[ Home page | Web log ]
An interesting piece about ID cards in today's Telegraph (may require bullshit registration to view). It reports the results of a survey by internet polling outfit YouGov; sadly, their website doesn't have a copy of the report, so the only copy of the results are presented as an image on the Telegraph website; again, it's obviously vitally important to keep this information from getting into the hands of blind people. (It's worth bearing in mind that internet polling is probably not very reliable on social issues like this; here, the errors probably underestimate support for ID cards.)
(Inevitably, the article is illustrated with a picture of two young ladies, who in this case are holding up identity cards and are apparently Germans, but may as well have been Britons holding up printouts of record A-level results. `© All Newspapers', as Private Eye are fond of saying.)
The striking thing about the government's ID cards policy is that they have no idea what the things are for. They don't even know what the card will be called. This confusion is reflected in the attitudes of the public, as measured by the YouGov poll. Here are the responses to a question about the preferred type of ID card scheme:
|Preferred ID card scheme||Percentage of respondents|
|compulsory to carry at all times (and presumably to present on demand)||39|
|compulsory to own card, but not to carry at all time||42|
|optional to own one||18|
And here are the responses to questions about the applications of a card:
|Objective||Percentage believing cards would help with objective||Percentage not believing cards would help with objective||Don't know|
|cut down on `health tourism'||78||14||9|
|cut down on benefit fraud||82||13||5|
|make it easier for police to catch criminals||60||26||14|
|make it easier for police, other officials to catch bogus asylum seekers, others attempting to avoid deportation||80||14||6|
Concentrate on `health tourism', meaning `eeeevil foreigners getting treatment in the NHS'. Ignoring for the moment that this isn't a significant cost to the NHS, that most of those branded `health tourists' are perfectly entitled to treatment, and that the `health tourism' issue is just another bit of production-line tabloid xenophobia whipped up by crank pressure groups like Health Watch UK, we should look at how ID cards would be used to tackle this `problem'.
Obviously the idea is that in order to obtain health care you must have an ID card, and must be carrying it when you go to hospital. No ID card: no treatment. (Ignore for the moment the practical and ethical obstacles to this, and try not to imagine being in a car crash where the emergency services aren't able to rescue your ID card along with you....) Approximately 80% of people believe that using the ID card in this way will cut down on `health tourism'. But only half that number believe that the ID cards should be carried at all times. Isn't that strange?
Even more bizarre were answers to the question, ``If there were identity cards, which of the following groups should/would the police and other public authorities target?'' 37% of people fairly reasonably concluded that the authorities would use the cards to harass `foreign-looking people', and 46% that they would harass `young people' -- but 6% of people thought that foreign-looking people should be targeted! Truly we are living in the multicultural society.
It's interesting to contrast idealism about the applications of the cards -- 60% of people accept uncritically that the card will `make it easier for the police to catch criminals', but there's no mechanism by which it would help in general -- with cynicism about their design. 66% of people correctly believe that the cards will be easy to forge; 72% of people that the confidential data on the cards will be passed to unauthorised people, and 63% that the cards will contain additional information not needed for their real function. Slightly strangely, the survey didn't ask about identity `theft', which the cards will almost certainly make easier.
98% of people think that £40 is too much to pay for a card -- note that if the implementation the cards is as successful as typical government IT projects, the real cost is likely to be between £120 and £400. (Of course, that's unlikely to be the sum you send off to Miniluv; most of the cost will, quite properly, be financed through general taxation. But let's not kid ourselves about the amount. The 86% who think the card should be `free' are, if they think a card should be introduced at all, just asking for a substantial increase in their income tax bills.)
The Telegraph identifies ID cards as, potentially, `Labour's poll tax'. This may not be so far from the truth; for once, the Conservative policy is actually less offensive than Labour's, with Oliver Letwin even warning that a current ID card pilot scheme may be the thin end of a wedge. 'Course, if they ever got into power, they might well put somebody rather more extreme in the Home Office, and in any case I did hear Letwin on the radio earlier trying to keep up his credentials with the Blunkett-Jugend by advertising the report of the `independent' (i.e., Conservative) Kirkhope Commission, which proposes yet another plan to put asylum seekers in concentration camps, this time in far flung `safe' locations or off-shore. No social liberal he. There's no reason to suppose that a hypothetical future Conservative government wouldn't change its mind on the issue and I see no evidence that opposition to ID cards will be a Tory manifesto commitment.
(And while I mention it, how come nobody's pointed out that having a .com domain may not create the impression the Conservatives are after?)
