« January 2005 | Main | March 2005 »

February 28, 2005

20th Century Spinozist : Rancière on Deleuze

As a companion piece to Infinite thought's review/report of Jacques Rancière's recent London lecture, I thought I would post a quick (=possibly flawed) translation of the short interview with Rancière which appeared in Magazine Littéraire's 2002 issue dedicated to Deleuze.

Rancière takes a unique and fascinating approach to Deleuze as aesthetician, focusing here on a characteristic tendency in Deleuze's aesthetic writing to apparently collapse the regime of representation whilst simultaneously using figures drawn from specific representations, or certain fictional traits, as abstract models for his analysis.

I remember being struck on first reading of this interview by how well this describes an aspect of Deleuze's book on Bacon that had bothered me vaguely: the way in which 'abstract machines' such as that of the triptych, or the observer, are said to organise the paintings whilst not corresponding to their figural equivalents (the 'actual' observers, the 'actual' panels of the tryptych.) This can be hugely puzzling, in that one asks (and the question is the same in many other eccentric instances in A Thousand Plateaus) how it is that these machines, supposedly operating on a level anterior to any figuration, nevertheless seem to be modelled after certain figurations, or "fictions"?

Rancière suggests (but very schematically, given the brief compass of this interview) that this movement and its tension, is a symptom of a sort of inevitably falling back onto the plane of radical immanence of the figuration, and a necessary symptom of Deleuze's attempt to consummate the end of representation and simultaneously 'fulfil the destiny' of the aesthetic.

Now this simultaneous disavowal of representation and inveterate dependence on allegory or metaphor can be read as nothing more than a dissembling, whether conscious or not - and this is arguably the position Badiou takes on Deleuze in his Clamour of Being (ie. that Deleuze claims to have 'done away with' metaphor whilst constantly using metaphor to back up his ideas). Although Badiou recognises as constitutive of Deleuze's thought the movement from the 'simulacra' to radical immanence, and the peripatetic narrative Deleuze makes of this movement, he apparently fails to see anything in this movement but a failure of philosophical probity...

A problematisation of this approach is consequent upon Ranciere's more general complication of the definition of the post-representative regime of art: it is defined not by the autonomy of the art object, a "freedom" which would be cognate to that of the radically free-floating sign of postmodernity (the very blithely-assumed unconstrained 'freedom' that is Badiou's nemesis), but a mixed regime in which the artistic will is bound to a certain automatism. In which, we can assume, the escape from representation is never achieved, but is expressed in a constant becoming.

(One last brief comment; this recourse to a falling back into representation, to a constant movement between absolute and relative immanence also connects with a very difficult chapter on Derrida I am currently trying to read, from Laruelle's Philosophies de la Difference. Laruelle argues that a philosophical decision is always made to set up such tensions, which in the absence of an interrogation into this founding decision can constitutively never be 'resolved' but rather act as the 'paradoxal motor' that drives the interminable philosophical peripitaeia.)

Posted by robin at 08:24 PM

February 22, 2005

Laruelle/Derrida

Controversy over the Possibility of a Science of Philosophy: Jacques Derrida and François Laruelle in conversation

Now with the proper title, and with the benefit of some substantial corrections and a great deal of syntactic defrancification, this should be a far more readable version of the fascinating debate between Derrida and Laruelle.

- Many thanks to Ray B. for taking the time to review the awfully hasty first version of this translation and thereby improving it by an order of magnitude.

Posted by robin at 08:36 PM

tic toc ya don't stop

What would Badiou's ontomathematics look like if it was turned inside out, subjected to xenoumenal pragmatism, and run through the hyperstitional mangle? Something like Dr Daniel Barker's Tic-Xenotation?

Posted by robin at 02:41 PM

February 18, 2005

Freedom of disinformation act

Although I still hold out the hope of holding a more indepth postmortem on this project, after recent discussions at hyperstition, and given that no-one can suggest anything particular to do with it, I have decided to release the burgeoning glossolalary data as open-source numbojumbo.

mulch1.csv (460k)

Posted by robin at 10:20 AM

When non-philosophers party

A non-conference is neither an anti-conference, nor the absence of a conference

[Warning:Images may cause offence]

- additional footage just in from sphaleotas

Posted by robin at 10:18 AM

February 17, 2005

On the Trail of the Oracle

I devised a fascinating but simple experiment with Google, featuring a recursive mapping through the google-database that constitutes a sort of monstrous misapplication of Vitanyi and Cilibrasi's intuition about the database's latent semantic properties. Essentially, the aim is to use the database in a way which is rigorous but does not presuppose the utility of the experimental outcome. The function I describe maps words to series of words using the Google database in a very rough-and-ready way.

Some technical parameters need to be set to control the experiment properly, and I will post a program to automatically calculate the output, but for now you can try it manually.

1. For the purposes of the experiment, we limit Searchterms to single words. We understand these Searchterms as numbers (which, of course, they are).

2. By the Trail function, or T, we designate the function which maps one Searchterm number to a larger number by the following, extremely simple algorithmic process:
- Search for the Searchterm (using syntax +"searchterm");
- Each main Google result consists of a Title (shown in blue) followed by what is called a 'snippet', a small extract from the page. Take the first word of the first snippet;
- Using this first word as the new SearchTerm, iterate the process;
- T for SearchTerm s is defined as the concatenation of the whole series of results, and may be finite or infinite.

3. As will be obvious, T(s) will be one of several types, comparable to types of real number;
- Numbers which terminate (when a SearchTerm fails to return any results)
- Numbers which cycle (when the last of a series of SearchTerms outputs one of the previous SearchTerms in the series)
- Numbers which recur on one term (when a SearchTerm returns itself) (this is a special case of the above. In any case, a SearchTerm-as-number really represents a multibyte binary number).

