visits to the SPPA website since 27 Apr 2005
Hubert Dreyfus is responsible for introducing Heidegger to cognitive scientists and AI research. In his famous book "What Computers Can't Do" he argued that Good Old Fashioned AI (GOFAI) or 'symbolic' AI is not possible, based on a pragmatist reading of Heidegger and Wittgenstein which thrust Heidegger's concept of Beingintheworld to the fore. The claim was, that strong AI on the model of a program with programmed beliefs, facts heuristics etc. would have no world, would not be embedded, situated and interested, and on Heidegger's theory could not think. His book takes the strong line that no digital computer could ever be a mind. He does not throw AI out tout court however, and he credits himself with forcing AI research in the direction of connectionism and neural nets. Even still, though he cannot produce a negative proof, he still claims it is hard to see how AI could ever work.
Mark Okrent is another pragmatist reader of Heidegger. In his paper Why The Mind Isn't a Program (But Some Digital Computer Might Have A Mind) he accepts Dreyfus's contention that Heideggerian beingin the world precludes strong AI, but he rejects the contention that no digital computer could ever be a mind. He says that the first is an ontological claim, based on what Heidegger sees to be essential deep ontological structures of thinking and being human. The second claim is in a sense contingent, not essential but rather ontic as Heidegger would say (a term meaning to do with mere things on a shallower level than the essential ontological level), and therefore there is nothing from Heidegger on the question of whether there could be a computer that thinks.
My paper presents a much stronger reading of Heidegger on issues of the mind and the possibility of AI. I argue contra Okrent that there are good reasons that Heidegger rejects any attempt whatsoever to model the mind. To do this I appeal to Heidegger's philosophy of technology, in particular his idea of reflective self understanding, that is, that how we see the world determines how we view ourselves. I argue that this is actually quite a common thesis, defended by many authors, in particular the phenomenologist Don Ihde, and the (philosophical) biologist Richard Lewontin. In our technological world, we understand ourselves technologically. I show that based on this thesis Heidegger must reject any attempt to understand or model the mind via a metaphor of machine, telephone exchange or parallel processing computer. He therefore must be seen to take issue with the likes of Searle and Pinker, and in fact anyone who understands the thinking or humans on mechanical terms alone.
Furthermore I show that my stronger reading of Heidegger forces us to go even further. Having rejected the mind as a program (a la Dreyfus) and in fact any artificial model of the mind (contra Okrent) I argue that Heidegger goes so far as to reject the possibility of even growing a thinker, in the sense of a clone, or as Heidegger puts it, "Artificially reproducing a human being". Heidegger has some fairly strong things to say about this, not that it won't technically work, but that this thing will not be a thinker, on his version of what thinking means. He contends that human subjectivity would disappear, and I give an interpretation of what he could mean by this.
I argue in conclusion that this means that Heidegger must even be seen to reject the idea of neural nets, or conectionism in AI research. Heidegger has a radically nonmechanical, antibiological view of what 'human' and 'thinking' mean. He doesn't believe that any type of artificially reproduced thought is possible. I conclude that this strong reading of Heidegger indirectly questions the pragmatist reading of Heidegger advocated by Dreyfus, Okrent and others, as it proposes a much deeper rejection of AI rather than one merely based on a practical understanding of one's environment.
To return to the page from which you came, use your browser's "Back" function
| visits to the SPPA website since 27 Apr 2005 |
