Erik Sandewall was born in 1945 and first heard about programming as a young teenager.
“I enjoyed solving crosswords and I read an article in a crossword magazine about how to program a computer. It sounded really exciting”, he remembers.
Computers and programming were to become a large part of Sandewall’s life. He was appointed Sweden’s first professor of computer science in 1975 at Linköping University.
Today, his stature as visionary in many fields is clear: he was, among other things, one of the first in Sweden to conduct research into artificial intelligence.
After upper secondary school, Sandewall studied maths at Uppsala, but found that his studies didn’t fully occupy his time. A friend suggested that he read numerical analysis, a subject that included a form of programming closely related to artificial intelligence.
Interesting and disappointing period at Stanford
Sandewall found the subject fascinating, and decided to direct his studies towards AI. This was a new field, and knowledge was severely limited in Sweden.
“My professor in Uppsala said that if I wanted to get into this field, I would have to travel somewhere where they mastered it.”
Such a place was Stanford, one of the most prominent universities in the world in AI. Sandewall’s supervisor was John McCarthy, the man who coined the expression “artificial intelligence”.
“At the time, the field was dominated by logic-based artificial intelligence and theories that emphasised problem-solving. Such AI is closely related to psychology, and language processing. These traditions are still highly active in the AI field today.”
Sandewall enrolled as a doctoral student at Stanford. He describes the period as both interesting and disappointing.
“John McCarthy held extremely stimulating lectures, and I learnt a lot. But he was a hopeless supervisor.”
It seems that this legend of AI gave priority to his own research above that of his research students, so Sandewall decided to return to Uppsala. Here, he published his first articles in artificial intelligence. One of these presented an extension to the Lisp programming language for AI, while another described how to search through a tree of alternatives.
In Uppsala he led a research group working in computer science and artificial intelligence, but was offered a professorship in computer science in Linköping. He collected the group together and travelled.
“The campus looked rather different then. Only the A, B and C Buildings had been built, and they were surrounded by fields. We felt strongly that we were pioneers.”
Artificial intelligence was often viewed with scepticism in the 1970s.
“Principally by the media and pundits. They resisted the idea of using words associated with human thought processes, such as ‘intelligence’ and ‘learning’. Their views were received well in circles critical to technology.
This meant that Sandewall and his group were careful when using the concept of ‘artificial intelligence’.”
“We talked internally about artificial intelligence, but to outsiders we talked more about the applications of what we were doing, also in cases where we otherwise used the term “artificial intelligence”.
Sandewall’s ideas are still alive
AI was one of the reasons for the unique design of the computer science programme, which started in the beginning of the 1980s. Sandewall and the others removed subjects taken from the natural sciences and replaced them with humanity-based subjects, such as psychology and linguistics.
“We considered this change to be necessary for artificial intelligence, and we had a more pressing purpose, namely the development of systems for person-machine dialogue. The Swedish Higher Education Authority of the day was, however, not interested in these ideas, and gave the programme very little support."
The authority has now been absorbed by another, but Sandewall’s ideas for a broad education in AI are still alive and well in several of LiU’s educational programmes.
It is not just scepticism that has been the response to AI: there have been periods of great enthusiasm. We are currently experiencing such a period.
“Unfortunately, this often means that the general public is given an inaccurate picture of AI through the media. Promises are made that can’t be kept.”
This is because many of the fundamental AI problems remain, despite advances in machine learning that have been made in recent years.
“There’s a huge difference between doing something that works some of the time, and something that works, in principle, every time. It’s easy to think that this is a small step, but it’s not. It’s a huge step.”
Translated by George Farrants