4. The first experiments with this method demonstrated the existence of recurring trails, but as yet no cyclical or terminating trails. Several observations follow:
- It is not necessary the case that terminating results for T exist since this would mean the existence of a word in a snippet that was not indexed for searching (this is a technical question concerning the Google database structure). However there is no reason why we should not find cyclical terms.
- the progress of a Searchterm along the series T owes very little to common semantic usage, because of the 'arbitrary' selection of the first word (arguably the result owes more to large-granularity artefacts than anything else - however it does represent all the same a thoroughly rigorous mapping)
- as a corollary to this, words which are not at all cognate or related in 'normal' usage converge to the same attractors.
- One of the easiest identifiable groupings of 'basin attractors' for T, which result in one-place SearchTerm recurring numbers, are tradenames. It is likely that once you reach a tradename (at least, we might say, if the e-marketing dept. has done its job properly) you will never escape.
- In real language some words, we might say, represent 'dead ends'. Some have a limited semantic connectivity. Some ('a','our','is') can connect to almost anything. This hyperconnectivity is heightened by the 'arbitrary' use of the first word of the first result. This could obviously be modulated in various ways but the present experiment provides an interesting and simple limit-case.
- There seem to be a very limited amount of attractors - we were surprised at arriving in the same 'loop' several times within the space of a very few experiments. It should be possible via experimentation to map the 'landscape' of T.
- The Google database is a dynamic entity, so this landscape is not static.

More to report later, but for now I leave you with a few more examples., please post other interesting results below.

Tradename trail basin:
know iknowthat.com iknowthat.com ...
london discounts travel 3 welcome whitehouse.gov whitehouse.gov

Genre 'ownership' through trail:
porn penisbot's penisbot's ...

Massive coverage through 'ownership' of a trail of common connective linguistic parts:
test cleveland governor's homepage The A our anne designer all alltheweb
alltheweb
calculus tutorials the a our anne designer all alltheweb alltheweb ...
robin us the a our anne designer all alltheweb alltheweb ...
ruth home the a our anne designer all alltheweb alltheweb ...
randomness writings eben academia next excellence copyright US The our anne designer all alltheweb alltheweb ...


Unexpected cognates (no comments please ;) via interlinguistic trails:
smell enter magazyn miesiecznik peryskop peryskop ...
dread enter magazyn miesiecznik peryskop peryskop ...

'Capture' of one tradename by another:
hyperstition further furthurnet furthurnet...
norton more in you free nedstat nedstat

'Capture' and partial escape into a subcategory:
lovecraft dedicated UK Yahoo! Finance Finance ...

'Singletons':
eleutheria eleutheria ...
liberty liberty ...
apocalypse apocalypse...
derrida derrida ...

Cognates, quickly-settling recurrences:
roulette outlines executable executable ...
anybody executable executable ...

Posted by robin at 01:01 PM

Nothing ever died...

(continuing the discussion from Hosegate, yeah?)

ANALYSIS: Dissection and full enumerative description of a system, with a possible view to exposing the facility of its robotic reproduction, and hence its inherent worthlessness beyond quantifiable production costs. Literally, dissolution. (nb, now synonymous, by a cruel twist of synthetic etymonomy, with the contemporary popular sense of 'deconstruction'. Cue JD spinning in grave.)

IRONY [1]: The reproduction of a given set of behaviours in the form of mere semiotic routine in such a way as to demonstrate the possibility of detaching them from any external epistemological commitments, or reversing their affiliations with such. Marking the de facto obsolescence of such commitments as vertically-integrated effective factors in the production of sense. Reflexive decoding.

PARODY:Intensification of a given trend to extents inevitably "absurd" according to common-sense, so retroactively positioning said trend as latently meaningless, arbitrary, and "unfounded". Efficient exhaustion through unregulated overexploitation.

Q1. Spot the inherent problems in using the above as tactics _against_ the generalised decoding apparatus of capitalism.

IRONY [2]: The simultaneous existence of two apparently antagonistic tendencies. Rarely fatal.

Posted by robin at 11:21 AM

Of the Refrain: Tramp with Orchestra

(maybe this will make it up to JC :)

I am eternally grateful to Robert Elms for introducing me to Gavin Bryars' Jesus' Blood Never Failed Me Yet. To hear this piece of music, so extraordinary in itself, emerge from the radio one morning was a truly astonishing experience. Elms apparently made a principle of playing the piece a couple of times every year, and the station was always inundated by listeners phoning up in tears asking for the title of the piece. Hearing it was a "life changing experience", not in the sense that it led to any momentous decisions or sudden realisations, but simply that it divided time into a distinct before and after, hearing it altered you somehow. It was not actually for some years that I found the CD, and earlier this year Eleutheria and I were lucky enough to see Bryars conducting a live performance of one of the very few pieces of 20th-century orchestral music that have touched me in any way.

In 1971, Bryars had isolated a small loop of tape from the offcuts of a documentary he had done the sound for, about the vagrant population who lived in the ramshackle 'cardboard city'in the centre of the roundabout at Waterloo. This, of course, was, until they were evicted and replaced in the centre of the roundabout by an Imax cinema, an architecturally null steel-and-glass number, and almost always empty of custom, a real symbol of (as the marketing has it) LondOn (presumably capital of Cool Britannia).

Having left a tape loop of one of the tramps giving an impromptu rendering of a (still unidentified, and possibly extemporised) hymn, playing in his studio, he returned five minutes later to find

the normally lively room unaturally subdued. People were moving about much more slowly than usual, and a few were sitting alone, quietly weeping. I was puzzled until I realised that the tape was still playing and that they had been overcome by the old man's unaccompanied singing.

It was obvious that, if his idea of creating an orchestral piece around this repeated phrase was to work, it would have to be treated with great sensitivityin order not to overwhelm whatever subtle magic was frozen forever in the medium of those few seconds of magnetic tape.

Posted by robin at 10:44 AM

February 12, 2005

Hosegate, yeah?

