Daniel Demski

I've long wondered about a reverse mapping of the Mandelbrot Set: give it a picture, and it tries to propose coordinates in the fractal-to-rule-all-fractals which will generate that picture. One of the chief reasons I want to do this is because everyone says the Mandelbrot set seems to contain every sort of shape you can imagine. It does contain many shapes, as you can see on the Wikipedia article.

But does it really, in any sense, contain every possible shape? I don't think so, and I'd like to see a fractal that does!

Of course there are probably infinitely many fractals which do contain every possible shape. Let's try and design one, shall we? I'll consider shapes = black and white (for now) images, that is, collections of square pixels. So a fractal would contain every possible image if some view of it was close to any given tiling of black and white squares.

Well. That seems pretty easy. Just take the possible images in order. The first one is all white, the second all black. Put the all black image somewhere in the all white one; say, the center, but really small so it's hardly visible when just looking at the all white image.

Then, in the center of the all black image, but the image represented by 0001; four pixels, one of them black. In the center of that, tiny, but the next image, 0010, which is a different pixel of the four black. Continue up to 1110 and 1111, then start on... 000000001.

Not a very satisfying fractal honestly. Maybe if it were to be constructed from the bottom up instead it would be a lot better. But something not based on pixels, but instead on curves, would be a bit more satisfying.

In other news, I plan on starting a blog elsewhere which will actually hopefully update weekly. The idea is going to be posting a new program every week, with the intention of that program actually being useful.

I use a lot of Google Alerts, and I feel like every time I take a look at them I find really interesting things and people, but I end up actually looking through the Google Alerts I receive for several reasons; first, because they are not well-organised or high-quality, second, because they are something of a timesink.

What I've been doing a bit of lately is looking at my google alerts only for terms which get very few hits. This is because of the strange phenomenon of me only having so much time, and only wanting to pay attention to things I have time to pay somewhat full attention to. It doesn't make that much sense, though, to want to pay full attention to an automatically generated list of Web pages- partial attention is just fine. But if I look at ALL my alerts, the topics people talk about more get overrepresented. What would naturally be nice would be if I could look through them in proportion to my interest in a topic and the quality of individual posts.

I got a new camera after dithering a bit about it. I always hesitate hoping for the best deal, but in this case it seems like what I wanted was available, so I just bought one. I am hoping that being able to take 1,500 12-megapixel pictures (4000x3000) before needing to empty it out (as opposed to 50 5-megapixel 2048x1536 images) will make it so that I carry it all the time. So far this has been the case. If I take very many videos I may actually run out of the 8 GB card, but not with just photos.

But what I'm actually writing about is my ideal camera. I was hoping 12 megapixels would mean when taking pictures I would get significantly more detail than I can myself see. But, of course, there are still circumstances where I can see more than the camera. I want a camera designed to see more than me in every way, a sort of scientist's camera (in that it would be really useful for gathering field data).

First I want a wider spectrum. I know that this would be difficult. Cameras can see infrared if their infrared filters are removed, but it is just picked up on the red sensors. That loses information about the picture in some ways; picking up the infrared makes it difficult or impossible to determine what the image looks like in visible light. To broaden the spectrum one would need either a longer exposure done with and without the filter, or, a new set of sensors, which I don't know if exist and which would lower the megapixels. I also don't know if good ultraviolet sensors exist.

Just engineering problems though.

The next thing is shutter speed. Obviously high-speed photography can see some pretty cool things, like the wingbeats of a bee. I took pictures of bees with my new camera and their wings were a blur. But besides this, really fast photographs would show the flicker in lightbulbs and, especially, fluorescent lights. This would enable me to see something I've long been imagining, the half (or so) of the time when the room is dark because the light bulb is not constantly shining. Good time-lapse video capabilities would be fun too, and would just be a matter of better software.

Better image stabilization would be nice but as it is I bought a camera with supposedly pretty good image stabilization and it just can't cancel out the large motions my hand makes. Understandable. But if you mount fairly massy gyroscopes in it, wouldn't spinning them up provide pretty effective stability? I guess I should just carry my little tripod if I want really high-resolution pictures.

The Internet makes it easy to make sure any question or curiosity I have can eventually turn into reading a fairly detailed account of that thing (right now I'm looking at the Wikipedia article on Mimristors). However, there is no similar guarantee that I use that knowledge for anything, and how can the usefulness of knowledge be tested without applying it somewhere?

