Lighthill report

The Lighthill report, officially called Artificial Intelligence: A General Survey, was written by James Lighthill and published in 1973. James Lighthill created this report for the British Science Research Council to look at how well academic research in artificial intelligence was doing.

In the report, Lighthill shared his concerns that many important parts of AI research were not making the big progress that people had hoped for. He said that the discoveries so far had not created the major changes that were promised.

Because of this report, the British government decided to stop supporting AI research in most British universities. This decision helped cause a time known as an AI winter in the United Kingdom, a period when interest and funding for AI research dropped significantly.

Publication history

In 1972, the British Science Research Council asked James Lighthill to review the field of artificial intelligence. Lighthill finished his report in July of that year. The Council talked about the report in September and chose to publish it, along with different opinions from Stuart Sutherland, Roger Needham, Christopher Longuet-Higgins, and Donald Michie. This decision came partly because there were disagreements at the University of Edinburgh, a major center for AI research in the UK.

On May 9, 1973, Lighthill discussed the report with several leading AI experts, including Donald Michie, John McCarthy, and Richard Gregory, at the Royal Institution in London.

Content

The report supported studying how the brain and mind work with computers but was very critical of basic research in areas like robotics and language processing. It said that AI researchers did not properly handle problems that become very big and complicated when trying to solve real-world tasks. The report showed a more negative view of AI after the early excitement.

The report split AI research into three groups:

• Advanced Automation ("A"): uses of AI, like reading written words by machine, designing machine parts, and guiding missiles.
• Computer-based Central Nervous System research ("C"): building computer models of how human brains and behavior work.
• Bridge, or Building Robots ("B"): research mixing the first two groups. This group was kept vague on purpose.

Projects in group A had some success but only in small areas where lots of detailed information was used to design the program. This was disappointing because researchers hoped for more general methods. Because of problems with too many possibilities, the amount of detail needed became too big to enter by hand, limiting these projects.

Projects in group C had some success. Artificial neural networks were used to model how the brain works. SHRDLU showed that how people use language depends on understanding meaning, not just the rules of words. This was important for psycholinguistics. Trying to make SHRDLU work for bigger topics was seen as too hard, again because of too many possibilities.

Projects in group B were seen as failures. An important project to make a robot that could copy human skills in using eyes and hands together and solving everyday problems was very disappointing. Also, chess programs were about as good as amateur human players. Because of too many possibilities, making general methods work fast became too hard, needing special tricks for each problem.

The report said that in the next 25 years, group A would become regular engineering technology, group C would join with psychology and brain science, and group B would be stopped.