170 comments

[ 4.5 ms ] story [ 112 ms ] thread
i know. when i found that out, i spent half the day on wikipedia trying it out with different pages. so i built a way to automate it :)
As they describe in the page, when you describe something, you are trying to place it within a general framework.

In trying to describe a domain in an encyclopedia, it does little good to use language specific to the domain. If your reader doesn't know what Biology is, it's little use trying to describe it in terms of Microbiology and Immunology. You have to use more general language.

There are are few words more general and meaningless in our language than "organization".

edit - grammar

Seems like it would be nice to cache some of these requests, instead of repeatedly hammering the Wikipedia API. And why is it grabbing images?

http://i.imgur.com/KSju9.png

(Incidentally, my test case for the screenshot did not lead to Philosophy!)

totally agree. i hacked this together pretty quickly last night. definitely a lot of room for improvement. thanks for the feedback!
sorry to say, but couldn't you have thought about that before posting it on HN? you're on the front page, if your site really runs without caching, you must be wasting an enormous amount of WP's resources.
As far as I know, this was noticed in May last year on the xkcd forums. http://forums.xkcd.com/viewtopic.php?f=2&t=71309 There's a writeup of it with a gigantic picture over here: http://www.mrphlip.com/wikiphilosophy/
(comment deleted)
True, the "wikipedia loop" itself is not a new discovery. But still, congrats for coming up with a nice GUI to demonstrate it!
Very first try, The Eiffel Tower, oddly enough, leads to Great Britain
This reminds me of a game I used to play as an intern with other interns to kill time. Someone would pick a Wikipedia article deep down in the bowels of Wikipedia and we would race to the article from Wikipedia only using links. We could get the article surprisingly fast, most within only 5 intermediate page visits from the homepage.
I can't replicate it, but I encountered a weird bug. I entered "Star Trek" and hit Enter. The suggested items list appeared and I clicked on "Star Trek". Page titles appeared as normal, but everything appeared twice, like so:

  Star Trek
  Star Trek
  Cinema of the United States
  Film
  Cinema of the United States
  Recording
  Data
  Level of measurement
  Film
  ...etc until the last page title, "Reality"
Seems like two processes were running in parallel and outputting to the same stream.
(comment deleted)
This is beautiful. I've always thought sites like IMDB and Wikipedia are hypertext expressions of the purest form.

As a nerdy kid, over a few years I read my way through a significant chunk of the early 1990's Groliers Encyclopedia set my Grandma bought me at Krogers. And it was never cover-to-cover reading, instead it was filled with hopping around, page-to-page, volume-to-volume. Start at oscilloscope, but what's a cathode ray? And then it's in television? How does broadcasting work? I'd sit on the floor with a half dozen open encyclopedias open around me, the same way we can do now with browser tabs.

Wikipedia has many flaws, but it's a fantastic tool and I'd have killed for it as a 12 year old with 5 volumes of an encyclopedia in my backpack!

Ah yes. To me (to my productivity's detriment...) quite often when I visit Wikipedia it becomes a learning adventure. From one thing you can hop to the other, finding out all sorts of interesting and obscure things. To a curious mind like mine, Wikipedia is heaven!
For myself, I would broaden this and say that almost every link I visit on HN, or every forum post I read leads to more than 1 additional link. Thus three windows of Chrome open with 10+ tabs each :)

I find this drill-down effect to be quite illuminating, if not productive. Strangely, it seems to provide both depth and breadth of topics in relatively equal measures.

Sometimes, I go onto wikipedia just to find out one fact... and end up with 6 or 7 tabs open on the topic, ranging from genetics to Judaism.
about an hour after posting this, someone changed the philosophy page so the philosophy-reality loop doesn't happen.

http://en.wikipedia.org/wiki/Philosophy

it still brings you to a loop, but it's not an immediate back-and-forth between the last two.

I quite like the loop between Doubt and Belief http://wikiloopr.com/Doubt
The loop found for "Psychology" is pretty neat: http://wikiloopr.com/Psychology

Edit: Looks like somebody is editing certain articles that don't have philosophy as the first link to have it. Psychology used to loop with itself through like 5 intermediates.