Pete points out another oddity from the poll. Consider two questions:
Are you in favour of, or opposed to, the introduction of national identity cards?
|Response||Percentage of respondents|
Taking everything into account, which statement comes closest to your view?
|Statement||Percentage of respondents|
|I would greatly welcome the introduction of identity cards||34|
|I would not mind their introduction||45|
|I would object to their introduction but would reluctantly go along with them||13|
|I would strongly object to their introduction and would absolutely refuse to acquire one||7|
If we interpret `would not mind' as `am apathetic', we have:
|Opinion||Percentage of respondents|
|Actually want the cards||34|
|Oppose the cards||20|
-- a rather different picture.
So, I shouldn't just propagate links, but whatever....
Alex has finished his new short film, Staying Pictures. Very cool, so drop what you're doing and watch it right now. Or don't. As you like.
(You'll need quite a modern video player to play the files; xine almost manages, but without the sound, which rather misses the point. Just go and find a Windows machine.... This could turn into a whole separate rant, best not rehearsed here.)
For reasons best not gone into again, I happened to need to get some money transferred from a bank in another European country to my own. While doing this I discovered two things which surprised me:
It reminds me of the last time I had to pay a US dollar cheque into my account. The US doesn't (as far as I can tell) have a small number of banks dominating the market as we do, so people typically hold current accounts with a small local bank. Cheques from these banks are treated by Barclays with considerable suspicion -- they perhaps imagine the bank headquartered in a tin shack in a one-horse town somewhere out west, subject to periodic raids by the Hole-in-the-Wall Gang or some other band of miscreants -- and even if the number of dollars involved falls far short of a fistful, Extensive Security Checks are necessary. (No doubt September 11th has given rise to even more Extensive Checks.)
So they spend an enormous amount of time trying to figure out whether the drawer's bank and money are for real. Naturally, the US bank don't process these requests with any urgency, perhaps imagining that Barclays is the UK equivalent of the First Bank of Dry Gulch, Wyoming and therefore not worth dealing with speedily and efficiently.
Result: six week delay and huge bank charges. Hooray for globalisation!
I malign Barclays slightly. The money took about one week to reach me from another European bank: still about 200 million times slower than the actual electronic signal, but about four times faster than I was initially told. And you won't be surprised to hear that their exchange rate wasn't exactly... competitive.
As a rule, logic puzzles really irritate me (you might have detected that from previous comments, I suppose). But I recently saw one which intrigued me (partly because I was offered cash money for solving it by somebody who'd managed to throw out the instructions...):
The puzzle consists of 16 triangular tiles which fit into a triangular recess in a piece of wood; the side of the recess is four times the length of the side of each triangular tile. Each edge of each tile is marked with a spot of paint which is either red, green, blue, white, yellow or black; along each side of the triangular recess are four paint spots chosen from same colours.
The problem is to arrange the tiles such that any two adjacent paint spots (lying on either side of the edge of a tile) are the same colour. (The photograph above shows the unsolved puzzle....)
This puzzle has a very large number of possible states. In particular, there are 16! (about 2.1×10**13) ways to arrange the tiles, and three possible rotations of each tile in each position, giving 3**16 × 2.1×10**13 or about 8.6×10**16 combinations. This is A Lot.
Obviously you can cut down on the possibilities by starting at a corner, where there is more than one constraint on which tiles can be used, but even so, solving the thing by hand would be a monumental pain in the arse.
(Incidentally, it's worth pointing out that no effort at all is required to make the puzzle-- you just paint the little dots on in place, making sure they match up as they should.)
I quickly became bored and wrote a computer program to solve it--
-- which performs an exhaustive search, starting from the bottom-left tile and working left and up until it completes the board. Producing the solution above required 37,293 separate tests (checking whether a given tile in a given orientation would fit in a given location). Sober, awake and enthusiastic, you might be able to do one test every second -- but after ten hours you might not be performing so quickly or so accurately.
So is there a trick which makes this problem manageable, or is it just stupidly difficult? Answers on a postcard much appreciated.
You'd expect the Torygraph to be good on the Patriotic Pedantry front. Not so. Consider, for instance, its suggestion for a citizenship test (presented as an image in order, presumably, to deter blind people from establishing their ability to become citizens of Telegraphistan) which included the question,
Q. Name the four countries in the United Kingdom.
A. England, Scotland, Wales, Northern Ireland.
-- Bzzt! Sorry, you lose. Wales is a principality; Northern Ireland is a province.
If the Torygraph are going to play the `accuracy' game, they should try a little harder.
Perhaps predictably, Anthony has written something about the Home Office's latest gimmick, citizenship tests for people applying for naturalisation in the UK. I'm testing a new bit of my web log software which is intended to copy links to articles into a Live Journal (don't ask...), and this is as good a subject to comment on as any other....
