AI Just Celebrated Its 70th Birthday. So Why Did I Just Find Out About It?

For many people, artificial intelligence seemed to arrive in November 2022.

That was when ChatGPT suddenly gave ordinary people a way to have a conversation with a machine that could write, explain, summarize, translate, brainstorm and answer questions in surprisingly human-like language.

Artificial intelligence (AI) was no longer something happening inside a laboratory, a technology company or a science-fiction movie. Millions of people could use it themselves.

It felt new.

It wasn't.

The field is generally dated to the Dartmouth Summer Research Project on Artificial Intelligence in 1956, where the term was introduced and researchers gathered around the proposition that aspects of learning and intelligence could be described precisely enough for a machine to simulate them.¹

But even 1956 was not the beginning of the idea.

Before AI Had a Name

Six years before the Dartmouth meeting, British mathematician Alan Turing published a paper titled Computing Machinery and Intelligence.

It opened with an extraordinary question for 1950:

Turing did not have a laptop. There was no internet, no smartphone and certainly no ChatGPT.

In 1950, computers were still enormous, rare machines largely confined to government, military, university and a handful of corporate settings. Ordinary Americans had virtually no direct contact with them.

Yet Turing was already seriously examining whether machines might display behavior that people would regard as intelligent. He proposed what became known as the Turing Test: could a machine converse well enough that a human judge could not reliably distinguish it from another human?²

He went further. Turing wrote that machines might eventually compete with people in intellectual fields.

That was written in 1950.

Not 2020.

Not 2010.

1950.

That does not mean Turing predicted modern large language models, autonomous vehicles, humanoid robots or the economic consequences of any of them. He did not.

But the central possibility was already on the table: machines might someday perform work that required human intelligence.

Then Somebody Gave the Idea a Name

In the summer of 1956, a small group of scientists gathered at Dartmouth College in New Hampshire for the Dartmouth Summer Research Project on Artificial Intelligence.

John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon had proposed the project. The proposal used a new term: artificial intelligence.

Their premise was audacious. They proposed to investigate whether learning and other features of intelligence could, in principle, be described so precisely that a machine could simulate them.¹

Dartmouth now describes that meeting as the birth of artificial intelligence as a field.

So if AI needs an official birthday, 1956 is a very good choice.

And that means AI is not a new technology in the way most of us experience it.

The World Around AI Was Changing Too

The timing matters.

The United States had emerged from World War II with enormous industrial capacity, powerful universities, a rapidly expanding scientific establishment and a new understanding that scientific leadership could determine economic and military power.

After the war, the federal government greatly expanded its role in supporting scientific research. The National Science Foundation (NSF) was established in 1950 as part of that broader postwar research system.³

At the same time, the baby boom was expanding the population. Colleges and universities were growing. Suburbs were spreading. Consumer prosperity was rising. The Cold War was intensifying.

AI was not developing in isolation.

It was developing inside a country increasingly willing to spend money, train scientists and engineers, build research institutions and treat technological leadership as a national priority.

Then Sputnik Changed the Mood

On October 4, 1957, the Soviet Union launched Sputnik, the first artificial satellite to orbit Earth.

The psychological effect in the United States was enormous.

If the Soviet Union could put a satellite into orbit, it also demonstrated technological capabilities with obvious military implications. American political leaders began questioning whether the country was falling behind in science, engineering and education.

The response was not subtle.

Federal science spending increased. Science and mathematics education received new attention. The National Aeronautics and Space Administration (NASA) was created in 1958. The National Science Foundation's appropriation rose sharply in the period after Sputnik.³

And on February 7, 1958, the Department of Defense established the Advanced Research Projects Agency (ARPA).⁴

ARPA was later renamed the Defense Advanced Research Projects Agency (DARPA).

The agency was created in the aftermath of Sputnik to help the United States avoid another technological surprise.

That mission would become enormously important to the development of modern computing.

The Cold War Helped Build the Infrastructure

In 1962, ARPA created the Information Processing Techniques Office (IPTO).

Its research portfolio included time-sharing, computer graphics, networking, advanced processor design, parallel processing and artificial intelligence.⁵

Those subjects may sound familiar.

They are pieces of the technological world we now take for granted.

The point is not that the federal government sat down in 1962 with a blueprint for ChatGPT.

It didn't.

The point is that artificial intelligence and the computing infrastructure around it became part of a long-running research effort involving universities, private companies and government agencies.

The technology was being built piece by piece.

By the 1960s, They Were Already Building Intelligent Robots

AI was not limited to software.