What would make it easy for me to eventually use knowledge for something, the way the Internet makes it easy to read up on something?

I've always avoided promoting myself. In high school I went through the normal track rather than going to the math-science center because I would have had to get letters of recommendation. I went to a college I knew I could get into, and similarly fell into my job. But I'm doing a resumé project for an English class and it's making me think about how I'll have to promote myself in the future, in particular to apply to grad school, but more broadly as well.

I'm going to want to set up a website and fill out my personal software portfolio, and list my website when I comment on blogs and forums, maybe even in an email signature. Google myself and make sure everything looks all right.

People who promote themselves get some advantage, which is some part of why everyone's encouraged to believe in themselves, but it's sort of a zero-sum game, isn't it? Though honest self-promotion serves a purpose: employers won't know about you if you don't get your name out there, and the more information you give them the better a decision they can make. But honest humility, applying for the job but listing your faults, would also let them choose more wisely. Humility, however, is quieter, and so people who don't think they're right for a job will not often apply, which makes it less effective. If everyone were too humble to apply for jobs nothing would happen.

I still hope self-promotion eventually becomes less necessary. I guess either the employees or the employers still have to try and get the information out there so agreements can be made.

A couple of days ago I saw a website for sharing code. http://www.openprocessing.org/visuals/?visualID=2356 It has the application for which you were viewing the code, followed by a box with the code. My first instinct was that it should be a wiki, so it would be easy to quickly develop code. I should think the page would include links to the first-posted or latest maintainer-approved version of the app.

This would be pretty appropriate seeing as javascript is becoming so capable. Google builds lots of dynamic interactive stuff in Javascript, though they use their Google Web Toolkit. Now, the GWT compiles code into javascript from Java. It seems like it would be wise for a software wiki to support that language, and others. Maybe it could be treated like the different languages of Wikipedia, or maybe different languages could coexist.

It might also be reasonable to support straight Java, and Processing, the language used on the site tat inspired this idea.

Anyway such a site would bring open source collaboration to a new level.

I know it's sort of going out of style to call onesself by a nickname, in forums and such. I actually got told to use my real name on a mailing list once. But I can't help but hope that by the time I'm looking for work, it is acceptable to put one's internet name on business cards to ease people googling me.

This entry is about a user interface concept, mainly as it applies to Firefox tabs (which is probably currently the dominant task management interface).

I've been thinking about proximity of information. I'm not sure proximity is the right word though; another one would be availability. One reason I think this is important is that it's one of the differences between hard-copy information, ie books and paper, and digital information.

If I buy a book physically, I will put it somewhere, like a bookshelf or a table, where I will see it. This serves as a physical reminder of my intention to read it. Intentions are not the only thing which we use physical reminders for, but they are probably the main one. (Anyone have some other ideas?)

If I download an e-book, typically I will save it to some folder on my hard drive. At one point I left that folder open as a visual reminder to read the books, but reading books is a long-term intention and leaving a folder open is short-term. The folder ended up closed because of a computer restart before I opened more than one of the books.

Maybe I should go look in that folder again...

Anyway, a more practical method for me of leaving something digital out to be read is leaving it open in a Firefox tab. Because Firefox can save sessions, as long as I'm willing to go through the trouble of reloading dozens of websites at once whenever I'm forced to restart firefox (I increasingly am willing) I can just leave tabs open for months.

Opening tabs is a good way to preserve an intention because then when I close a tab I might find the thing in front of me again. I, and probably many people, have the habit of opening a new tab with a google search only to switch back to what I was doing without looking at the results. Or a Wikipedia article. But there are intentions which can't be recorded by these. Another would be opening new blog posts, but this doesn't work to well for me because I don't tend to remember what I was going to write when I come back to it later. I almost always remember why I did a google search, and as for Wikipedia remember what I was wondering often enough that it's fine.

Another way I increase my 'proximity' to some information is with gmail web clips, basically an RSS reader which displays recent updates in a relatively out-of-the-way place while I'm reading my email. This is surprisingly nice, and covers an 'intention' which opening a new tab can't really: the intention to keep up with, or to randomly read, certain RSS feeds.

