3 ms·
Alchemy and Artificial Intelligence by Hubert L. Dreyfus (1965) [PDF]
- wrp 4y agoHubert Dreyfus had a good point to make, but unfortunately didn't do a great job of making it. He tended to be rather confrontational and lacked the patience to interact at length with his critics. His project with reference to AI was to collect the relevant results of Continental philosophy, mainly from the works of Heidegger, Merleau-Ponty, Lyotard, and Wittgenstein. He showed that many of the questions being asked in AI research had already been seriously treated by philosophers. However, his books were derided by the AI community, because the answers he found were not what AI researchers wanted to hear. I think it was not until the late 1990s that the AI community generally recognized the truth of his arguments. The submitted paper was his first report on this. He later expanded it into the book What Computers Can't Do in 1972.[1] Look for the much expanded MIT Press edition of 1992.[2] A good complement to the book is the detailed review by Zenon Pylyshyn, who called it "an exasperating mixture of nonsense, well-deserved criticism, understanding and insight."[3] I would say that the first half of the book, where Dreyfus summarizes the work of others is very good, but the second half, where he develops his own ideas, is best ignored. In 2007, Dreyfus wrote a paper summarizing the issues and describing how he influenced the research direction of Terry Winograd.[4] [1] https://www.goodreads.com/book/show/1039575.What_Computers_Can_t_Do https://www.goodreads.com/book/show/1039575.What_Computers_C... [2] https://www.goodreads.com/book/show/796506.What_Computers_Still_Can_t_Do https://www.goodreads.com/book/show/796506.What_Computers_St... [3] https://philpapers.org/rec/PYLMMA https://philpapers.org/rec/PYLMMA [4] https://philpapers.org/rec/DREWHA https://philpapers.org/rec/DREWHA