AI winters: why funding and interest in AI research collapsed
AI winters are periods when funding and interest in artificial intelligence research collapse. They stem from overoptimistic promises that fail to materialize, compounded by critical reports like the Lighthill survey. The field has repeatedly revived after each winter through new approaches and more realistic expectations.
The Optimistic Beginnings
Humans have dreamed of creating thinking machines from ancient times, a long-standing ambition reflected in folklore, as noted in IBM's history of artificial intelligence.
The field formally began at a 1956 conference at Dartmouth College, where the term 'artificial intelligence' was coined, according to the Stanford Encyclopedia of Philosophy. An early program, Logic Theorist, proved theorems and introduced heuristics and list processing, as documented in the Computer History Museum timeline.
The Gap Between Promise and Reality
Despite these early successes, machines that think like people have remained elusive. The Computer History Museum revolution states that getting machines to actually think like people has proved stubbornly elusive.
Early demonstrations often worked only in narrow settings. A robot squirrel called Squee, for example, could hunt for tennis balls but only in a very dark room, as described in the Computer History Museum timeline.
Critical Reports and Funding Cuts
Governments and funding agencies began to question the return on their investment. An influential assessment was the Lighthill report, a survey of artificial intelligence, which is preserved by Chilton Computing.
Such critical reviews, combined with limited commercial results, caused research funding to shrink. This period is known as an AI winter, when interest waned and many projects were cancelled.
The Cycle of Hype and Recovery
AI winters do not last forever. The field has repeatedly rebounded with new techniques and more realistic goals. The textbook by Stuart Russell and Peter Norvig, adopted by over 1500 schools, reflects the field's persistence and maturity, as stated on the Berkeley AI textbook page.
Understanding this cycle helps researchers manage expectations and avoid repeating past mistakes.
| Demonstration | Achievement | Limitation |
|---|---|---|
| Logic Theorist | Proved theorems from Principia Mathematica | Limited to propositional calculus |
| El Ajedrecista | Played a simple chess endgame | Only handled a king and rook versus king |
| Squee | Hunted for tennis balls using light sensors | Worked well only in a very dark room |
Frequently asked questions
What is an AI winter?
An AI winter is a period when funding and interest in artificial intelligence research decline sharply. It typically follows overpromising and underdelivering in the field.
What caused the first AI winter?
The first AI winter was largely due to unmet expectations after early successes. Critical assessments, such as the Lighthill report, questioned the field's progress, leading to funding cuts.
How long do AI winters last?
AI winters can last several years, but the exact duration varies. The cycle of hype and disappointment has occurred multiple times in AI history.
Can AI winters happen again?
AI winters can recur if researchers make excessive claims that do not materialize. The pattern of enthusiasm and retreat is visible throughout the history of the field.