This makes me think maybe I should have an out-of-the-way place on screen always display information I wanted to be at hand. Perhaps. After all, Firefox tabs can only display so much, and don't change what they're displaying at random.

However, concentrating on how tabs could do this job better, I think it would be really neat if the tab bar were always full. Close enough tabs that you can see them all at once, then close one more, and the empty space should get filled with something you previously intended upon doing. The easiest thing would be something from an RSS feed you've subscribed too, or a special group of RSS feeds you intend on reading every update from. Another possibility would be StumbleUpon pages, which would be relevant to your interests but not actually intentions.

Better would be a method of constantly acquiring intentions. When something isn't a really immediate intention, it could be moved into a reservoir rather than actually remaining open- like those random google searches. Users could be forced to put things in the reservoir by giving themselves a limited (constant) number of actually open tabs, by default the size which fits the screen. New tabs open next to the current one, and older tabs fall off... both sides, or the right, whatever.

I've been adding and deleting Google alerts all day (well it didn't take that long), and though I consider them really useful, they're a little bit inconvenient. I decided I wanted to view them in an RSS reader, so I had to create them and then add them to the reader, or delete them and then remove them.

I haven't tried more complex Google queries in them. That might be interesting, to use them to search within a particular site for updates or something.

What other really special-purpose computers are there than drawing tablet laptops? I mean, both special purpose and general purpose. Oh yeah, smart phones! So could there be a drawing tablet cell phone?

Another random thought, I really wish I could see which firefox tab was taking up my processing power or memory when my computer starts getting bogged down. Does Google Chrome do this, since it has a process for each tab? (I am finding no add-on for doing anything like this, unsurprisingly.)

Aaaanyway what I was going to write about was the idea of Reading Everything. This is something I've been thinking on a good while.

People are socially responsible to read their emails fully... well no, not really, plenty of people sign up for so much stuff that it isn't possible to read everything, but it's still considered pretty necessary to at least read all the subject lines, of stuff considered not spam, so that it's possible for people to get in touch with you. But people who are 'internet famous' or really famous just get too much email to deal with.

This is an example of I think a wider truth about the information feeds we try to 'keep up with'. It seems to me that as information becomes more prominent in our lives, we inevitably find that we cannot keep up with everything we're interested in. We learn to skim for either the most interesting or relevant information, or we have something sort out the junk for us, like with email.

But many of us still hold on to some things which we really fully keep up with. I've read every slashdot story for probably two years or more... I read the subject lines of all my email, except for the ones which gmail sorts out into folders for when I have more time.

I believe in being thorough, I always like to read something completely (I don't like it when sites hide comments from me, even though I almost never can finish reading comment threads). Yet it looks to me like thoroughness is quickly becoming a matter of degree; in more and more ways it's impossible to read everything. I'm hiding behind arbitrary distinctions, reading Slashdot because it doesn't give me more than I can read. Ultimately this sort of strategy makes me tend to read things that are shorter, webcomics instead of blogs, news summaries rather than more thorough reports.

So how should I try to stop the habit of keeping up with the things I read? By reading more things, I think, and less thoroughly- which requires more effort of me to determine which things are most valuable to read, or really, amounts to putting forth the effort continually to decide what to read thoroughly rather than assigning individual sources of information the status of worth fully reading.

I want to be able to look at the data Google has about me.

I want Google to tell me what I am interested in, to double check my own estimates.

... I want a StumbleUpon button for my google alerts, and I want to be able to 'stumble through' pages I've previously bookmarked on delicious, and tags I've used. I want it all in one button, and I want this button to also bring me back to comment threads I've commented on or forum threads I've written in, and my old blog posts.

egeo scribere!

I feel I have a couple of consistency issues to tackle regarding my use of some technology.

Firstly, I'm all for external tools which augment intelligence yet I actually tend to try not to use them in order to exercise my own mind's efforts.

So the question is, when should I draw the line and be efficient rather than treating it as practice?

I guess I don't know. First off, though exercising my own mind seems like an admirable goal, I'm exercising it in a way it wouldn't be used when it came to being more efficient; secondarily I'm making practicing take longer by doing it the slow way! So, well, maybe I should practice using external aids well rather than practicing remembering things or figuring things out internally.

In actuality what I'm talking about amounts to having a better system of writing down things I should remember. I don't take notes in class, for example, and taking at least rudimentary notes would not be too much effort. I also could use more aids when playing a game like Sudoku- taking notes on conclusions &c.

