3 ms·
Big Data is the new Artificial Intelligence
- mark_l_watson 13y agoCringley's article is a lighter treatment of the recent "debate" between Noam Chompsky and Peter Norvig.
- mathattack 13y agoYes - very clear that he lands on the side of Norvig, but he takes it to an extreme further than Norvig, no?
- mark_l_watson 13y agoI think Cringley is more on Chompsky's side since he is concerned about not being able to understand the internals of models created with machine learning. It may sound like a contradiction, but I agree with both Norvig's and Chomsky's positions, depending on context. I would like to see a model for real AI developed from an understanding of how our brains work, but real AI will probably not be created that way. When/if we see real AI, we probably won't understand how it works.
- hollerith 13y ago>When/if we see real AI, we probably won't understand how it works. If that's true, then it's very distressing since a smarter-than-us AI whose workings we do not understand would probably kill us all.
- ableal 13y agoMaybe the AI will be sufficiently intelligent to just become catatonic in short order. (I believe I'm filching this from Larry Niven, no precise quotation handy.)
- mathattack 13y agoThis is symptomatic of what's happening in math. Theorem provers figure things out that are beyond our comprehension to test or follow. Considering how many bugs we find when things are simple enough to check, we have to be careful when things get too complex to check.
- mathattack 13y agoI like this article, but I think he misses that part of the problem with early AI (in addition to not enough processing power) was poor mental models. AI became LISP programming. Another comment of his is a stretch: The system just figures it out. And that means there’s no need for theory. It isn't that there is no theory behind what Google is doing. It's that there's theory of how humans process things. There is a theory behind creating a pattern recognizer. It just doesn't force the pattern recognizer to conform to an existing theory of how humans process information. Building the black box still requires some knowledge of what you're trying to solve. I'm a big believer in the power of Big Data, but history is littered with failed projects of machines that replace the thinker.
- chazu 13y agoThanks for this. I cringed when I read "there was no underlying theory." That dog don't hunt.