AI Winter
During the early 1970s, interest in Artificial Intelligence (AI) started to decrease. This period, which lasted through to the 1980s, is often referred to as the AI Winter.
The advancements during AI Winter were mostly academic in nature. The computer systems were very limited at that time. DEC PDP-11/45 used commonly for AI research only had the ability to increase its memory to 128K. Corporate world focussed on FORTRAN programming language since Lisp language was not ideal for computer systems.
Another reason for the waning enthusiasm was the complexity of concepts involved in understanding intelligence and reasoning. For example, disambiguation, which refers to a situation when one word has more than one meaning.
In 1973, Professor Sir James Lighthill, in his report, downplayed the developments made in AI up until that time. He argued that the discoveries made far less impact than what had been promised.
Finally, the economic situation in the 1970s was also not conducive for the development of AI. Inflation, slow growth, oil crisis, and disruptions in supplies were very common during this period. Because of these reasons , the US government also reduced its research spending on AI. As far as the US government was concerned, a programme that can play chess and solve mathematical theorems was simply not a priority.
In 1973, Professor Sir James Lighthill, in his report, downplayed the developments made in AI up until that time. He argued that the discoveries made far less impact than what had been promised. The major issue stated by him was ‘combinatorial explosion’ – a problem where the models get too complicated and difficult. He did not believe that computers would be able to play chess or recognise images.
This report was controversial and led to a public debate which was broadcasted by the BBC. The debate featured Sir James Lighthill against Donald Michie, Richard Gregory, and John McCarthy. Even though Lighthill had valid points, he however underestimated the power of weak AI.
As the AI Winter took hold, many researchers were forced to change their career paths. Those who continued to work in the field began to refer to their work using other terms like machine learning, pattern recognitions, and informatics so that they were not perceived as people working in an out of fashion area of research and development.
Rise and fall of expert systems
During the AI Winter, there were still some innovations taking place in the field of AI. One of them was backpropagation, which was essential in assigning weight to neural networks. Another innovation during this period was the development of Recurrent Neural Network (RNN), which allows connections through input and output layers.
The growth of PCs and minicomputers in the 1980s and the 1990s paved the way for the emergence of expert systems. Based on the concept of Minsky’s symbolic logic, expert systems were developed by domain experts within in a particular field. Even though there were expert systems that dated back to the mid-1960s, their commercial use became possible only in the 1980s.
IBM used an expert system in the development of its Deep Blue computer. In 1996, Deep Blue beat the chess grandmaster Gary Kasparov in one of six matches played between them.
eXpert CONfigurer (XCON) launched, in 1980, is an example of an expert system. It was developed by John McDermott at Carnegie Mellon University. XCON can be seen as the first recommendation engine. From its launch, it helped Digital Equipment Corporation (DEC) to save a lot of money for its line of VAX computers. The success of XCON drew the interests of other companies, and it turned expert systems into a billion-dollar industry.
The Japanese government tried to seize the opportunity by investing billions to bolster its native market. However, the results were disappointing, and most of the innovations in this area continued to come from the US.
IBM used an expert system in the development of its Deep Blue computer. In 1996, Deep Blue beat the chess grandmaster Gary Kasparov in one of six matches played between them. IBM started its development from 1985, and it had the capability to process 200 million positions per second.
Expert systems also had problems. It was often difficult to apply across different categories. As it became larger, it became difficult to manage and feed data into it. This resulted in recurrent errors. The testing of the machines weres also a complex process.
Expert systems did not evolve over time. Constant updates to the underlying models became very expensive and complex, and, as a result, by the late 1980s, expert systems lost its place in the business world. This resulted in another AI Winter, which lasted till 1993.
Now put on your thinking hats and think about the following questions for a couple of minutes.
As a teacher, how do you describe the term "AI Winter" to your students?
Can you think of the reasons for the AI Winter?
What do you think of Professor Sir James Lighthill's views on AI?
Write down your thoughts and discuss them with your students, children and your colleagues. Listen to their views and compare them with your own. As you listen to others, note how similar or different your views are to others’.
Thank you for listening. Subscribe to The Scando Review on thescandoreview.com.
Happy Teaching!
Robo-teachers? Fundamentals of AI for educators