(I should first say that my immediate reaction -- can we make some of our existing, more disagreeable, compatriots take the exam and kick them out of the country if they don't pass it? -- was (a) uncharitable, and (b) misinformed, since the test is not designed to test eligibility for residence.)
Anthony links to the report of Bernard Crick's panel, which summarises the curriculum in a way that the media have largely failed to do; the whole report is worth reading, especially the first part which summarises current immigration rates and policies in a way that those who are disinclined to look up statistics themselves might find useful. (I was unable to find the report on the Home Office site after a few frustrating minutes of searching. You'd have thought that, of all the government departments, they'd at least make the effort to have an accessible and easy-to-use site, but no, 'tis not to be. Here is a mirror of the PDF file for your convenience....)
I'm divided on this idea. Superficially the idea that new citizens should speak English and know something about our history and culture (and `the NHS' and `how to get electricity') is fairly reasonable and perfectly compatible with an open and tolerant immigration policy.
I'm more troubled by the antecedents: the use of qualification tests like this has, historically, been subject to grave abuse (e.g. voting qualification `literacy' tests in the southern US, which for a century were used to implement the racial discrimination which had been banned after the Civil War). But there's no concrete reason to see the Crick ideas as the beginning of a slippery slope, of course.
The other Big Idea -- a public `citizenship ceremony' in which new citizens make their pledges of allegiance to the Crown -- is superficially appealing; as Blunkett said,
Becoming a British citizen is a significant life event. The Government intends to make gaining British citizenship meaningful and celebratory rather than simply a bureaucratic process. New citizenship ceremonies will help people mark this important event.
... but then again, such a ceremony isn't really very British, is it? We're not Americans, we don't salute the flag, the National Anthem is for when programmes on the wireless finish and we don't, I think, really need
a periodic civic ceremony in which the Mayor or Provost should make a speech of welcome to the new citizens and their families stressing the rights and duties of citizenship, and in the presence of invited local dignitaries such as Members of Parliament, candidates, councillors and leaders of community and religious groups. The Mayor, Provost or a local dignitary or celebrity should then present newly designed certificates of naturalisation individually (which we understand will be far more impressive and suitable to be framed than the current certificate) as the applicant takes the new Citizenship Oath and Pledge.
-- a quiet, bureaucratic ceremony in a registry office would set the right note of understatement and unfussiness by which we should all aspire to live our lives. (On the other hand, a public citizenship ceremony might serve as a welcome humiliation for any future BNP mayor obliged to preside over it. Every dark cloud has, if not a silver lining, at least the possibility of bringing some black humour into our lives.)
Anyway, the report is well worth reading -- especially the bit in the appendix which describes why people want British citizenship:
``Becoming British it elevates my status and I m part of a country that is existent in the map of the world now internationally. If I go to other countries with a Somali passport I might face a lot of difficulties because people do not recognise my country and they associate Somali with corruption and other sort of things but by becoming British I get VIP status when I go to other countries.'' -- Somali man
It's nice to know that, as a country, we're still on the map a little bit. The idea that Brits abroad are accorded VIP status will be met with frank astonishment by anyone who's had the misfortune of visiting a Mediterranean holiday resort, of course....
Roy mentions Shazam, a music-identification-by-mobile-phone service which is -- apparently -- both expensive and addictive. (My phone is on Virgin Mobile, which doesn't peer with Shazam or whatever the technical term is, so I've never used it myself. But my friends alternate between raving about it and complaining about the amount they spend on it at 59p per go.)
How does Shazam work?
(Here I should point out that I know nothing about music, so I'm going to talk about an entire field of cultural endeavour reduced to a vaguely diverting signal-processing problem. Whatever.)
Shazam's internals are, no doubt, `commercially confidential'. But there is some information on possible algorithms on the web. Cheng Yang from the Stanford database group has published a number of papers on a fairly simple music matching algorithm; this 2001 paper and this one from 2002 give a reasonable summary.
The basic game is to compute spectrograms at points of peak intensity in the music; the argument here is that a piece of music is, roughly speaking, characterised by how the loudest bits sound. At each peak in each piece of music, the spectrogram is divided into 180 buckets with frequencies between 200 and 2000Hz, and an entire piece of music is described by a sequence of such 180-value spectrograms, stored with their time position in the song.
Matching is done by computing a similar sequence of spectrograms from the query (as recorded from your telephone or whatever) and searching for similar spectrograms in the database of known songs. The idea is to record t1 = offset-of-match-in-song for each spectrogram against t2 = offset-of-match-in-query, and then assess whether the query comes from the song by seeing whether it has the same peak spectrograms in the same sequence.