The long-awaited (and by now somewhat anachronistic) Nathan Barley. Haven't yet reached an opinion as to whether it's funny/better/worse than x/y/z Morris product because still in shock from its horrible, cringe-inducing, will-to-exist-destroying accuracy. I'm not even sure, given its purely documentary importance, that its timeliness, funniness, or the inevitable sold-out/softened-up accusations matter much.

Yes, it was that bad (no doubt it still is, in pockets where economic correction has yet to take its toll), I have met all of those people and various hybrids of them personally. I could even name them*, but as Morris&Brooker show in forensic detail, the Idiots are unshamable, their hegemony is absolute within the bounds of the Hoxton/Shoreditch Triangle, and its tentacles spread outwards via the multitude of sad cases desperate to secure some 'edge' through desperate overbudgeted frottage with these loons (let us note that some of them number amongst the staff of the production companies Morris has contracted to supply the 'edgy' visual 'treatments' of the series - ironic, yeah?).

How do you satirise Ultimate Style-Man? "They say 'self-facilitating media node,' and blink".

The respective plights of Dan Ashcroft and Pingu are devastating in their poignancy. Just as Morris's savage newsmedia parodies were always outdone by the real thing, these characters painfully reveal the impossibility of combatting such hyperbolic levels of insular arrogance and wilful superficiality as those produced as secondary froth on the dotcom bubble. The first is a terminally-recuperated theorist of idiot-decadence trapped in the world of cool (he knows what's happening, but no-one is listening, he fears absorption but the only other exit is to renounce cool), the other an able, sincere-but-mild, code-drone slowly crushed to death under the ironically exploitative heel of the braying, superconfident incompetents.

Click it and weep.


*let's just say that a certain company's boardroom table was made out of a salvaged wing of a biplane and the director made a point of wearing skirts to all meetings. And that, being as I was employed by a corporate entity desperately seeking approval from the new ruling class of fashion-victims, I saw suits melt with ravenous delight and bust open the coffers at every swirl. Another major offender was a still-extant self-facilitating trustafarian goon who was paid richly by the hour to come and sit around in the offices of a certain dotcom company drinking coffee and exchanging hootingly loud excruciating faux-"street" banter with a hapless public-school developer.

Posted by robin at 02:18 PM

Decaplex Live

Decaplex is now open to all at:
http://www.urbanomic.com/decaplex.html

Please RTFM before participating, and also see the supplementary points below which refer to the virtual implementation itself. A player with no intellectual appreciation of the innate beauty of the axiomatic is liable to find the game pointless, confusing, dull, brutal and short. Remember: the game is more important than the player, but the more players that participate, the more dynamic the gamespace becomes.

Extra-axiomatic rules of engagement:

See you there...

Posted by robin at 01:23 PM

February 11, 2005

What is Philosophy, v1.0

It's hard to imagine a more frustrating book than Deleuze and Guattari's What is Philosophy. Hard to imagine that it was on the bestseller list for weeks in France (presumably in the same way Hawking's Brief History of Time was here, but perhaps even less of those who bought it read it, and of those, how many benefitted tangibly from the experience?). It may be just me (maybe suffering the aftereffects of Badiou's lucidity) but this book seems not to make any sense whatsoever, it seems to be a random combinatorial game with no systematic basis. I challenge anyone to explain their model of the concept without recourse to D&G's own elliptical formulations.

Anyway, a further maddening attempt to read this book, along with some stray thoughts about information theory, got me thinking. I wonder if I could discover the Kolmogorov complexity of What Is Philosophy?.
K(wip)
would be the shortest possible universal-turing-machine program that could produce wip, the text of this book, as an output.

We would, of course, converge to this notional minimum length through the use of an actually-existing programming language P, which would give an upper limit on K(wip) where:
|Kwip| < |Pwip| < |wip|

If it were possible to define such a program, even if we were still to subject ourselves to reams of impenetrable french-fried guff, at least it would be stored more efficiently.

So, as a first step towards this, I have invested half an hour of my valuable time writing a program in Perl, less than 50 lines long (including the all-important randomexamplebracket() subroutine), and I present here the highlights of its first run.


It has been said that affects are always percepts, at least when they enter into milieus of absolute externality. This is why these perpetual variations are not always surfaces, at least when they constitute milieus of incorporeal immanence at infinite speeds on the affective plane.

We say that concepts are always volumes at infinite speeds in the affective chaos(a rat, the broomstick of spinoza, an old lady's nose). On the other hand these perpetual variations are not strictly mediations , at least when they participate in consistencies of speed (a bronze spear, the cogito of Descartes, the prickles of a hedgehog).

This is why thresholds are not, however, volumes, at least when they enter into milieus of absolute speed. On the other hand affects are strictly volumes, at least when they engage coordinates of absolute operativity in the perceptual function.

These thresholds are not , however, transcendent , except when they enter into zones of externality (the trunk of an elephant, the love of a good woman, an old piece of pizza). This is why concepts are not always volumes , at least when they enter into thresholds of absolute speed. On the other hand thresholds are not strictly surfaces. This is why these perpetual variations are always transcendent , at least when they enter into milieus of absolute transformation at infinite speeds (an axe, the faculties of Kant, the entire human species). Certain such figures are not strictly immediate in the affective consistency.

It has been said that affects are not always volumes , at least when they enter into milieus of maximum externality at infinite speeds in the affective field. On the other hand variations are not biunivocally transcendent , at least when they constitute consistencies of immanence in the perceptual field(a bronze spear, the cogito of Descartes, an old lady's nose). These are expressions, both subjective and real, which circumscribe the plane in a creative process of creativity. These affects are strictly percepts at infinite speeds. This is why such figures are not always surfaces, at least when they enter into consistencies of immanence in the conceptual plane of immanence.

These variations are not always surfaces in the conceptual vicinity( a bird, the cogito of Descartes, the empty set of Badiou). On the other hand variations are , however, transcendent , insofar as they enter into thresholds of externality at infinite speeds (a rat, the cogito of Descartes, a juicy plum). These are singularities, both possible and determined, which circumscribe the plane .

