Second AI Winter

The second AI winter was another period of reduced funding and interest in artificial intelligence. It was triggered by the collapse of the specialized Lisp machine market and the realization that expert systems were expensive to build and maintain and were only applicable to a narrow range of problems.

The second AI winter was triggered by several factors. In the 1980s, several companies had built specialized computers (Lisp machines) to run AI programs written in the Lisp programming language. These machines were expensive and optimized for symbolic AI. By the late 1980s, general-purpose workstations from companies like Sun Microsystems had become powerful enough to run AI programs, and they were much cheaper than Lisp machines. The Lisp machine market collapsed, and companies like Symbolics and Lisp Machines Inc. went out of business.

While expert systems like XCON were successful, many others were not. They were expensive to develop, difficult to maintain, and brittle (i.e., they failed when faced with problems outside their narrow domain). The 'expert systems boom' of the 1980s turned into a bust as companies realized that expert systems were not the panacea they had hoped for. The technical limitations of rule-based systems became increasingly apparent as companies tried to scale them to more complex domains.

As with the first AI winter, the second was marked by a significant reduction in government and industry funding for AI research. The Strategic Computing Initiative, a major DARPA program that had funded AI research in the 1980s, was scaled back. Symbolics, once valued at over $100 million, went from profitability to bankruptcy. Apple canceled their AI research division, and many other companies followed suit.

The second AI winter had a similar effect to the first. It led to a further decline in AI research and a loss of confidence in the field. Many researchers left AI for other areas of computer science, and the term 'artificial intelligence' continued to be associated with hype and failure. Some researchers began avoiding the term, preferring 'informatics,' 'knowledge-based systems,' or 'machine learning' instead.

However, the second AI winter also led to a shift in focus towards more practical and tractable problems. Researchers began to focus on subfields of AI, such as machine learning and computer vision, rather than trying to build general-purpose intelligent systems. This shift laid the groundwork for the successes of the 2000s and 2010s. The second AI winter reinforced the lessons of the first, highlighting the importance of realistic expectations and the need for a solid theoretical foundation for AI research.