A perfect match -- in which the same peaks occur at the same times in the song and query -- would have t1 = a + t2 (since the query can come from any part of the song); a match between songs in different tempos would be t1 = a + b t2. The best matching a and b are found using linear regression in the first paper, and in the (more sophisticated) second paper, using the Hough transform beloved of computer vision researchers to pick up linear features. (The present implementation doesn't search for spectrograms which are similar after a pitch shift, with the result that it won't match music which has been transposed. Doing so would add another free parameter to the matches, obviously.)
The search step is computationally intensive. Searching for matching 180-element spectrograms is not easy. Yang uses an efficient procedure called `locality-sensitive hashing' to make this tractable. (There's a brief description in the second paper linked above, and a Google search for the term yields lots of information -- often from people using the technique for matching DNA sequences, in many ways a similar problem. The basic idea is something like this: pick a large number of random basis vectors v_i, and then for each data vector d record the result of d·v_i > 0 as a hash value h_i. Catenate the results of comparisons against all the v_i as H; if H for the query and H for a data vector match, compute the similarity between the two properly (and slowly) and add the data point to the matches if it is close enough. To find all (rather, most) of the similar data vectors, repeat the above procedure for lots of sets of v_i. There's a clear but much more mathematical description in these lecture notes, which also proves why this actually works, if intuition isn't enough.)
Unfortunately, even the optimised search procedure takes about ten times the duration of the input clip to search for matches; that is, a 30-second clip results in a search of about five minutes (see graph at end of 2002 paper). This could probably be improved using some DSP-tastic custom hardware but for a large-scale application like Shazam more optimisation would be needed to make the searches economic.
With a small corpus (220 piece of mainly classical music) Yang obtains better than 60% accuracy (defined as `correct result among top five matches') for thirty-second samples, with over 80% accuracy when the exact same piece is being searched for (the algorithm works for variants and derivative works, too, but less well; the idea is that, given one piece of music, it could find, for instance, the same piece recorded by different performers). Obviously it's the second case which is more important for a Shazam-type application, since Shazam push you towards buying CDs of the music you hear.
I don't know if this is exactly how Shazam actually works but I'd guess it isn't far off. Obviously something must have been done to improve the execution time. I suspect that if we only really care about exact matches then a less sophisticated search step could be used; in particular, we wouldn't need to match pieces in different tempo and it may be possible to just represent each song as the catenated peak spectrograms and search those.
If all else fails, well, computers are fast and memory is cheap. Throwing money at the problem works for that other information-retrieval success story, Google.
(Two other thoughts occur to me. Yang's algorithm -- at least in a toy version -- wouldn't be that hard to implement for someone with some spare time and a passing familiarity with the fast Fourier transform, and the problem is quite interesting. Yang bemoans the lack of a widely distributed and standard corpus of music to test against. But -- and at this point I am compelled to point out that using it would probably be an infringement of copyright -- the material available through file-sharing networks like Gnutella could provide an extensive and already machine-readable corpus with some descriptive text for each track. Samples from pop music radio stations would be a useful source of test queries, and some stations will provide details of their playlists, which would make automated testing tractable.... Anyway, if you do wander off and implement this to compete with Shazam, I'd be interested to hear how you get on -- especially if you're raking in 59p/query and want to give me some of it....)
So, last Thursday I went to Brussels to lobby the European Parliament against software patents. I'll leave the FFII to talk about that.
Flying to another European capital for the day makes me quite the worldwide traveller, I suppose, which would explain why I didn't enjoy the experience very much.
To get to Brussels, I flew on Ryanair to Brussels South Charleroi Airport, a contradiction in terms like `military intelligence' or `London Luton'. Charleroi Airport is just like a bus station, except with fewer bus services, which was a bit of a pain when it came to onward travel. To get from Charleroi to Brussels you get a transfer coach which is timed to coincide with the occasional Ryanair arrivals and which costs Ten Of Your Earth Euros.
I had made the mistake of assuming that I'd be able to get money out of a cash machine when I arrived, which would have worked fine if Charleroi's single auto-teller had any money in it. As a result I had no cash to pay my bus fare, until I realised -- little helped by the surly bank official, though at that time in the morning everyone is surly -- that the bank was open and was in fact able to change money, but by the time I'd established that this transaction was possible and completed it, I'd missed my bus and had to wait over an hour for the next one.
I suppose this is an argument for the single currency, but it's equally an argument for never getting on an aeroplane at half-past seven o'clock in the fucking morning and being completely incapable of rational thought when it arrives. The one advantage of flying at such a ridiculous time was that I was able to catch a little bit of sleep on the 'plane, which probably enabled me to stay -- barely -- awake whilst ensconced in the throbbing heart of our great European democracy.
