Scott Aaronson tears NKS apart [pdf]
scottaaronson.com
scottaaronson.com
It's a good review, no doubt, but I think Aaronson, while catching Wolfram using a bit of hyperbole on occasions, misses the essence of what NKS is.
It's not simply a review of existing CS theory.
The Feynman quote was quite illuminating. Wolfram does take the discrete conversation to the next step. He doesn't answer the questions, mainly because what he is proposing is, well, a New Kind of Science. Which means that there are going to be hundreds or thousands of questions.
What might be missing the most is a methodology for exploration. But I think overall Wolfram is just a bit ahead of his time. Those things will be worked out.
he is proposing
And what Aaronson analyses is that book's statements form no proposal.
Which may not be a bad thing, except it all comes at a price of credibility. Why does one need to create all that fuzz? Isn't it because the revolutionary product/idea cannot win followers and reputation solely on its merits?
Same applies to his book. With a bit of modesty its impact could have been so much greater. Especially since its subject is indeed fascinating.
Why can't you get http://www.wolframalpha.com/input/?i=how+many+college+gradua....
But you can get it for each university with http://www.wolframalpha.com/input/?i=harvard+graduates
Seems like it'd be a simple aggregate...
Little things like this would be flushed out with a bit of testing over time. The hardest part for WA is going to be figuring out how people want to talk to it, and that will take time.
It is not.
Or at best in a sense that Pythagorean Harmony is.
Subject has been explored to death long before ANKS. In sixties by every garden variety computer scientist, and in eighties by physicists, giving no insight for either science. Even as an art, his book was few years too late to "generative art" market.
Outcome of this popular fascination is now known as AI-winter.
That'd be it. Can you define the paradigm? State the idea and the method?
The classical AI approach was more about modeling, knowledge representation, inference rules, decision trees and so on. The only thing related to NKS the classics dabbled with were neural networks. Most of them were also very restricted, though - in order to have a shot at analyzing them, the common neural networks have relatively simple structures.
The idea of cellular automata is of course not from Wolfram, but he studied them to an impressive extent. To me they are just one variant of the theme, though. For me NKS is not the invention of evolutionary computation. But it portrays one part of it very well. And he argues that many things could be handled in that way, which was not obvious (I remember how he could imagine the real world to work like an cellular automaton, and still agree with quantum mechanics).
You still haven't said what is the scientifical idea.
Because I for one don't know. John McCarthy doesn't too, he once hoped to gain such knowledge by careful study, and later remarked regarding Wolfram "In the 1950s I thought that the smallest possible (symbol-state product) universal Turing machine would tell something about the nature of computation. Unfortunately, it didn’t. Instead as simpler universal machines were discovered, the proofs that they were universal became more elaborate, and [so] did the encodings of information."
how he could imagine the real world to work like an cellular automaton, and still agree with quantum mechanics
Oh, and he got that wrong too... (by the so called No Hidden Variables theorem).
But I am also not sure exactly what you want. For example, would it be strange to you if I found ants fascinating? Even though there is no scientific idea at first. Maybe if you look at ants for a while, you'll generate all sorts of ideas (and even in Goedel-Escher-Bach, which I guess is very classic AI, they make an appearance).
So I am not sure that I accept your complaint that you can see no scientifical idea. Actually I am not sure, what is a "scientifical idea"? (Honest question)
Looking at cellular automata I get all sorts of ideas. So I find them fascinating. And why is it not a "scientifical idea" for you to build complex things from simple rules? I don't think it is such an obvious idea.
I am willing to try to answer your question, but first please explain to me what aspect exactly you want to hear about ("scientifical idea").
As for McCarthy, I don't really worry much about what he thinks, especially since LISP is actually from the "classic AI" side of things (I don't mean that as disrespectful - but he is a specialist for LISP and computation, not for complex systems). I think finding the smallest Turing machine was just something to be done for fun - I don't think Wolfram thought much more about it either. So that finding it did not blow McCarthy's mind hardly disqualifies the whole subject in my opinion.
Also, could you link to the No Hidden Variables vs Wolframs suggestion please? It was only a wildly speculative thing, even in NKS, though - not the cornerstone of his world view.
1/ Aristotle's classification of forces and the idea of quintessence;
2/ Plato's idea that planets move necessarily in circles, because it's the most beautiful and simplest shape. That one witheld progress in astronomy for 2000 years;
3/ Beautiful idea that the market is a statistical system, and thus can be modeled by tools of mathematical physics. That the assumptions of these models are not met in economy, no one bothered much till just recently.
So you take some guiding rule first, and then force world around to accept it.
