Hacker News Reader

Comment by CaliforniaKarl | Hacker News Reader

Please help me understand this.

If I understand this correctly, the Wikipedia snippet is saying that Mr. Cook developed this theory while under a related NDA with Wolfram Research; and that by publishing a paper, Mr. Cook violated that NDA. The paper in question, the conference, and the injunction blocking publication all happened in 2015.

If I understand that correctly, then should this post be marked "(2015)"?

Also, this seems to me to be a rather run-of-the-mill NDA violation case. The only things making it more more noteable is it being an NDA related to math, and how one of the parties involved is Wolfram Research. Otherwise, this seems to me to be like any NDA involving a technology company, and HN has seen discussions on over-broad NDAs recently.

Finally, I'm not liking the quality of the Wikipedia article: I'd appreciate more information, and maybe links to court documents!

View on HN

This all went down in the late 90s. Basically, Wolfram had been employing a lot of people to continue the basic research he did in the early 1980s on cellular automata. Wolfram's claim-to-fame is that he discovered 4 classes of cellular automata (random, stable, periodic, complex). I claim he took these straight off nobel-prize winner Ilya Prigogine who identified these 4 classes for far-from-equilibrium thermodynamic systems - see Nicolis paper (figure 1): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC388547/pdf/pnas.... (Nicolis was Prigogine's student).

Anyway, Wolfram thought he was on to something really big (meaning-of-life type big). But he also founded Mathematica, thought - "hey, i'll just get rich first". And like Jeff Hawkins who put his Palm Pilot money into his first love of brain research, Wolfram funded basic research on cellular automata. And Wolfram wanted the scientific credit for that basic research - hence the NDAs. Now artists (like Andy Warhol and Damien Hirst) have been doing this successfully for years - taking credit for their employees art. Wolfram reckoned he could do the same. But Mr Cook thought otherwise. Fair play to Cook, I say. It's a very interesting exercise, because there's nothing Wolfram can do now. Cook was the one who made the discovery and he told us so. Wolfram wants to undo that, but with the Internet, he can't.

Regarding the actual discovery. It's significance is that it shows us how very simple rules and interactions can produce something complex and regular (a universal Turing machine) upon which intelligent systems can be built. Given that the brain is almost certainly not a universal Turing machine, I believe that Wolfram over-estimated the long-term historical significance of the discovery.

Awesome, thank you very much for the description!

How is the brain not a universal turing machine? Is there a task that turing machines can do that brains can't? It has been shown that rnns are turing complete [1] so it is certainly believable that brains are

[1] http://stats.stackexchange.com/questions/220907/meaning-of-a...

You can calculate rule 110 by pen and paper, so at least the brain-pen-paper system is Turing complete.

However, the brain is not a Turing machine, probably. There aren't discrete states, no transition table, no tape, etc. We probably don't store information remotely like a Turing machine does (with arbitrary symbols).

There's a remote chance that the opposite is true, and there are functions humans can calculate that Turing machines can't.

Lambda calculus aren't turing machines either but they are turing complete

I won't go into details on why the brain isn't a UTC - this is useful: https://www.quora.com/Is-the-brain-a-turing-machine

Care to mention which reply you think is insightful? I didn't find much of interest at the end of that link; some say yes, some say no, and some answer the inverse but related question (can a Turing machine emulate the brain - because obviously people can laboriously emulate Turing machines).

If you think the brain can compute something a Turing machine can't, it would be interesting to link to something that justifies the position (I personally think a TM is enough).

I don't think a brain can compute something a TM can compute. The brain, however, is not a simple universal turing machine in the sense that it doesn't have a finite set of instructions and an infinite tape-band. I don't subscribe to the Penrose school of thought on this matter.

Well clearly the brain isn't infinite, but nor is any Turing Machine we implement. When we say something like C is Turing complete, we don't mean that any correct implementation can compute what a TM can compute, because no implementation is finite. Eventually both the stack will be exhausted and malloc() will fail; there's a finite amount of state available on any concrete hardware.

My point being that that seems to be such a trivially true statement - conveying no information - so that we must understand the sentence to mean something else; that it can compute what a TM can compute, up to some arbitrary limit related to available state. So, would you say that the brain isn't Turing complete either?

I think the key detail is that the brain is actually more powerful than a turing machine, as demonstrated by our ability to solve halting problems

> our ability to solve halting problems

Is that true? Do you have a source?

I'm not sure this is correct at all. Turing machines can solve certain subsets of the halting problem (i.e does a given program without recursion and with only certain types of loops halt?), but it's impossible to create an algorithm which will say whether any program halts. I see no reason why the Human brain would be able to solve the halting problem, or even any problem a Turing machine can't.

Interesting.

How does Wolfram's NKS book fit into all of this?

It's interesting but not earthshaking level interesting and does not warrant the expanded title of 'A New Kind of Science'.

No, I actually read the book, so I know what it is.

What I meant was how the book fits into the timeline and the authorship? On one hand it's largely about rule 110, which was Cook's finding, but in the other Wolfram said he was working on it for 10 years.

When Wolfram said he was working on it for >10 years, he meant that his minions (with signed NDAs that prevented even acknowledging the existence of a proof for rule 110) had been working on the research for >10 years. He did, however, work on writing the book. I read it, too. Didn't enjoy it at all.

The significance is also that cellular automata might be a good model of physical reality. The idea is probably more interesting than practical though.

The docket for the court case seems to be this one: https://www.plainsite.org/dockets/20370j96/illinois-central-...

>Finally, I'm not liking the quality of the Wikipedia article: I'd appreciate more information, and maybe links to court documents!

It's Wikipedia! If you are not liking the quality then feel free to edit it and make it better.

That's true! But I didn't have any additional information to add, so my edits would likely reduce the quality of the article.

I don't have any experience in looking up court records.

How kindly would you look at someone who said they didn't have experience looking up documentation?

Have you ever tried looking up documentation from a field that you aren't familiar with? It's not trivial.

When I did some initial dabbling with microprocessor programming, reading the docs made absolutely no sense to me. (My field is desktop software) Most things are still a mystery to me. Most docs expect that you are familiar with the field, and in most cases you won't even find them if you aren't.

I would say that anyone who would make this analogy actually doesn't have experience looking up documentation in any field they're unfamiliar with.

I would LOVE a primer on how to look up court documents! Please hold forth!

Let's say Apple hires you to work on Siri, and you discover that a certain construction of a deep neural network allows you to construct a smooth (differentiable) Turing machine which can do wonderful new things.

Then, without asking for permission, you go on to publish the details of your research.

What do you think Apple might do in this case?

Cook's proof was presented at a conference in 1998 and published in 2000. NKS came out in 2002.

Reply on news.ycombinator.com