In the 1960s, researchers at the Stanford Research Institute worked on a machine later known as Shakey the Robot. The Advanced Research Projects Agency began supporting the project in 1966.

DARPA describes Shakey as the first mobile robot with enough artificial intelligence to navigate rooms autonomously.⁶

By modern standards, Shakey was primitive.

But look at the problem researchers were already trying to solve: combine perception, reasoning, computing and physical movement so a machine could act in the real world without a human directing every movement.

That sounds remarkably similar to the challenge facing today's robotics companies.

Progress Was Not a Straight Line

If all this was happening so long ago, why didn't most of us hear much about AI for decades?

Because progress was uneven.

Early researchers made ambitious predictions. Computers were slow. Memory was expensive. Data was limited. Many systems worked in carefully controlled demonstrations but failed when the real world became messy.

Enthusiasm rose and fell. Funding rose and fell. Periods of disappointment became known as AI winters.

This history matters because repeated disappointment may have taught society the wrong lesson.

AI kept promising more than it could deliver—until, increasingly, it didn't.

Underneath the disappointment, the supporting technologies kept improving.

Processors became faster. Storage became cheaper. Networks connected the world. The internet created enormous quantities of digital data. Machine learning improved. Cloud computing made huge amounts of computing power accessible. Specialized processors made it practical to train increasingly large neural networks.

The individual pieces began to compound.

Then the Pieces Started Coming Together

In 2017, researchers introduced a neural-network architecture called the Transformer. Instead of processing language in the older sequential ways, the architecture relied heavily on a mechanism called attention and could be trained efficiently in parallel.⁷

The Transformer became one of the foundations for the large language models that followed.

Five years later, on November 30, 2022, OpenAI released ChatGPT as a public research preview.⁸

That was the moment the long history of AI suddenly became personal for millions of people.

For decades, artificial intelligence had largely been something researchers, engineers, companies and governments worked on.

Now you could type a question into a box.

And it answered.

We Spent Decades Asking Whether Machines Could Think

Look at the questions researchers pursued over those decades.

Can a machine reason?

Can it learn?

Can it understand language?

Can it see?

Can it navigate?

Can it plan?

Can it make decisions?

Can it perform tasks that previously required a human?

These were legitimate scientific questions, and answering them produced extraordinary technology.

But there is another question.

That question was not completely ignored. Economists, technologists and government officials have worried about automation and technological unemployment at different points in history.

That distinction is important. It would be wrong to say nobody thought about the problem.

The harder question is whether our economic planning ever became remotely proportional to our technological ambition.

We built institutions to fund science.

We built research laboratories.

We trained generations of engineers and computer scientists.

We built semiconductor industries, global networks, data centers and enormous technology companies.

We spent decades making machines more capable.

Where was the comparable long-term effort to understand what happens if those machines eventually reduce the amount of human labor the economy needs?

This Is Not an Argument That Someone Should Have Predicted ChatGPT in 1956

Hindsight can make history look easier than it was.

No policymaker in 1956 could reasonably have predicted cloud computing, smartphones, modern graphics processors, large language models or the exact capabilities of ChatGPT.

That is not the argument.

The argument is narrower.

The possibility that machines might perform intellectual work has been discussed openly for generations. The possibility that automation might displace human labor is even older.

As machine capability improved, the obligation to examine the consequences should have grown with it.

The Same Competitive Logic Is Still With Us

There is another piece of this history that matters now.

Sputnik helped convince the United States that technological leadership was a race.

Falling behind could carry economic and military consequences.

That logic helped produce extraordinary scientific progress.

It also sounds familiar today.

Governments and companies increasingly discuss artificial intelligence in terms of leadership, dominance and the danger that a competitor or another country might get there first.

The race helped create the technology.

The race may now make the technology harder to govern.

That deserves its own examination. Government accountability, national competition and the question of whether any country can meaningfully slow down alone are too important to squeeze into a history lesson.

The Technology Was Slow. The Realization Was Fast.

Alan Turing was asking whether machines could think in 1950.

Artificial intelligence got its name in 1956.

Sputnik shocked the United States in 1957.

The Advanced Research Projects Agency was created in 1958.

Its Information Processing Techniques Office was funding artificial intelligence and advanced computing research in the 1960s.

Researchers were building autonomous robots in the same decade.

Machine learning developed over the decades that followed.

The Transformer arrived in 2017.

ChatGPT arrived in public in 2022.

And suddenly millions of us asked: Where did this come from?

The answer is that it came from a very long road.

We just weren't all standing beside it.

AI did not come out of nowhere.

What may have come out of nowhere is our sudden realization that it might actually work.