Certain affects are biunivocally transcendent at infinite speeds (a bird, the Ideas of Plato, the delirium of Schreber). They form events, both subjective and energetic, which affect states of affairs in a perpetual transformation of creation(a cats whisker, the Ideas of Plato, a dry-stone wall).

We can say that concepts are always mediations , at least insofar as they participate in transformation at infinite speeds in the affective vicinity(a forest, Proust's girls, the taste of cheese). On the other hand these perpetual variations are not strictly percepts , at least if they can be said to participate in immanence in the oscillatory field.

Meanwhile, certain planes are not always surfaces....


[ENDS]

I think you'll agree I'm onto something here. Although it seems almost certain that there are a few warwick graduates who have already mastered the process but are keeping their algorithms under wraps, I firmly believe the code should be open-source, to prevent the development of a monopoly.

Try it yourself (a prize every time!): WIPpingboy.pl (feel free to post highlights in the comments)

And now I shall go back to translating Badiou. Ithangkew.

Posted by robin at 11:09 AM

February 10, 2005

Refined Products

[x-post hyperstition]

All the products developed by our software are matched to an author, the difference being that ScriptGenerator©®TM treats the author in the same way as all the other elements which make up the contents it delivers; ScriptGenerator©®TM makes up the author from existing elements and orders its functions within the dictates of the marketplace-It is impossible to overlook the fact that the author is a key element of any commercial strategy-

Is the book jacket-s marketing epigram a description or a lyricised algorithm? -The love child of J G Ballard and Baudrillard-; an unknown author, with an intriguing glamorous past (-he worked as a corporate detective in the US-) that bolsters the apparent verisimilitude of his account of financial alchemy and clandestine traffic.

On the opposite flap the author claims to run a newsletter -Africa Energy Intelligence-, so he knows what he-s talking about when he describes Liberian mudslides or namedrops "coltan, a mineral used in mobile phone microchips and currently the subject of an international embargo as it is extracted from a war zone, the frontier between the Congo and Rwanda".

He knows the places he describes, where fathomless quantities are extracted and massive exchanges are transacted, Canary Wharf office blocks, Liberian mines, Swiss hotels, the queasy, jetlagged world of demoniacally-driven men in suits, supremely confident masters of the elements: Gold, cocoa, barley, milk, pork bellies, all in biblical quantities. He knows the cold glow, the strained, artificial radiance of striplights seeping into carpeted offices in the early hours, the mud-strewn, primitive huts of researchers in rainforests where multi-billion-dollar interests are invested. Individuals who bargain with governments for release of their cargoes. The opaque nobodies whose only job is to pass commodities through themselves in order to separate them from the real players. Certainly Ballardian enough, these shadowy pharoanic beings happy to stay in the shadows, never touching what they trade, shifting megatonnes as lightly as digits.

Outlandish, yet utterly convincing, given the author-s background. And timely, too. The Parisian literary scene needs a -new Houellebecq- and as if on demand, he appears. Phillippe Vasset is as convenient a fiction as his creation (ScriptGenerator©®TM Paris, Librairie Artème Fayard, 2003; English translation Serpent-s Tail, 2004)

'Everything has been said'. It's time that you, entrepreneurs, use this mantra of artistic circles to your advantage. If everything has been written, filmed, and acted, if the flow of stories has effectively come to an end, it means that narrative has finally become raw material, a commodity. Therefore, its treatment can be mechanised. This manual will demonstrate how this can be done'

ScriptGenerator©®TM is not merely a program for the assembly of pre-existing elements of -content- but a market where agents can trade these elements as commodities; not just all the fictions in the world, but documents created by individuals following a spectacular event -" photos, amateur videos, recordings, oral descriptions will all be added to the stock, valued, floated, evaluated for potential profitability and risk-managed for potential litigations, checked for congruence with the ZEITGEIST index, and recombined into product designed to perfectly balance the public-s expectations with the range of available materials, create a slightly differentiated product from known elements, and ensure the potential of its exploitation in every one of the media channels that ScriptGenerator©®TM -s target clients own.

Searching for the source of the ScriptGenerator©®TM manual, on a long hallucinatory journey around the demimonde of abstract trafficking, disparate locations linked only by hidden monetary chains, characters who know each other only as accounting acronyms, the protagonist never manages to be absolutely certain whether ScriptGenerator©®TM has been realised, and who is behind it. Uncovering russian dolls of shell companies, hotels with no guests where commodities are laundered, docks full of tonnes of containers which change hands thousands of times in transit, between owners who will never see them. All we are sure of is that beneath our feet and above our heads occult traffic is shaping the global state of affairs, steered by a ruling class the vision of which is both seductive and terrifying. Codes, fictions, instant recombinations, crowd alchemy.

ScriptGenerator©®TM is no more unlikely than this already-existing realm to which we are more-or-less blind; indeed, as the ScriptGenerator©®TM manual argues, all that is wanting is a formalisation, and the concomitant sloughing-off of the -creative- process:

ScriptGenerator©®TM is revolutionary in that it obliterates the -creative- process, or more specifically, it transforms the production of content into one long treatment of raw material-This new tool will finally strip the entertainment industry of its unique status; it is the only industry which still develops a product from nothing, ultimately it will be transformed into a genuine industry, one which transforms commodities.-

An impossibly ambivalent, definitively loopy 'creation'. How's it done? As they say in the software industry: RTFM.

Posted by robin at 11:04 AM

February 08, 2005

Our little secret, Lisa

Kotsko on Lisa Simpson (in the comments of IT's thread):

The example of Lisa Simpson tells little girls:
1. If you're smart, you'll always be alone.
2. If you try to break into a new group of "smart" friends, you'll still feel out of place, unless you've always been surrounded by smart people.
3. Your family is always going to be at least vaguely annoyed at, if not outright hostile toward, your deepest interests and beliefs.

