Transmission 004

The failure of Artificial Intelligence

From a philosophical standpoint, artificial intelligence as originally conceived was the height of the Enlightenment project: the definition of all of intelligence as pure rationality, and mechanized as pure logic (Dreyfus, 1972). Understanding human thought required investigation into symbolic knowledge representation and its constraints in space-time. Planning algorithms were developed, AI programs made for everything from vision to reading, and progress seemed unlimited. The search for artificial intelligence internationalized, and soon Americans were even terrified that they would be economically overcome by Japanese industry controlled by “Fifth Generation” AI computers. Companies were founded to commercialize such expert systems (systems that formalize human knowledge and transfer it to machines), and the entire artificial intelligence industry appeared to take off.

All was not well for the artificial intelligentsia, as predicted by the Heideggerian philosopher Hubert Dreyfus, employed by RAND to determine if AI was truly a good investment (Dreyfus, 1973). While in limited and highly formal domains such as mathematical proof-proving and fairly immobile factory robotics, AI made great strides forward. On simple problems, such as getting a machine to walk across a cluttered floor, AI failed miserably. AI failed because it could not unify the concrete intuitive and kinesthetic skills with its abstract formal rationality, or in other words, because it was a mind without a body (Dreyfus, 1973). Results at machine translation came back as failures, and knowledge representation languages became so powerful they could not reliably draw inferences.

abstract of the article " Digital Sovereignty. The immaterial aristocracy of the World Wide Web" by Harry Halpin
http://www.geocities.com/immateriallabour/halpinpaper2006.html