By the time I got to the European Parliament (optionally referred to as the `Europarl' in 1984-speak) I'd wasted about three hours, most of the time being spent in the company of a crowd of overenthusiastic American tourists who perhaps thought they were taking their last chance to see Old Europe before Donald Rumsfeld `liberates' it. In any case their verbose enthusiasm did not help me to cope with early morning travel.
The Parliament itself is an enormous building from the `monumental shopping mall' school of architecture: marble and mirrored glass, and security guards in silly hats. That said, it would be hard to claim with a straight face that it's any worse than our own Victorian-gothic monstrosity. Horses for courses, I suppose.
(It is also equipped with at least one statue of unique and -- at the risk of making a value judgment -- horrid design: a bronze depicting a woman wearing a nightshirt holding up an enormous euro symbol while a number of male heads grow out of her free arm. Or at least, that's what I think it's supposed to be:
That said, there's a statue of great parliamentarian and, uh, noted military dictator Oliver Cromwell outside the Palace of Westminster, which is hardly an aesthetic improvement and certainly in worse taste.)
Getting in to the Parliament requires all sorts of searches and X-rayings and slightly baffling interactions with the security guards. It took me a while to determine that the impassioned gesticulations of one of the guards -- let's call him Captain X-Ray -- were intended to convey to me that I'd dropped my keys and should pick them up. (For this service, many thanks; returning to Cambridge only to discover I couldn't get in to my flat would have been exceptionally irritating.) I suppose this illustrates why my general strategy for dealing with security people -- walk obliviously and purposefully away as if I really know what I'm doing and their concerns are no concern of mine -- is not always a good idea. Also they were probably armed. Whatever. (Another tip: sharing a language with the security people is probably a Good Idea. Naturally, I studied French for many years at school, but it doesn't seem to have had quite the educational effect that I'm sure was intended. Here would be the point to insert a long rant about GCSEs and `dumbing down', if I were so inclined.)
The security is, I guess, fair enough, though a cynic might suggest that typical European reactions to news of a terrorist attack on MEPs might be open astonishment followed by apathy.
The Parliament is huge -- our European representatives get two offices each, one of which is allocated to their staff of one or two assistants -- and the building is generously equipped with lifts, photocopiers, and notices imploring MEPs to please not steal any more cutlery from the restaurant. (Each photocopier is also supplied with an enormous stack of forms for reporting `photocopier incidents' -- I guess these must have been for when the photocopier breaks down, though I suppose it's conceivable that a form was supposed to be filled out for each incident of photocopying. The Parliament uses so much paper that it has its own brand....) From the upper floors of the building you get a fantastic view, but unfortunately it's a view of Brussels, which seems to consist chiefly of building sites. The old joke--
Q. How many people work in the European Commission?
A. About half of them.
certainly can't be true of those who are planning new buildings for Eurocrats....
I won't bore you with any details of the actual process of lobbying; suffice it to say that this was quite stressful, especially since the majority of MEPs seem to take the end of the week off and trekking around the building only to find them absent is a pain. I wasn't able to see any of my constituency MEPs, for instance, which was pretty poor going.
Returning to the UK was relatively painless, largely I think because I had time -- finally -- to get a coffee before catching my flight. (By that time I was acutely aware that my general crankiness was probably connected to the fact that I'd had no time for any caffeine all day. Bad move.) Happily my plane landed after the recent blackout and I expect Stansted wasn't affected other than peripherally anyway.
My recommendations: if you want to see the throbbing heart of our great Euro-democracy, go to Strasbourg: at least that's a pretty town. And don't just go for a day. And don't economise on the coffee....
Suckdot extracts the links -- the only even occasionally useful content -- from Slashdot, so that you never have to see the offensive colour scheme or inane, stupid commentary. Also available as an RSS feed for those using aggregators.
Surprising RSS aggregators, unlike most of the technical contributions of the ``'Blogosphere'', are actually quite useful, though they are hamstrung by the crappy specification of RSS and the fact that typical web site administrators are too lazy to generate correct RSS anyway. Part of the problem is that RSS must be XML, because that's the current fad; a more sensible format would have been plain text with lines of year, month, day, hour, minute, URL, title, description. Even the most doltish of web log software authors could hardly get that wrong.
Anyway, here's my headlines aggregator, which I've been meaning to package up for a while. It emits HTML. See the README in the tarball.
This is all done with wwwitter.
Copyright (c) Chris Lightfoot; available under a Creative Commons License. Comments, if any, copyright (c) contributors and available under the same license.
Hosted and supported by