I think I get an idea where you are coming from, though, and I lost all interest in discussing with you.
Oh, so sorry.
PS. I got that as consciously and purposely insulting.
What I wanted to say, and probably failed, is that beautiful patterns do not form any interesting science, that is they lead to no knowledge. This is a classical dichotomy between intuitionist/inductive and deductive philosophies of science.
I think cellular automata provide at least one way to study complex systems, which is a relevant field of study.
Isn't it because the revolutionary product/idea cannot win followers and reputation solely on its merits?
If the odds are worse without promotion, what rational human wouldn't use it?
I believe the last time was Cuil.
Being able to work the hype machine is great, but if you generate more hype than you can deliver on it hurts your image. Since A New Kind of Science was similarly underwhelming, not to mention even more pretentiously hyped, many are cynical.
Cuil's sin was failing at a solved problem, and most plainly, trying to be a "Google killer" (I believe they lectured Google on the size of their search database before they even launched).
NLP answer engines aren't a solved problem, and WA was a pretty decent 1.0. Sure it failed techcrunch's vanity test, but for what Wolfram promised in those videos, it performs.
It is a rather little known fact that, trained through TM catastrophe, its founders went on to form a stealth company, called Ab Initio, which is wildly successful despite having no marketing. Indeed they're actively avoiding it through NDA contracts to the point that general public doesn't exactly know what exactly their products do.
I am not ridiculing Wolfram, just wanted to point that above is as an interesting anecdote.
Since when polluting valuable community information resource with paid commercial fluff became acceptable?
Aggressive promotion poisons the well for everybody. People come to HN for the honest information and opinion exchange with the like-minded, not to be bombarded with PR.
And by the way, I remember time when NH core were not get-rick-quick wannabes but engineers and, well, hackers. Which does not exclude being entrepreneurial. It just means being not easy for some 'smart business' to manipulate.
If the odds are worse without promotion, what rational human wouldn't use it?
The one that is familiar with game theory. Defection does work in the trusting, cooperative environment. First time. After that the community switches to 'tit-for-tat' strategy and bans the aggressor into oblivion. Or core of the community decamps to greener pastures leaving 'smart businesses' and their shills to their task of filling communal cesspit with more crap.
I fail to see what pollution has happened here. Both NKS and Wolfram Alpha are pretty serious endeavours, that most hackers should be aware and probably interested in.
Aggressive promotion poisons the well for everybody. People come to HN for the honest information and opinion exchange with the like-minded, not to be bombarded with PR.
You're mixing various concepts here... "Aggressive promotion" is, effectively, what most entrepreneurs need to do at some point to get some users on the system. The problem is not with promotion or how aggressive with it. The problem is with "unethical", brutal, blunt promotion, like spam.
I don't think anyone here has something against "aggressive promotion" in the form of interesting information that happens to relate to something that someone is building. Except for you, perhaps.
I think you are the only one here enjoying reading PR releases. This sure looks like spamming (or a good PR work):
http://www.google.co.uk/search?q=site:news.ycombinator.com+w...
What makes you think it was Wolfram's talent per se and not that of some marketing exec ?
They say that love is blind, ego must just be plain blinding.
This is the best theory I've seen yet to explain his behavior:
http://groups.google.com/group/sci.math/msg/37f51351f5340e58...
It makes a lot of sense, in light of the man's biography.
See this:
http://www.cscs.umich.edu/~crshalizi/reviews/wolfram/
for an informed critic's perspective on the details.
Also, SW hasn't gotten anywhere near a 'greatest discovery', so I'm not sure the comparison is valid.
The first publication on the Mobius strip was written by a guy named Listing, who did the same work as Mobius at the same time. I have to warn you that googling "Listing Strip" is slightly NSFW.
Kalman was only the popularizer of the Kalman Filter.
Grassman, who invented vector spaces and developed a lot of linear algebra, only got credit posthumously.
Some people, like Fourier, only inspired the theory named for them. Modern theory is so much more general than classical theory that entire careers can be built by translating an old mathematical idea to a new domain.
Sometimes, a new idea gets labeled as an extension to an existing idea if it so much as resembles an old idea. For example, the Hamilton-Jacobi-Bellman equation was Bellman's work. This is dangerous if you're after immortality -- there are very few ideas in math that aren't even slightly hinted at in the works of: Euler, Gauss, Jacobi, Hamilton, Lagrange, Laplace, etc.
Sometimes, though, the new guys can usurp the old. When set theory being laid down as the core idea in all of mathematics, the mathematicians who reformulated ideas in rigorous set-theoretic language often got the new names. That's why we now have the Haar Measure instead of the Hurwitz Invariant Integral.