Whilst I.T. (whose trojan-horse of a blog has crashed my computer twice today already) argues for her as, minimally, least-worst role model available:

1. She gets to be cynical and knowing at the age of 8, with some killer lines.
2. In all the futurustic bits, she gets to be doing extremely well, studying and having British men fall in love with her (imagine!) (while Bart will fall pray to the Simpson's 'male disease', which means running around with a saucepan on his head will be the apex of enjoyment). Lisa holds out the hope that intelligent working-class girls will carve out a place for themselves, whilst retaining the love of their family (she dumps the British guy when he won't wear the pig cufflinks Homer gives him).
3. Besides, if you're smart, you should be alone. How else will you get any work done? Anyway, when Lisa goes to the library, all the old folk say hello, so she's merely misunderstood by her peers, not folk in general.

Lisa would perhaps be unlikely to appeal to anyone who's not already destined for library loneliness, and it's quite true, no-one ever listens to her cleverdick advice. But K and IT are both right, really.

One has to yield to the truth of the representation The Simpsons offers us: In fact, I'm tempted to say that Kotsko's three points above seem hauntingly familiar (so it's not particularly female-specific). Of course such pouting identification with this tragic model would merely constitute a pernicious compounding of intellectualism as separate from the 'real world', an affirmation of the necessary destiny of the thinker as the ineffectual outcast, at best allowed a peculiar glamour (marrying English men!). It may be 'the truth', but in general, does 'representation', however witty, suffice for satire? We might allow ourselves a gratuitous Badiouism and say that 'telling the truth' amounts to mere reflective accounting, and that 'creating a truth' is what's required of great satirists such as Swift. The question is less what LS's character represents, than the effect of the massive diffusion of that representation and whether it achieves any status above that of a compounding simulacra.

I've always thought the out-of-control meme of The Simpsons being 'subversive' is misguided; it rather seems to be to be an inoculatory shock that confirms the world's shitty corrupt and hypocritical and there's not much you can do about it except (a)laugh ironically and have another beer (b)return to the bosom of the nuclear family for succour. The more I watch, the more I appreciate the wit and humour and characterisation, but it hasn't changed my dissent from this received opinion.

Isn't it, in fact, a parable of resignation to the state of mediocrity? One doesn't come away feeling outrage and horror and determined that the real world should never become like Springfield - of course, I'm not arguing that one should (i don't suppose it would be that funny), but rather against the received cult-studs interpretation of the show's "importance" in this respect.

And even Lisa is in a certain sense consigned and resigned to mediocrity since her destiny is prescribed - there is a certain 'place' for thinking, for reason; separate from and submissive to the inpregnable heart of family-love, for a start (how would Lisa react if she was being abused by Homer, I wonder? That's a serious question. But doubtless a step too far for this mild ironic 'transgression'). Thought that surpasses the bounds of common-sense laziness and the bounds of family values is an object of amusement at most, because we know it can never prevail.

Is 'cynical and knowing at the age of 8' really to be our ideal for the 21st century, is that the extent of what we can hope for from our children? And how will the media-studies courses cope with the explosion in student numbers?

The most that can be said is that there is a formal innovation at work. In bringing a complex character like LS's, bound up as it is with issues of social family and class constraint, personal development, and environment, into the eery stasis of the cartoon-world (where no-one grows older except in imaginary flashforwards) the problem is posed in a distinctively disturbing and insistent way. But I'm not sure that it changes anything. Lars von Trier as guest scriptwriter, perhaps....?

Posted by robin at 12:27 PM

Funds of faith

There is perhaps nothing that better characterises the at once comical and lamentable blind adherence to anthropic intuition than the continued existence of fund managers in contemporary financial institutions. These people exemplify, if anything does, the proper target of any "anti-capitalist" gesture, insofar as it seeks to rid the world of illegitimate power and control, of the blind faith of the moneyed in the wisdom of the more-moneyed, and of the continued existence of corpulent priests guarding the gateway between cash-money and capital-money, and claiming a share by simple virtue of their talent at bluffing. Rather than capitalism "itself" being a source of corruption, it is discredited throwbacks such as these human agents that represent a failure to 'live up to' the conceptual torsion that the contemporary reign of hypercurrency insists upon. They symbolise our clinging to reasons (but not reason) and to the humans that generate those reasons for our comfort, in an arena where these reasons are at best, equal to shots in the dark. Those individuals who we used to call priests.

Imagine if you had a choice of two mathematical models which would be used to steer your investment for the next ten years. One had consistently grown at a steady rate, the other had wavered erratically, according to no particular pattern, but in general had yielded less than the first. Now, imagine that in the case of the latter, you were not given access to the actual rules of the model, except for in cryptic phrases that could not be interpreted in any systematic way. OK then, it's obvious which we would choose. Now imagine that the second, erratic, losing model is bipedal, with eyes and a booming voice. Oh, of course, you'd choose him.

A British building society sponsored a contest (as you will be aware, this is only one of many examples) where a group of teenage students made as much money as a professional fund manager lost, over the course of a year (+4.8 against -4.4%). Consider the staggeringly blithe arrogance with which the 'professional' dismisses this: "The students have benefited from a strict stop-loss strategy," said [ ] as a nod to the winners". Hmmm. Yes, stopping loss, that outlandish strategy, well, it could work, I suppose... As for his own choice of shares, he believes they are all solid selections hit by general market sluggishness". So what is he paid for, we wonder (if not to anticipate such sluggishness or to pick shares that will not be affected by it)?

The fortune of the fund manager's portfolio is a classic model of human error, a catalogue of 'hot picks' that turned sour, and reasoned 'value' investments made on the basis of 'knowledge' about the companies, which knowledge was obviously incomplete and so essentially useless. Whereas the students intuitively moved to a day-trading strategy, clipping profits from short-term movements in shares (a strategy that forms the basis of the high-risk high-profit derivatives market where traders accept there is little to 'know' about the hyperabstract options they are buying and selling - a market that is therefore considered somewhat less 'respectable' than the primary market in shares). The sponsor of the competition says "I honestly could not call their stock selections a well-grounded portfolio..Their strategy is more about avoiding disaster than long-term growth."

