Registered: Nov 2004
Posts: 36

Could someone please explain the connection between the new program Wolfram Alpha and NKS?

__________________
Craig

Report this post to a moderator | IP: Logged

Registered: Aug 2003
Posts: 712

It's a fair question. There are some layers to the answer and Wolfram himself might answer it differently. A short answer is biographical; Wolfram got the idea for it from some insights from NKS; many people who worked on NKS went on to be key people in making Wolfram Alpha; and some of the underlying techniques use "NKS ways of thinking". We didn't discover one "special mcguffin" rule 88569988 and get all of it for free, though. Wolfram Alpha includes a lot of traditional science and contributions many other people have made to their subject areas, whose shoulders we are standing on, so to speak.

NKS expects that the world will be computationally tractable in pockets of simplicity, rather than exhaustively predictable in every detail. Even if there is a simple rule for many systems, you couldn't do all the calculative work to evolve those rules in all cases, only in a selection of simpler ones. Those would also be the places human knowledge would find something it could bite on and figure something out. OK, so if you expect finite pockets of computational tractability, can we just get them all, or a reasonable fraction of those that are actually known? Key Wolfram's encyclopedic imagination - he starts asking, how much space do the world's libraries use up, how much in scientific sources have actual formulas, can we just go get them all? You might have thought they'd be practically unlimited, but from NKS intuition and a sense of history, he didn't think so.

The traditional AI problem called the "frame" problem was to teach a computer the right surrounding reference or context that humans see easily. That sort of generality was found to be particularly difficult to get right. But if you have the idea that there are only a finite set of places you want to land, where you can do something meaningful in computational terms, you face a significantly simpler problem. Pick out from free form inputs those that should funnel to each of those domains and admit you don't know what to do with the rest. You still have to get from A to B. There one can take a page or three from branching-possibility NKS-style systems and how one programs them. That's maybe the most direct use of NKS-ish ideas, though some of it would be familiar to anyone who has worked on traditional parsers.

At the bottom of any Wolfram Alpha output is some traditional algorithm in some established domain - usually "OKS" but of course you can also ask Wolfram Alpha about cellular automata or other NKS systems. Getting to that specific traditional algorithm as opposed to N others around it uses some applied NKS and some traditional computer science techniques. The overall idea might have seemed impossibly broad without a bit of NKS way of thinking. Ask five people who worked on both and you might get three or four different answers, but most would probably mention those points.

I hope this helps.

Report this post to a moderator | IP: Logged

Registered: Jun 2012
Posts: 1

Hi WolframScience people-

I have a lot of questions about the connections between NKS and WolframAlpha. I'm curious about what is, and is not, appropriate to discuss on this forum. I understand if it is not possible to discuss some or all of these topics, for whatever reason.

Here are some things I've been wondering about this topic:

Question 1: Is the WolframAlpha core AI derived from NKS-enlightened science? Was it built 'in house', or was it licensed from an external source?

Question 2: Are there any known real world physical systems in which a Universal pattern operates directly on the physical substrate of the system, at a scale with exploitable quantum mechanical properties? E.g. Have any Rule 110-like patterns been discovered that operate in an environment of braided, entangled anyons interacting in a Two Dimensional Electron Gas (2DEG)? Has anyone looked for NKS patterns in such an environment?

Question 3: The WolframAlpha core AI sure is pretty smart. Does it qualify as Strong AI?

Report this post to a moderator | IP: Logged