"Greg James" As in Radio one DJ.....Doesnt work!
It's funny, but it's true that everything in life ultimately leads back to philosophy.

As noted by XKCD, everything is based on something else: http://xkcd.com/435/

And I think after Maths on that should come Philosophy.

"Nematode" and "Flow (psychology)" stop at "psychology" which I think is just another example of that discipline putting on airs.
I agree. Thought that at the time. Was a philosophy student at the time too though.
I don't think the loop is the interesting part, what's interesting is that articles lead to philosophy.
Bangkok is 25 and Chopsticks loops with itself.
Keith Chegwin gets stuck in Great Britain, which is probably for the best.
Either philosophy is a dead end, then, or it is the base of all knowledge. Depends on whether knowledge is a tree or a graph, which is... a philosophical question!

My brain exploded.

If there is some characteristic Q that is true of all knowledge, then Q must be true of Q. A cycle!
Trees are connected acyclic graphs, so knowledge could be both. ;)
Yes but graphs can very well be cyclic, so eventually a path from a node could lead back to the node itself.
From a mathematical perspective, this is actually not that surprising.

Think of the act of clicking the first link in the Wikipedia article as a function that takes in the page you're on and outputs another page.

If you call this function on itself over and over again (ie use the output of one step as in the input of the next step), you will eventually enter a loop. The proof of this is simple: there are only a finite number of Wikipedia articles, therefore you must eventually reach an article you've seen before. (This is the same reason systems with a finite number of states cannot be chaotic.)

Since it's necessarily true that all articles will eventually reach a loop when you iteratively click the first link, we have to ask: how unusual is it that they usually reach the SAME loop?

Never thought about it as a finite state system, but that's a great way of looking at it.
Agreed. I'd love to know how many different cycles the wikipedia article graph contains. If most articles lead to 1 out of 20 possible cycles, it's much less interesting than if it's 1 out of 5,000,000.
I've been looking to do something like this with the Networkx library in Python, though on an internal company MediaWiki with a much smaller article base. Going to try to visualise much the same thing, the major loops and cliques in the graph.
Unless the articles are edited before you come back!
Another way of thinking about it: There are N nodes that loop back to themselves. When starting at a given node, what are the odds that you'll never hit one of these N nodes? Additionally, we know that many of the N nodes are very general ones, like "truth", under which many topics roll up.
Unless an article is reached that has no link. [A dead-state in the stated finite-state machine]
This is the same reason systems with a finite number of states cannot be chaotic

Without more information, I don't think this is true. It is true if the transition function between states depends on a finite number of previous ones (WLOG if s_n is a function of only s_{n-1}), but I think that it isn't if the transition depends on an infinite number.

(Though, in this case, clearly there is no history, so what you say is true.)

If your model only depends on at most a maximum fixed length of history, you can always model your finite systems to only depend on the most recent state. (Just make a copy of the history part of that state.)

In the case of depending on up to an infinite history, I wouldn't call that system to have a finite number of states any longer.

The fact that many lead to the same loop(1) of Argument <-> Logic in unsurprising, too. Wikipedia articles usually define their subject in in the first sentence. Defining things works by saying what general category they belong to, then by differentiating them within that.

E.g.

A fish is any member of a paraphyletic group of organisms that consist of all gill-bearing aquatic craniate animals that lack limbs with digits.

Google Inc. (NASDAQ: GOOG) is an American _multinational corporation_ which provides Internet-related products and services [...]

A table is a form of _furniture_ with a flat and satisfactory horizontal upper surface [...]

And so on. If you always walk up the abstraction chain because you're picking the first link (the general category), you'll end up at the root of the categorization tree, which is likely something along the lines of Argument, Logic, Fact etc.

Note however that the current state of wikipedia has a different root loop for me, Logic leads to Philosophy.

(1) or one of a very limited number of loops, I also saw Science <-> Knowledge.