Textbook of AI

The Lighthill Report, formally Artificial Intelligence: A General Survey by Sir James Lighthill, the Lucasian Professor of Applied Mathematics at Cambridge, was published by the UK Science Research Council (SRC) in 1973 in response to growing concerns within the British research-funding establishment about the slow progress and inflated promises of artificial-intelligence research. It is one of the most consequential single documents in the history of AI policy and is widely credited with triggering the first AI winter in the United Kingdom.

Lighthill divided AI research into three categories. Category A covered advanced automation, practical applications such as industrial robotics, automated chemical analysis, and computer-aided medical diagnosis, which Lighthill judged useful and worth funding. Category C covered computer-based central-nervous-system research, using computers as tools to model the brain, which Lighthill considered a legitimate strand of neuroscience. Category B, however, encompassed everything else, including most general-purpose AI research aimed at producing genuinely intelligent machines: heuristic search, planning, problem solving, machine learning, robotics, and natural-language understanding. Lighthill concluded that progress in Category B had been disappointing, that the field had failed to deliver on its grand promises, and, in his most influential argument, that no significant further progress was likely because of the combinatorial explosion of the search spaces involved. Toy-world successes, he argued, would not scale.

The report recommended sharp cuts to general-purpose AI funding in the UK, with resources concentrated instead on Category A applications. Its recommendations were largely accepted by the SRC, and AI funding was duly cut. Many British AI researchers, Donald Michie at Edinburgh, John McCarthy (visiting from Stanford), Christopher Longuet-Higgins, and Roger Needham, argued vigorously against the report, pointing out that Lighthill had ignored important successes (notably in machine learning and robot vision) and had built his combinatorial-explosion argument on caricatures. The famous Lighthill debate with Michie, McCarthy and Longuet-Higgins, broadcast on BBC2's Controversy programme in 1973, remains one of the most public airings of the field's internal controversies.

The Lighthill Report, combined with the contemporaneous ALPAC Report's similar effect on US machine-translation funding (1966), brought about what is generally called the first AI winter of roughly 1974–1980. The Edinburgh Department of Artificial Intelligence was reorganised and partially dismantled. UK AI research did not recover until the rise of expert systems in the early 1980s and the response, coordinated through the Alvey Programme (1983–87), to Japan's Fifth Generation Computer Systems project. The episode left a lasting cultural memory in the field: a generation of AI researchers became wary of grand public claims, and successive practitioners have invoked Lighthill's warnings during each subsequent boom, including the deep-learning era from 2012 onwards.