I would not give up walking to school, though driving would save time, because I think the exercise there is necessary. I do not do anything else much for physical exercise than walking.

A similar issue is asking other people for help, though this is more complicated... I do like taking classes rather than just trying to study things on my own, but I wouldn't usually think of getting a tutor when I'm having trouble in a class (um, this is pretty rare which doesn't help, well it doesn't help my learning methodology.), or even studying with people. When I think about learning something on my own I also don't think about tutors or private teachers etc. But these sort of aids are actually really helpful, I learned far more Latin when I had someone going over my homework with me.

Secondly— It appears I've forgotten the second thing. Just goes to show I don't use external memory aids enough. Fact is I've delayed writing about it for two days, because I've been doing other things, namely reading about Minesweeper.

I started programming a minesweeper solver a couple weeks ago. It appears most cases are solved by looking at just individual squares, and a good portion of the rest by looking at two adjacent squares. However, minesweeper is actually much more difficult than that would suggest, and it can be possible that to win without guessing, the player has to look at very large portions of the board. (But more often, it's impossible to win without guessing.)

I often ignore easy clues in Minesweeper and try to develop solving methods which will get me the most information just from looking at what information I already have. (Often opening more squares immediately gets you enough information to solve a formerly hard-looking area with only simple reasoning methods, whereas reaching all possible conclusions from only present clues can involve pretty sophisticated reasoning.) (Should I play more efficiently by opening squares for more clues immediately, or challenge my mind more by not?)

Anyway, my goal was to make a Minesweeper game which always generated a solvable game, and furthermore threw out games which were easy, always generating something requiring more than just local reasoning.

This would bring minesweeper, which I already enjoy, closer to the level of Sudoku for me – Sudoku being good because it's always solvable; losing is directly due to a mistake and is in your control. Sudoku isn't inherently like this, mind you, but rather it has become the practice to always filter out cases which can't be won without guessing, cases with only one solution.

However, Sudoku may be inherently easier to filter like this. Actually they are pretty difficult to compare.

Solitaire is another game I'm interested in the properties of... I'd really like to look at the computational difficulty of generating and of playing various games, and of determining solvability. It seems like a more general theory of when games are fun should exist.

I say, there should be a Mnemosyne plugin which allows one to view websites at intervals!

Now, Mnemosyne is a flashcard program based on theories of long term memory and forgetting curves which helps the user to review information for optimal memorization. I am not suggesting that I want to memorize various websites.

But, what I am suggesting is that sometimes I bookmark websites or even just leave tabs open intending to look at them again. I even read pages and intend to look at them again because I want to do something such as contribute. And I think it would be just groovy to have the revisiting automated. Actually I want it to be integrated into the StumbleUpon button.

Um, yes my idea is a little incoherent. I still think it's sorta important.

I constantly write random things, when I'm pretending to be taking notes in class or just sitting at home, about various things I never feel like writing about here since I'm never sure what audience I'm writing for here. Sometimes I feel like I should start separate blogs for separate interests like my Latin software, the swarm research I'm currently doing, philosophy, etc. But I wouldn't really have time to write regularly on them.

So while I'm indecisive about giving up rambling on paper for the sake of rambling in a publicly available way, I still am going to write about some things I normally wouldn't write about here.

Number theory does a lot of studying when numbers are divisible by each other. The most common real-world application of divisibility seems to me to be sharing a number of items evenly between some people. But there are more shades of possibility here than something being divisible or not; something can be 'almost' divisible by five, in that if there were one more or one less it could be distributed evenly. The normal number theory notation would say n = 1 mod 5 or n = -1 mod 5. (n = ±1 mod 5, if that symbol shows correctly.)

The first thing that comes to mind is that everything is either divisible or almost divisible by 2 and 3. So there are no 'almost-primes'; that is, no number has no almost-divisors and no divisors except one and itself.

But there would be other ways of formulating almost primes. I will say a-divisible means either divisible or almost divisible. An a-prime could be a-divisible by 1,2,3 and itself. In fact I'll be more generous, and say it's also a-divisible by the numbers one above and one below it; sort of seems symmetrical.

