"Alchemists were so successful in distilling quicksilver from what seemed to be dirt, that after several hundred years [...] still refused to believe that on the chemical level one cannot transmute metals. To avoid the fate of the alchemists, it is time we asked where we stand"
[Nowadays, the alchemy metaphor is of course well-known due to Rahimi and Recht (https://archives.argmin.net/2017/12/05/kitchen-sinks/). I find a specific sense of irony in the fact that while R&R "alchemy" critique was mostly of contemporary DL practice -- an extreme version of the distributed/fuzzy/statistical/"connectionist" approach, the Dreyfus' "alchemy" critique is mostly of symbolic/formal/rule-based approaches, implicitly advocating _for_ more statistical approaches! And so the pendulum keeps on swinging...]
It was a truly thought-provoking article. The first thought that came to my mind was, "Why do so many researchers succumb?" As a Ph.D. student, aside from the fact that AI research encompasses very interesting topics, I feel tempted to enter the field because of the "expected" quantitative and monetary rewards that this research is likely to provide to researchers in the future. Especially in these recent volatile domestic and international circumstances, I think that many people might be more easily swayed. While our freedom to act for personal gain should not be restricted, it is hard to predict what effects such behavior might have on society as a whole when it becomes collective. I think this article suggests one such effect. Thank you for the post.
"Alchemists were so successful in distilling quicksilver from what seemed to be dirt, that after several hundred years [...] still refused to believe that on the chemical level one cannot transmute metals. To avoid the fate of the alchemists, it is time we asked where we stand"
--- Alchemy and Artificial Intelligence, Dreyfus 1965; https://www.rand.org/content/dam/rand/pubs/papers/2006/P3244.pdf
[Nowadays, the alchemy metaphor is of course well-known due to Rahimi and Recht (https://archives.argmin.net/2017/12/05/kitchen-sinks/). I find a specific sense of irony in the fact that while R&R "alchemy" critique was mostly of contemporary DL practice -- an extreme version of the distributed/fuzzy/statistical/"connectionist" approach, the Dreyfus' "alchemy" critique is mostly of symbolic/formal/rule-based approaches, implicitly advocating _for_ more statistical approaches! And so the pendulum keeps on swinging...]
This is beautiful and needs to be WIDELY read.
It was a truly thought-provoking article. The first thought that came to my mind was, "Why do so many researchers succumb?" As a Ph.D. student, aside from the fact that AI research encompasses very interesting topics, I feel tempted to enter the field because of the "expected" quantitative and monetary rewards that this research is likely to provide to researchers in the future. Especially in these recent volatile domestic and international circumstances, I think that many people might be more easily swayed. While our freedom to act for personal gain should not be restricted, it is hard to predict what effects such behavior might have on society as a whole when it becomes collective. I think this article suggests one such effect. Thank you for the post.