In the second part of our summer series “AI before AI”, the broadcaster looks back at the early days of artificial intelligence. Click here to read Part I on the forgotten Soviet programme that developed early AI concepts.
Billions of dollars are now pouring into artificial intelligence, and AI start-ups have little trouble convincing investors to back them financially.
But back in 1955, the Rockefeller Foundation rejected a $13,500 request to fund a conference submitted by John McCarthy, a young American mathematician who, although promising, was still at the start of his career.
Read moreAnthropic to restore global access to most powerful AI models
Rockefeller Foundation administrators balked at approving so much money – the equivalent of more than $160,000 today – to organise a workshop for a field of research that did not yet exist. What was “artificial intelligence”, and why devote two summer months to it at New Hampshire's Dartmouth College?
Nevertheless, McCarthy was backed by some high-profile figures and had teamed up with Marvin Minsky, another young mathematics prodigy. The two rising stars managed to convince other heavyweights to join the project. Nathaniel Rochester, who invented the first commercial computer for IBM, was one of the co-signatories of the funding application. At the time, the few computers in existence resembled large cupboards overflowing with electrical wires and bore little resemblance to today's modern computers.
Read moreEU ‘tech sovereignty’: Uncoupling from the US and China?
McCarthy and Minsky also managed to secure the involvement of Claude Shannon, whose reputation as a mathematician was already well established following his seminal 1948 paper on the mathematical theory of communication. Shannon ranks among the names most frequently cited when discussing early AI pioneers alongside Alan Turing and John von Neumann, according to Philippe Mathieu, director of the Multi-Agent Systems and Behaviour team at the University of Lille.
They promised to explore topics at the conference including automatic computers, neural networks, the creation of a computer “programmed to use a language” and even the “self-improvement” of intelligent machines – which might today be described as “deep learning”.
The birth of AI
One man at the Rockefeller Foundation offered a ray of hope: Warren Weaver, director of the Division of Natural Sciences, had already approved funding for several landmark projects of the time, including in genetics, molecular engineering and agriculture. Weaver had also written an analysis of Shannon’s work.
“Weaver was a friend of Shannon's, which is why the application was addressed directly to him,” said Rudolf Seising, who specialises in the history of science and technology at the Deutsches Museum in Munich.
But although Weaver found the proposal interesting, he felt it was focused on how the brain works, and that brain research sounded more like medicine than mathematics, according to Seising. “He therefore forwarded it to Robert S. Morison, the director of medical research at the foundation,” he added.
Morison decided to award $7,500 to this biomathematics project, making the first AI investment in history.
One of your browser extensions seems to be blocking the video player from loading. To watch this content, you may need to disable it on this site.
The term “artificial intelligence” was still not yet a fait accompli. In a joint article written with Shannon, McCarthy suggested referring to “Intelligent Machines”, but “Shannon felt the term was much too big,” Seising said. They eventually settled on “Automata Studies”, which McCarthy found less promising.
“‘Artificial intelligence’ was a good term for raising money and to interest the public and policymakers. But it was not purely a marketing decision, as it also served to define a common vision; it was the starting point for a community of researchers on artificial intelligence,” said Hartmut Hirsch-Kreinsen, a sociologist who specialises in technology.
That was the aim of the Dartmouth Summer Research Project on Artificial Intelligence. The group of four – McCarthy, Minsky, Shannon and von Neumann – wanted to bring together mathematicians, philosophers and economists alike to work on the future of AI.
Read moreAI before AI: The forgotten Soviet programme (1/2)
Disappointing results?
In June 1956, the team took up residence at Dartmouth College apartments for five weeks. The true significance of the research programme is debated to this day: was this really the founding moment of AI, or were the results disappointing overall?
The conference itself yielded little of substance. Its organisers had hoped for an exchange of ideas that would in the near future enable them to demonstrate in concrete terms that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it”, as McCarthy wrote.
Watch moreOpen AI models go rogue, ecape and and hack online AI sharing hub
In reality, the various accounts of this historic conference show that most of the participants – the exact number of whom is not known – interacted with one another only very sporadically and were present for just a few days, sometimes simply to make use of the computers.
But the Dartmouth workshop went down in history, and the 50th anniversary of AI was celebrated in 2006.
And at least one breakthrough was unveiled at the conference: the “Logic Theorist”, considered to be the very first AI programme, was presented by its creators, Allen Newell and Herbert Simon, to the Dartmouth assembly.
Many of the workshop's participants went on to have brilliant careers, and several went on to win Turing Awards, the equivalent of the Nobel Prize for AI.
The conference even influenced the silver screen in the 1968 film, “2001: A Space Odyssey”. Director Stanley Kubrick consulted Minsky as he was developing his concept for the HAL 9000 – the hyper-rational, potentially deadly AI computer and centerpiece of the film's spaceship.
This article has been translated from the original in French.