Hard to say whether 1,2, or 3 are a-prime. Probably not. 4 is, probably; it's only a-divisible by 1,2,3,4, and 5. But it's the only a-prime with only 5 a-divisors. 5 is a-divisible by 1,2,3,4,5, and 6. All numbers above 4 are a-divisible by at least 6 numbers.

The next a-prime is six, and there are no more. All numbers are only one away from a number 't' divisible by 2; if they are above 6 then t/2 is a divisor above 3 and so there are more than 6 a-divisors.

Well. How about we throw in this number, say a-primes have 7 (or less?) a-divisors? All numbers are still a-divisible by 3, so there is a number t one away from them where t/3 is an additional divisor. Say, how about 8 a-divisors?

Now it seems like there's a chance, but is there? OK, with 8 or less a-divisors (by now I've written a program to list them) we get 1 2 3 4 5 6 7 8 9 10 12 14 18... 22 38 ... and (looking at all numbers below 200) 58, 158, 178.

Interesting that they are all even. If it weren't getting late I'd try to prove that there are an infinite number of them. But in what sense do they build other numbers the way the primes do?

I'd also try other formulations of almost-primeness if I had time. What happens if you count the number of almost-factors not including actual factors?

For now, here are all the ones below 2000 and the hardly efficient common lisp code which produced them.

(defun 1-n (n) (cond ((eq n 1) '(1)) (T (append (1-n (- n 1)) (list n)))))

(defun a-divides (a b) (apply (lambda (x) (or (eq x 1) (eq x 0) (eq x (- a 1)))) (list (mod b a))))

(defun a-divisors (n) (remove-if-not (lambda (x) (a-divides x n)) (1-n (+ n 1))))

(defun num-a-divisors (n) (length (a-divisors n)))

(remove-if (lambda (x) (> (num-a-divisors x) 8)) (1-n 2000))

(1 2 3 4 5 6 7 8 9 10 12 14 18 22 38 58 158 178 382 502 542 718 878 1202 1318 1382 1438 1622 1822)

And hopefully later (but probably not) I'll write about the other random things I considered for this entry. Actually I can't really remember what they are.

I feel like I have a tendency not to finish projects I start, though I blame it on being a student and so concentrating on other things. But anyway.

Last semester I was working on a Latin parser for the purpose of identifying sentences with particular words or grammatical structures. Parsing Latin would be almost crucial even for the task of identifying particular words because there are situations where you need to look at grammatical context to determine whether you are looking at a verb or a noun.

However, doing computational linguistics / corpus linguistics research for my Old English class (taught by a computational linguist) I found that of course there were Latin corpora which were already parsed, so I should be able to write a program using that to do what I wanted- which is, retrieve example sentences for study of particular words or concepts.

Instead of doing that, I started on a simple program to quiz on Latin numbers, which I've always thought I should do. It generates random integers in a range 1-2000 (you can set a different range if you want) and asks you to write them in Latin. However, I've started adding to this program all the underlying framework ideas I've been thinking my ideal Latin quizzing software should have, and it's going really well but I should really make a useable interface for it and release it somewhere.

Besides this, I should make a website of what software I've already made- the Mac widget Latin calendar and the verb quizzing. I guess I'll do that today.

A couple of things I think there should be buzzwords for:

-- discovery: 'discovery' is obviously not catchy and doesn't sum it up, but the idea is things like Stumble Upon, Pandora radio, and netflix where some algorithm tries to randomly lead you to things you like or are interested in. Personally I think the idea is pretty widely applicable, for example I wouldn't mind being able to be randomly taken to a previous bookmark I've made out of my delicious bookmarks.

-- artificial fiction: There are tons of random name generators out there. That is artificial fiction at its simplest, creating a plausible name randomly. There are also programs for creating pronounceable combinations of English letters, as a way of automatically generating nonsense words. This can be useful for naming fictional countries or creating bits of fictional language. There is a program called Dramatica which can randomly generate characters and a plot outline; that's pretty good artificial fiction, though it's about the story rather than world building. I've always thought about software for designing realistic fictional worlds, continents, climates.. and recently I've thought about software for helping generate fictional languages. You could randomly generate a basic dictionary and then create random grammatical rules within some simple realm of possibilities based on variety of real-world languages.

I can imagine serious authors of the future using artificial fiction software packages to help them make rich, realistic worlds... though of course the idea seems pretty funny.

