In our last post, we stated, “When the early founders of AI set extremely high expectations, they invited scrutiny. With the passing years, it became obvious that the reality of artificial intelligence fell short of their predictions. In 1974, funding for AI research began to dry up, both in the United States and Britain, which led to a period called the ‘AI Winter,’ and optimism regarding AI turned to skepticism. The first AI Winter lasted until the early 1980s.”
In early the 1980s, researchers in AI began to abandon the monumental task of developing strong AI and began to focus on expert systems. An expert system, in this context, is a computer system that emulates the decision-making ability of a human expert. This meant the computer software allowed the machine to “think” equivalently to an expert in a specific field, like chess for example. Expert systems became a highly successful development path for AI. By the mid-1980s, the funding faucet for AI research was flowing at more than a billion dollars per year.
Unfortunately, the funding faucet began to run dry again by 1987, starting with the failure of the Lisp machine market that same year. MIT AI lab programmers Richard Greenblatt and Thomas Knight, who formed the company Lisp Machines Inc., developed the Lisp machine in 1973. The Lisp machine was the first commercial, single-user, high-end microcomputer, which used Lisp programming (a specific high-level programming language) to tackle specific technical applications.
Lisp machines pioneered many commonplace technologies, including laser printing, windowing systems and high-resolution bit-mapped graphics, to name a few. However, the market reception for these machines was dismal, with only about seven thousand units sold by 1988, at about $70,000 per machine. In addition, the company, Lisp Machines Inc., suffered from severe internal politics regarding how to improve its market position. This internal strife caused divisions in the company. To make matters worse, cheaper desktop PCs soon were able to run Lisp programs even faster than Lisp machines. Most companies that produced Lisp machines went out of business by 1990, which led to a second and longer-lasting AI Winter.
If you are getting the impression that being an AI researcher from the 1960s through the late 1990s was akin to riding a roller coaster, your impression is correct. Life for AI researchers during that timeframe was a feast or famine-type existence.
While AI as a field of research experienced funding surges and recessions, the infrastructure that ultimately fuels AI, integrated circuits, and computer software, continued to follow Moore’s law. In the next post, we’ll discuss Moore’s law and its role in ending the second AI Winter.