steps Toward Artificial Intelligence’ to various Dartmouth visitors (Minsky 1961b: 10.i.g). Dartmouth turned out to be the naming-party for AI. However, in contrast with Christopher Langton’s 1987 party for A-Life (15.x.a), the new name wasn’t welcomed by every guest there. The event had been announced as ‘’The Dartmouth Summer Research Project on Artificial Intelligence’’. But there was no little disagreement at the meeting about which term should be adopted for the field: cybernetics, automata studies, complex information processing (the Newell–Simon favourite), machine intelligence, or (McCarthy’s suggestion) artificial intelligence. (The phrase knowledge-based systems was dreamt up a quarter-century later, in the mistaken belief that this raises fewer philosophical problems: 11.v.c.) McCarthy had insisted on using “artificial intelligence” in the meeting’s official label partly because the book he’d just co- edited with Shannon, and which the senior author had insisted (against McCarthy’s pleading) be called Automata Studies (1956), had attracted only highly mathematical papers.
Pamela McCorduck’s Machines Who Think (based substantially on interviews with the scientists who developed the field) says:
A dispute occurred over what the new field should be named. Although the conference was officially called The Dartmouth Summer Research Project on Artificial Intelligence, many attendees balked at that term, invented by McCarthy. “I won’t swear that I hadn’t seen it before,” he recalls, “but artificial intelligence wasn’t a prominent phrase particularly. Someone may have used it in a paper or a conversation or something like that, but there were many other words that were current at the time. The Dartmouth Conference made that phrase dominate the others.” McCarthy had not chosen accidentally to call the conference by that name. During the summer that he’d worked for Shannon, when they had put together the book of collected papers on some of the subjects that McCarthy was interested in, to be called Automata Studies (Shannon and McCarthy, 1956), McCarthy wanted to use a term different from automata studies for the papers he hoped to get for the book, but Shannon objected that any other phrase was simply too flashy, that the theory of automata would be sober and scientific. McCarthy went along with that, thinking that it probably didn’t make that much difference. “The original idea was that Claude Shannon would be the name to attract good papers, and I would do the work, but it ended up that he did the work too,” McCarthy now recalls. “One of the reasons why he did all the work was I was unenthusiastic about the papers.” Most of the papers they received for the book were in fact about automata theory in the narrowest sense, that is, mathematical principles underlying the operation of electromechanical systems, and not about the relation of language to intelligence, or the ability of machines to play games, or any of the other topics McCarthy was becoming more and more fascinated by. The Shannon and McCarthy volume contained many significant papers and is still a reference in automata studies, but McCarthy felt he’d learned his lesson. Thus in the proposal and again at the conference, he argued strongly for the term artificial intelligence to distinguish it from automata theory, though to this day there are persons in the field, including some of the original participants in the Dartmouth Conference, who object to it.