AI is already inescapable: From your community, where data centers may be going up, to your stock portfolio, which is dependent on it; from classrooms, where it’s preparing essays and presentations, to offices, where it’s being used to eliminate entire departments.
Of course, we all already know that it’s supposed to be the most society-changing technology since fire.
But what is artificial intelligence really? Where does it come from? What risks and opportunities does it truly offer?
While writing and editing stories for Roca over the past year, I‘ve repeatedly found myself diving deep into the AI ecosystem, trying to understand what exactly it is and what it will mean. While I’m yet to have any answers about the future, I have found a number of answers about the past: Where AI originated, how it developed, who got it to where it is today.
It’s important and interesting information, and we want to share it here. So, starting today, we’ll be writing a weekly story about the history of AI.
This first installment begins not in 1955, when the term “artificial intelligence” was coined, nor in 1950, when Alan Turing pioneered his machine intelligence test, but in 1943, with the publication of an abstract paper that, in hindsight, is shaping the world we live in today.
That paper – “A Logical Calculus of the Ideas Immanent in Nervous Activity” – was authored by neurophysiologist Warren McCulloch and the logician Walter Pitts. The duo’s achievement was this: To model the neuron as a simple “threshold” device that fires only when the sum of its inputs exceeds a certain level.
Such networks, they showed, could implement any logical function – think Excel descriptors like AND, OR, NOT – that could be expressed through propositional calculus. The paper’s biological accuracy was limited (real neurons are far more complex), yet its impact was profound: “Thinking,” or at least logical reasoning, could be reduced to networks of remarkably simple elements.
In other words: McCulloch and Pitts had shown that brains could be logical machines, and that logic could be mechanized.
As McCulloch and Pitts theorized about the biology of thinking, engineers were building the machinery that could eventually make it real.
The demands of World War II had produced a slew of new technologies, among them the first large-scale electronic computers. These machines were programmed the hard way: Technicians physically rewired plugboards and flipped dozens of switches for every new problem. The computer’s “instructions” were literally part of its hardware.
But in June 1945, John von Neumann proposed a drastic simplification in his “First Draft of a Report on the EDVAC.” Both the numbers a computer worked on and the instructions telling it what to do should live together in the same high-speed memory. In other words, programs would no longer be fixed wiring; they would become flexible data – something that could be read, written, changed, or even created by the machine itself.
This stored-program idea transformed what a computer could be. A machine could now hold a set of rules inside its memory and treat those rules like any other information. It could follow one procedure today and an entirely different one tomorrow simply by loading new instructions. Far more than clever engineering, this was the essential precondition for artificial intelligence: To learn and reason, a system had to be able to represent, store, and dynamically revise its own processes.
Two other ideas laid AI’s foundation in 1948.
One came from Claude Shannon at Bell Laboratories, who published “A Mathematical Theory of Communication.” He reduced communication to something a machine could handle: “The fundamental problem of communication is that of reproducing at one point either exactly or approximately a message selected at another point.”
In other words, transmitting a message from one place to another.
This was groundbreaking because it turned information into a quantifiable commodity – measurable in bits, encodable, and transmittable. Shannon’s framework gave engineers the mathematics they needed to handle symbols, uncertainty, and representation inside an information-processing device.
That same year, Norbert Wiener published “Cybernetics: Or Control and Communication in the Animal and the Machine.” Drawing on wartime work in anti-aircraft gunnery and discussions with many of the same thinkers, Wiener coined the term “cybernetics” to describe the study of control and communication in both animals and machines. He emphasized feedback loops, circular processes in which a system’s outputs continuously shape its future inputs.
Wiener’s argument was simple: Organisms and machines both run on information. They sense, process, and act to stay stable in a disordered world. Information, he asserted, was neither matter nor energy but the currency of control itself.
By the close of the 1940s, a small network of scientists had reframed what information was and how it could be processed, stored, and conveyed. It was no longer fringe to view the mind as a predictable, logical device. Brains could be understood through logic and networks, and machines could, in principle, embody the very rules of thought.
But this research was still young, and big questions remained. Chief among them: How do we test whether something is actually “thinking”? In 1950, Alan Turing would give an answer.
Editor’s Note
Thanks for reading. We hope you found this interesting and learned something new. As always, if you have thoughts on today’s story, send them here or leave a comment below.
And, as always, find our latest articles below:
We’ll see you back here tomorrow.
—Max and Max
Machines think isn't new anymore when calculators were invented from simple to scientific and still one can test it if someone has one!! You observe the time lag a calculator needs for more complicated calculations, especially in scientific one but the time diminishes so much so that it could be noticed and it's instant when it's a computer to see the lag time anymore like our brains do in most responses but still brains have the time lag for more complex problems they have to encounter sometimes but for mechanized process of the Al's but we now know Als could outsmart us in near future!!
This will be a cool series