Thus the belief that human judgment embodies a sort of 'respectability' that has anything to offer above and beyond numerical trigger-mechanisms cashes out (or not) in the assumption that different strategies work over different chronological terms, and that wisdom does not trouble itself with short-term gains. But the sad fact for the professionals is that long-term, simple averages of the market (tracker funds) consistently outperform managed funds anyway, so they have no advantage there either. (Incidentally a tracker would actually have outperformed all participants in the contest). There simply does not seem to be any middle ground between massive-scale generalised expansion and an intensively-tended opportunism based on senseless micro-movements. The obvious background to this observation is the Efficient Market Hypotheses: that is, if there is anything to be known, then everyone will know it, and there will be no advantage in knowing it; or its converse, that there simply is nothing to know. All evidence suggests that the truth is distributed on extremely obscure and treacherous islands of order somewhere in-between these two extreme positions.

You have two choices: simply ride the long wave blindly, or plunge into the fathomless density of wavelets-within-wavelets without being worried about "knowing" what you're doing.

Either way, if reason was to prevail it would assure us analytically that reasons could not (that is to say that arguments from common-sense, empirical induction, or the interpretation of mathematically-undefinable qualities, could only be a hindrance, not a help).

Unless we take a different approach, from the side of the human, and treat investment not as a rational activity but as a sort of personality test or lifestyle choice. This would be a strong version of the conclusion John Allen Paulos came to about traders and their clients, that their decisions were often based on wanting to confirm (or recover) their own image of themselves rather than on hard-headed reason. In which case, as with all belief-systems, you get the results you deserve based on the congruence of your faith with the reality of the world you live in. And the lazy will always rely on priests to make their choices for them.

Posted by robin at 10:36 AM

February 04, 2005

All joking aside...

...and we don't yet know the effects of our blithe assumption that people like kilroy are 'merely' a joke.

You want to reach out to a large group of people you suspect exists, but the only way to realise them as a coherent group would be by appealing to instincts whose existence the PR-machinery that cloaks politics will not acknowledge, cannot acknowledge. This is one of the the dubitable benefits of liberalism, of course: there are certain things you just cannot say.

So how do you go about it? You can learn a lot from T.Blair here. "Make a mistake". Say something you shouldn't say, the resulting "controversy" is bound to be mired in "opinion", "debate" and the rest of the media machinery. What happens to the person who "made the mistake" doesn't even matter, what matters is that the object has been achieved, the latent grouping has been galvanized, it now knows that it exists, and knows that certain people in power know that it exists. Now, leave a trail...launder the "mistake" with apologies, or by passing the baton publicly from the original agent to another party.

Everyone knows what's going on. A conspiracy that can't be proven, a plan that hangs on apparently accidental or misinterpreted words publicly spoken. A clandestine teledoxy, comprehensively deniable, using open public channels of communication to organise around 'unacceptable' causes...Just a word that slips out, an appearance with the wrong people, a series of little mistakes. They know, of course, that the problem with liberalism is that the worst it can do is to disapprove, and it does that very loudly, alerting everyone who needs to know. The liberal audience is all-too-happy to accept the apology, because they know it couldn't possibly be true, no-one really thinks like a fascist, that was all a long time ago. But the other audience, the silent minority (or majority?) knows different.

And these people have their eyes open, they don't miss the signs, in fact they've been waiting for them, so it's no joke.

"But what if they won't go? What then?"

Posted by robin at 01:24 PM

February 02, 2005

Automatic Ontology

What follows is a gloss on and interpretation of Rudi Cilibrasi and Paul Vitanyi's paper 'Automatic Meaning Discovery Using Google'. Prepared for my own notes, it might be useful as a relatively de-equationed digestible version of the research, and secondarily as a discussion of how it links to other of my own (and others') research themes.

The paper has its background in the researchers' previous work which used 'feature-free' compression-based methods to identify similarities amongst 'texts' in varied domains including genomes, computer programs and music. In the most accessible and widely-reported research result, a compression-based use of information theory was used to define a similarity measure for pieces of music expressed as MIDI files: Given a number of pieces by Beethoven, the method was able successfully to identify further pieces by the same composer.

The idea of 'feature-free' compression methods takes us right to the heart of important problems of information, definition, explanation, and meaning.

Compressibility can be thought of as a measure of information. (We treat a binary string as the generalised form into which all information can in principle be translated.) If you can compress a string of binary digits S of x bits into a smaller string S' of x' bits, and subsequently recover S from S', it is reasonable to assume that the information content of S is measured by x', not x - or, in other words, that the articulation of x in S is redundant. Compression equals the elimination of redundancy.

Discussion of compression techniques brings into play the question of the level of contextual assumption about the data to be compressed: a compressor such as mp3, developed specifically for use with digital audio data, is successful because its compression algorithm relies on certain empricially-derived assumptions about the type of data it will be asked to compress. In fact, mp3 works by throwing away information it 'knows' to be redundant to the ultimate musical 'message' (a message which, unusually, is not to be identified with the digital source code itself) - complex frequency groupings are simplified by taking out those which the user would not hear anyway.

No true (information-theoretically speaking) compression algorithm is afforded such an extravagant disposition. A general-use compressor must be able to compress and recover in full a binary string about which no assumptions as to usage can be made. They work, therefore, on a 'pure' definition of redundancy with indifference to contextual meaning. "Such compressors implicitly assume that the data has no global, structured meaning," looking only for "statistical biases" to exploit.

