Norvig vs. Chomsky and the Fight for the Future of AI(tor.com)
tor.com
Norvig vs. Chomsky and the Fight for the Future of AI
http://www.tor.com/blogs/2011/06/norvig-vs-chomsky-and-the-fight-for-the-future-of-ai
4 comments
I feel like Chomsky's argument is being mischaracterized a little here. I don't think Chomsky is necessarily looking for 'simple' rules, but he wants models that explain. Norvig is looking at statistical models that can predict what is going to happen, but don't explain why.
I think Chomsky is bringing up a good point here, but he is also being a little hard on people analyzing the data. We are gathering more data now than ever before, it's impossible to carefully analyze every piece of it to understand exactly why things happen how they do.
We can use data and statistical models to predict stuff, so they are useful for the short term. I think in the long term, analyzing this data will give leads at where to look for understanding.
I think Chomsky is bringing up a good point here, but he is also being a little hard on people analyzing the data. We are gathering more data now than ever before, it's impossible to carefully analyze every piece of it to understand exactly why things happen how they do.
We can use data and statistical models to predict stuff, so they are useful for the short term. I think in the long term, analyzing this data will give leads at where to look for understanding.
Great article. This is a question that's come up before in Mathematics, for example the Four-Color theorem. "Can a proof that's not comprehensible by humans really be called a proof?"
Personally I'm with Norvig. While finding simple patterns is extremely nice and useful, there's no inherent reason why the Universe should follow an easily described set of rules. If it's the difference between an elegant but useless theorem or a useful but incomprehensible pattern, I'll pick the latter every time.
Personally I'm with Norvig. While finding simple patterns is extremely nice and useful, there's no inherent reason why the Universe should follow an easily described set of rules. If it's the difference between an elegant but useless theorem or a useful but incomprehensible pattern, I'll pick the latter every time.
"Why learn anything when you can look it up?"
I'm certainly no linguist but isn't this confusing learning with remembering facts and details?
I'm certainly no linguist but isn't this confusing learning with remembering facts and details?
* The political debates between Foucault and Chomsky in the seventies: Chomsky wholeheartedly believes in the application of justice to injust actions, Foucault was interested in how an act becomes just or unjust.
* The revolution in historical method effected by Foucault lies eliding the descriptive apparatus with which we usually explain the transmission and transformation of ideas (author, book, oeuvre, influence, tradition, consciousness); or, imagine if we did history as if consciousness and its attendant concepts were disqualified as descriptors.
The key to comprehending Foucault here is to suspend the following ideas:
* There exist a particular set of universals regulating historical description of a domain of knowledge
* "Consciousness", whatever that is, is a constitutive principle for that historical description
* The particular set of universals and consciousness exist inextricably, the latter buttressing the concepts of the former
Basically you can sketch a considerable amount of human history without reference to any human subject. You can accomplish a lot just by looking at what was said. What's more, you'll discover a completely set of rules and regularities when you do.
So I have a speculative interest in what discoveries for linguistics might be discovered with Google's method.