AIThe Story of AI
Chapter Two · 1974–1993

The Frozen Years

When AI got put on ice — twice. Big promises, bigger disappointments, and the chill that followed.

Imagine you're super excited about a new toy. You tell all your friends it can do amazing things, but when they try it, the toy doesn't work as well as you promised. Your friends feel tricked, stop playing with it, and stop giving you money to make more toys. That's basically what happened to AI scientists, not just once, but TWICE! These chilly times are called "AI Winters."

Just like a real winter freezes plants and stops them from growing, an AI Winter froze research. Money dried up, smart scientists left the field, and people stopped believing AI would ever be useful. Companies went bankrupt, and labs closed their doors. Brrr!

The reason AI Winters happened was simple: scientists made HUGE promises they couldn't keep. In the 1960s, famous researchers like Marvin Minsky said machines would be as smart as humans within 20 years. When that didn't happen, governments and businesses got angry and pulled their funding, like turning off the heat in winter.

§

Two Winters, Two Decades

1973
The Lighthill Report
Sir James Lighthill writes a report in Britain saying AI had failed to deliver on its promises. Britain cuts almost all AI funding.
1974
DARPA Pulls Funding
The U.S. military's research agency slashes AI funding because speech recognition projects flopped. The First AI Winter begins.
1980
XCON Saves Millions
Digital Equipment Corporation starts using XCON, an expert system that saves them $40 million a year. AI gets exciting again!
1982
Japan's Big Bet
Japan launches the Fifth Generation Computer Project, spending $850 million to build super-smart machines and beat America.
1987
LISP Machine Crash
The market for special "LISP machines" (computers built just for AI) collapses. The Second AI Winter begins.
1992
Japan's Project Ends
Japan's Fifth Generation Project ends without reaching its big goals.
1993
Another Funding Cut
DARPA again cuts AI funding, deepening the freeze. AI wouldn't fully thaw until the 2000s!

What Was an "Expert System"?

An Expert System was an early type of AI from the 1980s that tried to copy how a human expert thinks. Programmers would interview a real expert (like a doctor) and write down ALL their rules as "if-then" statements. For example: "IF the patient has a fever AND a sore throat, THEN check for strep."

Two famous expert systems were:

The problem? Expert systems were like a robot that only knew ONE recipe book. If you asked them anything outside their rules, they were totally lost. They couldn't learn new things, and updating their thousands of rules became a nightmare. That's why they eventually froze out in the Second AI Winter.

§

Three Lessons From the Cold

01

Don't Over-Promise

Scientists learned to be honest about what AI could and couldn't do. Big bragging leads to big disappointment.

02

Computers Need Power

Early AI failed partly because computers in the 1970s and 80s were way too slow and small. Today's AI works because computers are MILLIONS of times faster.

03

Real Life is Messy

Expert systems failed because they couldn't handle surprises or learn new things on their own. Modern AI uses huge amounts of data to learn, which works much better.

An icy season. Twice in twenty years, the AI field froze over — funding stopped, labs shut, and the dream had to wait.

The Two AI Winters

Excitement and funding dropped sharply — twice — before the modern AI boom.

WINTER 1 WINTER 2 19561974198519932025 FUNDING / HYPE
Each winter began with broken promises — and ended when computers finally caught up to the dream.
The reason AI Winters happened was simple: scientists made HUGE promises they couldn't keep.
— A lesson the field would have to learn twice