Hofstadter's Law: "It always takes longer than you expect, even when you take into account Hofstadter's Law".
Korbo, Lorbo, Jeetbo.
You can get updates to this blog on Twitter: follow @intrcnnctd.
Okay, given you were all so bloody useless when I asked about Principal Component Analysis, here's another question (from flat James): Why do you get bubbles in a glass of water when you leave it out overnight? Tell me?
This mail delivery program uses a statistical learning algoritm (Bayes) to file your mail to the correct folders, based on word count and your manual corrections: ifile. Looks like it's worth a punt [cheers Nick].
And... breathe. My personal gestalt hasn't been integrating terribly well over the last week (or more). And what's that literary device whereby the inner state of being is reflected by the environment as a whole? So much for holidays. Talk about journey of discovery. Talk about dog shit. Talk about being on my own in a house.
And another thing: In the long-time limit, there is no redundancy in nature. There is no it doesn't do anything, it's just to look good in the real world. It's all there to feed the gestalt baby.
(All my going on about nature, evolution, interfaces and so on isn't some thinly veiled commentary on the complexities, to-and-fro-ness, loopbacks and meaning in my life, by the way. I am just going on about nature, evolutions, inferfaces. And so on.)
Okay, given that questions and audience participation are the flavour du jour, here's one for you: Given I have a rather large number of items each associated with a vector of high dimension, does anybody know the algorithm to perform Principal Component Analysis on my dataset to get a the principal components themselves, and b the positions of my items in principal component space? I've been hunting on the www, but nothing's jumping out. Tell me?
The new BBC2 idents are -- well. What is it with purple and gold? It's not a colour combination with any kind of resonance. It's interesting that the new trailers divide down the middle and don't centre everything, as they used to, but the "BBC TWO" purple box is just ugly. It'll be interesting to see where they go with it though. (See also: old BBC2 idents.)
"The desktop is dead," says David Gelertner in The Next Computer Interface [via blackbeltjones]. Various other concepts are covered, all coming from the basic idea that the desktop/filing-system metaphor is limiting and outdated. Two of the products mentioned are: Gelertner's Scopeware, a browser-based piece of software based on the Lifestreams model; and, Star Tree, a way of navigating/viewing the web using nodes and arcs.
Naturally, I've been thinking about this.
Although I think the filing cabinet model is dated and not suitable to the way we store documents on computers, from a UI point of view I believe it's extremely strong due to its rigid map component. The brain is very good at analysing and remembering terrain. I've noticed this particularly moving from Mac OS 9 (strong spatial desktop UI) to Mac OS X (weak spatial desktop UI) and it's a significant difference.
So, let's decouple the two parts of the filing system: The storage system is nonsense. We've got a computer -- the computer should be the secretary and do the filing. Finding documents/information is a human job and we should keep the interface to do that, namely the spatial desktop. The human should influence/construct the filing layout; the computer should file information appropriately based on past behaviour which can be derived from either which documents are where, and what the folders mean. Oh, of course: folders are dynamic search results, not just buckets. What I'm saying is that like documents should accrete near like documents, in the corners of your file system, in the same way my pens "naturally" tend to end up on the right hand side of my desk (because I'm right handed). Whenever I want a pen, there's one to hand because I used one there last time! That's important.
The other big point is something that both these systems are reaching towards but I don't think has been explicity mentioned. A computer monitor is replacing all our senses when interacting with the universe on the other side of the LCD. We're left with sight (in most cases) and a very crippled form of it; real vision doesn't work by having everything you see commanding equal attention. Computers aren't even as good as books here -- you haven't even got that contextual "how many pages are left" sense. We're building from the ground up.
What is needed is peripheral vision. This may or may not couple with zoomable interfaces, but it does enforce one technology constraint and one UI constraint. Firstly, that the computer needs to be capable of finding contextual and associating information and displaying it in a way that making it clear said information is secondary. Maybe when looking at an email, associated emails glow or the folders they're in appear "well trodden". Maybe you can overhear IM conversations [cheers Phil G]. Maybe the bottom left of your screen is dedicated to your contact-information-memory, and contextual information about everyone mentioned on the screen is displayed in that portion. Remember everything has to be rebuilt here: your reality associations layer, tunnel vision, concentration.
The second point: To make this metaphor clear, we have to lose action-at-a-distance. Menu bars don't work in this model. Seeing bold text, you should look closer and see buttons, levers and handles on it that affect its weight, or typeface. Or maybe you can pick up a tool to influence the text, and carry it with you, but if the cursor is your point of concentration, your centre-point of tunnel vision, you shouldn't be affecting items miles away by an action in a menu option. It's a rigidly enforced box model we need. Boxes can choose what boxes they contain, and boxes have their own behaviour and can decide what to do based on their context. But boxes can't dictate what happens outside them. Generally, that's not how the world works.
We need high-level concepts about how the universe works here, to emulate them on the screen. Systems are systems afterall, and I think our metaphors so far have been slightly too parochial.
Evolution and nature | Of course, the model of evolution as being driven by environmental change is too simplistic. A better model would incorporate more of the give-and-take of nature. A species in adapting to its niche adopts some other unnecessary characteristic as a side-effect, thereby changing its niche. The species will not be perfectly adapted to its new niche, which will drive further change.
The point is this: A [a species] figures out/ evolves a way to take advantage of [extract energy from] B -- B being either another species, or "everything else" a.k.a. "the environment". The niche for B has therefore changed; a successful B will evolve a way to take advantage of A, changing the relationship from a parasitic one to a symbiotic one. Successful Bs are the only kind there are, given time. (This symbiotic relationship can be viewed as a curious kind of altruism. Seeing A do something that helps B, ask: how is A feeding off B?)
There are no parasitic relationships. How? Because the environment is not static, it's just everything else. (In an entropy diagram we would refer to A as the system, and B, the environment, as the surroundings. It's a convenient line to draw for certain arguments, but not a real one.)
Es updated the Roast Dinner HOWTO, and I added some photos. This is my public service contribution for people that don't come from the womb fully able to cook. People like me, in fact.
Digital Time Machine Spec: "Buy some time. Then re-sell it at a profit". Or, How To Get Even More Ads on Television. Wow, the system really will stop at nothing.
The 8 latest posts are named
How any of the Big 3 could own connected products, Pricing hardware and changing business models, Orbits and hardware, BERG Cloud press, Testing, Facebook should make a camera, Instagram for webpages, and Ze Frank on ugly.
2013 May. 2012 July, May, April, March, February, January. 2011 May, March, February, January. 2010 December, January. 2009 February. 2008 December, November, September, August, July, June, May, April, March, February, January. 2007 December, November, October, September, July, June, May, March, February, January. 2006 December, November, October, September, August, July, June, May, April, March, February, January. 2005 December, November, October, September, August, July, June, May, April, March, February, January. 2004 December, November, October, September, August, July, June, May, April. 2003 December, November, October, September, August, July, June, May, April, March, February, January. 2002 December, November, October, September, August, July, June, May, April, March, February, January. 2001 December, November, October, September, August, July, June, May, April, March, February, January. 2000 December, November, October, September, August, July, June, May, April, March, February.
Interconnected is copyright 2000—2013 Matt Webb.