-- there was another one I thought of but I forgot it.
EDIT: Ahh, the missing one was: integrated recall / association. The one thing that makes storing information in the human brain better than looking it up when needed is being able to make connections between pieces of information spontaneously. This is often needed in conversations; we'll suddenly think of something relevant to say. We can sort external information and record what we can think of it being relevant to, but all those relevancies are produced by us, by our brain's ability to make those connections.

Computers can do everything but this, as far as storage and retrieval of information goes, so we should always focus on minimizing the time humans do other things, and maximizing what factors make this association possible. Store lots of concepts and understanding in the brain, lots of raw data on the computer. And get the most out of connections by letting them be stored and shared.

People have been making cheaper and cheaper ways of ... well, there are the microscopes on a chip, the cell-phone-camera-blood-screening, LED flourescence microscope... all way cheaper ways of viewing the very tiny. What I want to figure out is a cheap way of hearing higher and lower wavelengths of sound, and viewing higher and lower wavelengths of light.

Now, a little googling reveals that most digital cameras can see in infrared.

http://www.youtube.com/watch?v=PMpgeFoYbBI

Just point a tv remote at your camera and you should be able to see the flashing light that your tv sees.

But can cameras see ultraviolet? I suspect they can a little... I couldn't be sure testing with a blacklight-looking lava lamp and blacklight-looking LED-- I need to find my actual blacklight, which is lying around somewhere.

If I can manage to filter out just the visual spectrum I should be able to get a nice infrared + ultraviolet pic.

Toying around with black-looking plastic, it appears the difficult thing will be simply determining the quality/properties of various filters. Also, attaching the filters; my camera doesn't have much of any good spot to connect them. But, anyway, the plastic off the front of a random remote definitely filters out most light & doesn't filter out the remote's light, but I can't tell if the rest of what I'm seeing is just bright enough to get through or is actually infrared.

On a unix system, to read a randomly generated text file type at the command line

less -f /dev/urandom

There is a phenomenally small chance you'll get something meaningful. :)

I'm still a beginner at emacs, even though I use it constantly for programming. But here are some useful, if basic, customizations I've done to use it better on my XO.

First, I set my background to white and font to black. I know the other way around is sometimes considered easier on the eyes, but whatever. I like it better this way.

In my .emacs:

(set-face-foreground 'default "black")
(set-face-background 'default "white")

Actually this somehow ends up setting the background to grey? I use emacs on the tty (ctrl+alt+'world') and it doesn't seem any background colors are allowed to be at 100% bright.


The cooler thing that I figured out was how to enable complex key sequences on the 'game keys'. I use the screen rotate button, which apparently sends "/C-@" (equivalent to control+@) as a prefix key by setting (once again in my .emacs)

(global-set-key "\C-@" ())

This sets it to nil so it can be combined with other keys. The game keys are "\C-[[5~" for circle, "\C-[[1~" for square, "\C-[[4~" for check mark, "\C-[[6~" for X. These are page up, home, end, and page down, respectively. I still want those movement functions from them in tablet mode, so I add

(global-set-key "\C-@\C-[[4~" 'next-buffer)
(global-set-key "\C-@\C-[[1~" 'previous-buffer)

so that I can also switch between buffers with rotate-check and rotate-square. I haven't decided what to use circle and X for yet. For now they're

(global-set-key "\C-@\C-[[5~" 'other-window)

(fset 'switch-to-last-buffer "\C-xb\C-m")
(global-set-key "\C-@\C-[[5~" 'switch-to-last-buffer)

How I figured all this out was through the keyboard macro system. I typed "M-x(", meta-x open paren, and then hit the key I want to know the code for. Then "M-x)" then M-x name-last-kbd-macro. I give it some random then M-x insert-kbd-macro . I get something like (fset 'meh "\C-[[4~") and then use the part in quotes. Or in the case of switch-to-last-buffer above, I keep the fset in my .emacs and simply use the new macro.

There is probably a better way of checking how to specify a particular key, but I don't know it.

Tags: emacs, lisp, olpc, xo

After what seems like it must have been two weeks, I have internet again. In celebration I googled 'cats' for the first time and watched videos of cats sleeping and failing to jump far enough to reach their target.

A surreal thought: what if all your passwords mysteriously stopped working? Or worse I guess, you couldn't remember any of them.