Kolmogorov Complexity;

Kolmogorov complexity represents the maximum compressibility of a string. K(x); stands for the shortest possible algorithm that will produce x; as an output (like Chaitin's program-length complexity). K-complexity can be understood best as a regulative ideal; we cannot actually discover or define the maximum compressibility of x;, but we can approximate it from above with the use of any number of programming languages or compression methods. So that for any computing system C;, we can say that:

K(x) &lt; C(x) &lt; x;

As should be clear, programming languages and compressors are theoretically identical. The difference is wholly on the side of practicality for the user, where programming languages need to present to the progammer some reassuring kinship with natural language in order to be usable. Actually, all programming languages consist of a syntax and lexicon together with a compiler that transforms the program into machine-code, constituting two distinct stages of compression, one facing towards natural language, the other towards the pure binary seriality of the processor opcodes.

So, K(x); is approximated from above by C(x); where C; is some compression mechanism, programming language, or computing system whereby x; can be transformed into x'; with the guarantee of complete recovery. "The choice of computing system C; changes the value of K(x); by at most an additive fixed constant" that corresponds to the weight of assumptions built into C;. Regardless of this, C(x); can always be considered a linear approximation of the unknown K(x);. The authors give the practical example of the familiar compression program gzip, and the lesser-known but more powerful bzip2 and PPMZ, where gzip(x); gives xg; where |xg| &lt; |x|; (meaning that the bit-length of xg; is less than that of x;); but the other two programs b; and p; give:

|xp| &lt; |xb| &lt; |xg| &lt; |x|;

By definition, |x|; and |Kx|; given by K(x); limit the series:

|Kx| &lt; .... &lt; |xp| &lt; |xb| &lt; |xg| &lt; .... &lt; |x|;

- "and our task in designing better and better compressors is to approach this lower bound as closely as possible." To amplify and generalise, then: our task in writing computer programs is generally to approach optimum efficiency or source-code to output ratio;and our task in any rational scientific endeavour is to account for as wide a field of phenomena as possible with the simplest rule possible.

Normalized Information Distance;

The authors next (leaving aside a great deal of technical detail) define the shortest program that will compute output y from input x as the Information Distance E(x,y);. Defined in terms of K; this is shown to be:

E(x,y) = K(x,y) - min{K(x),K(y)};

So the quickest way to get from x; to y; algorithmically is found by the difference between the shortest program for producing x; and y;, and the lesser of the lengths of the shortest program for producing either x; or y;.

Normalized to yields a value between 0 and 1, E; is proportional to K; and has the same limiting properties - all ways of getting from x; to y; using a given computational system C(x,y); will converge at the lower limit of E(x,y);. Normalized Information Distance; (NID) "discovers for every pair of strings the feature in which they are most similar, and expresses that similarity on a scale of 0 to 1" where 0 = identical and 1 = sharing no features.

The Normalized Compression Distance; (NCD) is straightforwardly an implementation of NID for a given C;. "Thus the NCD is actually a family of compression functions parameterized by the given data compressor C;0".

Now, the method used to compare MIDI files analysed (with no assumptions) all statistical features (all measures of NCD between two MIDI pieces), just as if the purpose were to find the most effective compression method, and used the most predominant of the statistical features as comparison operators. The likelihood of piece c; being by the same composer as piece y; was defined as NCD C(x,y); - this latter assumed to approximate NID E(x,y); which would name the optimal way of determining the 'Beethoven-ness' of a piece (and hence, arguably would encode the meaning of Beethoven-ness) Similarity, and thus relatedness, were deduced from Kolgomorov-complexity or compressibility.

This was a method that dealt directly with literal objects (pieces of music) in order to discover mathematically the distinctions that a musician would intuitively make between pieces of music. However, the proposition is to apply a similar technique to more abstract meanings "that cannot be given literally, but only by name and acquire their meaning from their contexts in background common knowledge in humankind." Two obvious intuitions stand out with regard to such semantic; analysis, roughly corresponding to the founding intuitions of structural linguistics: that context and usage is as or more important than the literal content of an object (a word), and that linguistic meaning is actually a function of the relationship between 'names' for objects, rather than between objects themselves.

So an analysis of the statistical features of the names of objects; will yield more significant results in this field than one based on words themselves as concrete informational entities, and this is what the Google method sets out to achieve.

Normalized Google Distance;

In effect the researchers use Google as a set of assumptions for the building a compressor. Their initial intuition is that, encoded in the probability distribution matrix constituted by the Google database, is a mechanism which can 'define', 'explain' or compress the semantic relatedness of millions of words (the 'low level' basic semantic knowledge embedded in the web) into a relatively short program-length (the Google database).

Thus "in essence, the choice of compressor brings a particular set of assumptions to bear upon an incoming data stream, and if these assumptions turn out to be accurate for the data in question, then compression is achieved. This is the same as saying that the probability mass function defined by the compressor concentrates high probability on these data. If a pair of files share information in a way that matches the assumptions of a particular compressor, then we obtain a low NCD."

Remembering what was said above concerning the shortcomings of generalised compressors such as gzip, it would be unlikely for them to attain optimal compression for a characteristically-featured (non-ergodic) domain such as natural language. Kolmogorov asks "what real meaning is there, for example, in [applying a generalised compression approach] to 'War and Peace'?" Generalised compressors cannot take advantage of any feature in the source material, no matter how obviously such features may seem, to us, to distinguish that material from random noise. We might conjecture that for a feature-rich source x;, |x'|; given by C(x); with a generalised C; is bound to converge to |x'| = |x|;, that is, no, or minimal, compression is achieved, compression being inversely proportional to the 'featuredness' or information in x;.

"Everywhere gzip looks it only finds a loaded die or biased coin, resolute in its objective and foolish consistency...Thus instead of starting with a standard data compression program, we start from a probability mass function that reflects knowledge" - namely, the function that maps search terms to number of pages returned in Google - giving Normalized Google Distance; or NGD. "We have replaced the obstinate objectivity of classical compressors with an anthropomorphic subjectivity derived from the efforts of millions of people worldwide." giving a highly adaptable, non-domain-restricted measure that uses names of objects to discover the structural relationships between those objects.

