The Second AI Winter and the Lisp Machine Collapse

The second artificial intelligence winter, spanning from the late 1980s to the early 1990s, was a period of severe funding cuts and widespread disillusionment in AI research. This downturn was directly triggered by the sudden collapse of the specialized Lisp machine market in 1987, the severe practical limitations of early expert systems, and the cancellation of major government funding initiatives like DARPA's Strategic Computing Initiative. Together, these factors popped the commercial bubble that had inflated around AI throughout the early 1980s, leading to a multi-year freeze in both venture capital and academic interest.

During the early 1980s, AI experienced a massive commercial boom driven by "expert systems"—programs designed to mimic the decision-making abilities of human experts. These systems were predominantly written in the Lisp programming language. To run these complex programs efficiently, companies like Symbolics, Lisp Machines Inc., and Texas Instruments built specialized, high-priced computers known as Lisp machines. These workstations cost upwards of $100,000 each and formed the hardware backbone of the AI industry.

By 1987, the hardware landscape changed dramatically, rendering Lisp machines obsolete. Mainstream computer manufacturers like Sun Microsystems and Apple began producing desktop workstations and personal computers using general-purpose microprocessors. These new machines were a fraction of the cost of Lisp machines and, thanks to rapid advancements in chip design, could run Lisp software just as fast. Furthermore, software companies developed tools to translate AI programs into standard programming languages like C. Practically overnight, consumers had no reason to buy expensive, specialized AI hardware. The half-billion-dollar Lisp machine industry vanished, bankrupting the leading AI hardware companies and signaling to investors that the AI sector was unstable.

Simultaneously, the expert systems that drove the market began failing in real-world applications. While they worked well in narrow, controlled scenarios, they proved to be brittle and lacked common sense. When presented with unexpected inputs, they produced bizarre errors. Additionally, maintaining them became a financial nightmare. As knowledge expanded, updating the thousands of hard-coded rules within an expert system caused conflicting logic, making large systems too expensive and complex to maintain. Iconic successes, such as the XCON system used by Digital Equipment Corporation, ended up costing millions of dollars a year just to keep operational.

The final blow came from government funding agencies. The Defense Advanced Research Projects Agency (DARPA) in the United States had poured hundreds of millions of dollars into the Strategic Computing Initiative (SCI) with hopes of creating intelligent military hardware. Similarly, Japan had heavily funded its Fifth Generation Computer Systems project. By the late 1980s, new leadership at DARPA concluded that AI was not delivering on its promises. DARPA drastically cut AI funding, pivoting resources toward projects that showed immediate, practical results.

The combination of the Lisp machine market crash, the unmaintainability of expert systems, and the withdrawal of government grants created a domino effect. Venture capitalists fled the market, hundreds of AI startups went bankrupt, and "artificial intelligence" became a tarnished buzzword in the tech industry, cementing the second AI winter.