Given their simple example it is easy to see how superior Google is as a predictor for texts which are not indistinct and ergodic but contextual: given 'the quick brown', a purely probablistic compressor would give the most likely NCD outputs own, row, she, the;; where Google gives fox,vet,jot,hex;.

Of course, using a 10-billion page database as a compressor is bound to make things a little inconvenient! But, theoretically speaking, since the size of the computing system increases C(x); by a constant (C; is static as x; increases), at larger program-length sizes this will become less important, and we will still converge to K(x);. The trade-off between compressor-assumptions and compressive power changes as a function of |x|;...

The distinction the researchers make is between a literal-object based NCD and one based on contexts and names; of objects, where information about the context is built into C;. And the germane historical fact is that the Google system has unprecedented latent semantic properties;. Despite the fact that many projects have been set up to construct just such semantic webs, Google has produced a superior one 'for free' (and keep in mind that with this research we are only dealing with using the pagecounts in this research, not even the page contents!)

"Google events [the mapping between a search term and the number of pages it returns] capture in a particular sense all background knowledge about the search terms concerned available (to Google) on the web. Therefore it is natural to consider code words for those events as coding this background knowledge" This last sentence is suggesting the consideration of the "Google compressor function" as a unique mapping between all search-results S; for given terms and background-knowledges, or if you will, contextual meanings; of those terms M; so:

G(S) -> M;

or, once more, Google constitutes a computing system C; such that the function C(x); mapping a word to its meaning approaches the supposed Kolmogorov-complexity function K(x); giving the optimally-compressed definition of x;'s meaning. Google would then be the closest turing-compatible system we have to the optimal understanding or definition of a term such as (or better than) that which a human being builds up in the context of lived experience.

Experiments;

The experiments proceed to give some examples of how this latent semantic content can be extracted, using feature-free knowledge-extraction methods. This approach basically takes a set of terms, and on the basis of the NGD between all terms builds a 'best-guess' topological model of their relatedness. The elegant and impressive results are shown in Figure 2 in the paper reproduced here.

<img alt="tree.gif" src="http://blog.urbanomic.com/dread/archives/tree.gif" width="520" height="641" />

The authors refer to this as a 'best-matching unrooted ternary tree' (unrooted tree? rhizome, surely!) and refer to the result as an "example of automatic ontology creation". Here we have an interesting analog to the question of 'popular numerics', and also of its compromise with anthropology, insofar as one is tempted to question whether, in building in such a vast amount of information into the compressor, we do not simply minimise the influence of the source-code on the result, and merely reap the benefits of an overloaded computational language or mapping that has become more like a gigantic lookup dictionary than an operative algorithm? The equivalent situation arises with programming languages, where the compromise is measured in the ratio of programming time-to-runtime-cost: in 'high-level' languages where the compiler must translate from near-natural-language constructs, runtime and compiler size is increased, whereas with 'low-level' languages the programmer must invest greatly in coming to terms with the unnatural syntax of the inner workings of the processor.

However let us rehearse again the fact that since the complexity of C; increases C(x); only by a constant, the difference between a low and a high level C; will matter less in proportion to a larger |x|;. In computing terms, you only have to load the compiler into memory once, no matter the length of your program.

Next, the authors show a series of more elaborate tests where a training procedure is used to 'teach' a system by providing a set of terms divided into two categories as a basis for a web of pairings that can then be run through Google yielding NCDs (or rather, NGDs). Here where the system is given a 'headstart' (analogous, perhaps to the certain amount of learning-by-rote a child has to do before being able to spontaneously add to their knowledge) the results are even more impressive (the experiments section of the paper should provide no difficulties to the non-mathematical mind). Lastly, and rather cruelly, they compare their results to those from WordNet, a semantic model that has been built up deliberately over years with the contributions of experts, and find that the Google-derived results fall within the standard deviation from the WordNet results.

As they conclude, it is generally assumed that there is a "many-to-many relationship between...utterances and their many possible meanings," and what Google is spontaneously providing wil converge to a ready-to-use model of the entirety of this massive web of interconnections. Whereas a "context-free statistical compressor" such as gzip must assume the source code to be essentially random, its probability mass function corresponding to "the probability that the reference universal Turing machine outputs the string if its input program is generated by fair coin flips" - the universal distribution - assumes, then, a totally featureless 'web' of equal probabilities between all source-code-words (absolutely-decoded space), the NGD "is a comparable notion, in the context of the world-wide-web background information, to the universal distribution" i.e. it offers an actual; model of the quasi-universal distribution of probabilities rather than an idealised pure one. What is important to consider is that the historical transformation in computational circumstances have only now made it possible, through the collection of massive-enough sets of data, to analyse such a real distribution. The relation between the fixed constant |C|; and source-code |x|; indexes the impact of increased computational power on the applicative distinction between source-text and coding-machinery. This suggests the need to radicalise our everyday notions of the distinction between data and program in line with the theoretical model, and, along with it, the distinction between what intelligence is, what it does, and what it numbers or names.

Finally, to put it at its most speculative, what has been shown is that it would be possible for an alien with no foreknowledge or asumptions whatsoever, and no background in human culture or 'ontological assumptions', to analyse the data returned by Google as a response to random ASCII-strings, and eventually to discover the structural properties of any domain of natural language (the examples given in the paper are of automatic ontologies of whole numbers, colours, painters, and religious beliefs).

We could imagine such an alien, accidentally tuning some outlandish binary radio device into some stray emanations from earth, and being filled with wonder at the strange but palpably meaningful patterns returned by this binary oracle, intimations of an occult ordering from across the cosmos, an intelligence from outside...we also should share its extra-terrestrial wonder!

Posted by robin at 06:55 PM

Du sperme, j'en veux encore!

Lou Salomé, part IV, Berlin

salome.pdf - 360kb

I challenge you to read this without being overwhelmed by admiration for this outrageous, brilliant - and for many, catastrophically attractive - free-thinker...what a woman!

Posted by robin at 06:44 PM