Category Archives: ai

Hush! Caution! Echoland!

It’s a cliché to speak of Ulysses as an endpoint, the culmination of the nineteenth century novel, finally bled dry of plot and incident and to a certain extent character (we only feel that we know Stephen and Bloom so well because we spend so much time with them) and erupting into a monstrous growth of period detail and stylistic parody. It’s not the last station on the line, but at least, unlike with Finnegans Wake, one can still pretend it’s something like a readable work of fiction.

But Ulysses was my gateway into mainstream literature: before that, excepting what I was forced to read for educational purposes, I’d only read science fiction and fantasy. Literature was too boring, just a bunch of normal people doing grown up stuff. Ulysses was different: the first handful of chapters were pretty slow and contained a great deal of matter relating to Thomas Aquinas which I let slide by in peaceful incomprehension. But once the newspaper headlines started in the seventh chapter it started to get fun, if not easier to understand.

So for me, it’s always felt like a starting point, not a conclusion. It’s not exactly a friendly introduction to the Western canon, but there’s a lot of writers I first heard of, or was exposed to parodies of, under its influence. And its attitude of “hey, keep up with this if you can”, the sense you get of being complicit with someone taking everything they knew about every book they’d ever read for a dance, is exhilarating.

This post started out as another very short science fiction story, which is what I usually post here on Bloomsday, but it felt like it was getting into territory I’ve covered too often: a sort of dystopian scenario where after the Singularity, or some parody thereof, the AIs really do reconstruct Dublin from Ulysses, and put a bunch of human consciousnesses in it, and it’s terrible, like being trapped in a Bloomsday costume party for all eternity. I gave it up because, for one thing, I was unconsciously plagiarising part of a short story by Ian Watson from the 80s called, I think, “The Bloomsday Revolutions”. (I thought of Ian Watson for the first time in years the other day. He’s a good writer, look him up if you get the chance.)

The other reason I stopped was that I’m weary of science fiction being about computers and AI. I think that the real event underlying the Singularity is the collapse of the sf imagination into the computational. Too many of my own attempts to write longer pieces of fiction have gotten stuck or faltered for the same reasons.

I was remembering Jorn Barger, the guy who coined the word ‘blog’ and had a kind of internet celebrity which then dissolved into anti-Semitism and silence. Barger was an autodidact Joyce fan and had a site called “IQ Infinity”, the central thesis of which was that in Ulysses and Finnegans Wake, Joyce had solved AI, that through sheer brainpower he’d comprehended how the human mind worked. I don’t remember, if I ever really understood, exactly in what sense Barger thought that the works themselves constituted artificial intelligence: could one create the personality of “Leopold Bloom” from the text if it were somehow transformed into software? As another Joyce fanboy I can understand Barger’s reverent awe — without sharing it to that extent. And looked at dispassionately, it’s a ridiculous and self-infatuated idea: if only everyone else loved my favourite author as much as I do, they’d understand how consciousness works, too.

I was also thinking of Ted Nelson’s Xanadu project, the origin of the term ‘hypertext’: I’d known about it for decades, and based on what I’d read of his writing at various times, I thought of him as a crank, embittered by the success of the web. Xanadu proposes a much more complex way of linking and embedding documents within one another, with links that go both ways, and an elaborate system of building a top-level document from a variety of sources. Having come across this summary by a recent participant made me sense, in an obscure way, the allure of this vision of a global network of interpenetrating words. But in another way, it feels nightmarish.

In my mind, Xanadu’s “transclusion” is a codified and rigid version of the sort of association of ideas which the reading mind does in a flexible way all by itself. All writing depends on this, but it’s essential to a text like Ulysses, and even more so Finnegans Wake. Rather than narrating, “Stephen thought about Aquinas’ doctrines of sense perception as he walked along Sandymount”, Joyce interpolates “the ineluctable modality of the visible” and so on, all those weird terms I didn’t understand the first time, leaving it to the reader to either follow the echoes, if they are aware of the reference, or, if not, to fold the unusual texts into their own memory, to be echoed later or in other texts.

I love this process: to some extent, what I’ve just described is what being literate means to me. But I enjoy doing it with my own mind, or letting my own mind do it for me, and the thought of it being made explicit, with coloured markers joining the texts in different columns, makes me queasy, as do the very few working xanalogical demos.

I should add that sometimes just reading Joyce gives me the same feeling of vertigo. There’s a central image, or nightmare, behind these different incarnations, a cousin of Borges’ total library, the idea of mind as a sort of infinite glossary. It makes sense that my imagination, in trying to come up with a response to Ulysses as Bloomsday comes around each year, would return to the machines with which I work, and the fantasy that one day they’ll be able to read our favourite books so well that they’ll bring them to life.

I often get the same feeling reading blogs from the rationalist and AI risk communities. I suspect that these are not so much a school of philosophy as a literary genre, in which people with a very particular form of intelligence — discursive, articulate, fond of numerical arguments, insistent that any discipline can either be reduced to economics or physics, or is empty or misleading — imagine, with the same kind of self-infatuation, that magnified forms of this form of intelligence will either save or wreck the world. Earlier this year, I got so compulsive about reading this kind of thing that I had to use a site-blocking extension to stop myself.

I console myself with the idea that Joyce, had he lived in our era, would have been very bad (one imagines with glee his towering contempt and exasperation) at using computers.


Neural Streams of Consciousness

Style-extraction algorithms having reached the level of popular smartphone apps which could take the small-scale features of Hokusai’s wave or a Lichtenstein cartoon and apply them to a picture of one’s pet, it was only a matter of time before the technique was successfully applied to textual, rather than graphic, works. These first neural networks were mere mimics, more sophisticated versions of elementary Markov chains, which could produce plausible but nonsensical imitations of existing texts with no semantic content.

A breakthrough came with the Antal functor, which used a form of iterated adversarial machine learning algorithm to extrapolate multiple versions of a given text along many dimensions and then aggressively prune this ramifying cluster of words into “fixed points”, an unfortunate piece of mathematical jargon for what could be quite subtle and profound features of the source material. (The story of the functor’s use in extracting “virtual characters” from apparently objective and non-fictional texts, and the subsequent effects of this discovery on journalism and politics, have been told elsewhere.) Once this basic technique had been mastered, it could then be applied in an analogous way to that used in vision, sorting the qualities of a text on an approximate scale which ranged from such minutiae as idiosyncrasies of spelling or word frequency, to the characteristic syntactic patterns employed, and then on, with decreasing accuracy, into such large-scale qualities as extended metaphors, symbolic structures and plot.

The most famous application of this technique was the urDay service. The user registered his or her various social media accounts with urDay and allowed it to apply a battery of neural functors to the texts and images which flowed from them, taking these as a modern and technologically-mediated version of the stream of consciousness which had been pioneered in literature by Woolf and Joyce. The abstract versions of these could then be expressed in any number of ways: to generate wry or amusing animations with a cast of adorable algorithmically-generated mascots; inserted into an ever-changing roster of movie clips and viral videos as sarcastic commentary, witty cameo or heartfelt dialogues.

A set of textual plugins had been provided, more out of the curiosity of some of the development team and a sense of pride in their antecedence than any hope that urDay would have any serious impact on literary studies, much less kindle in its users a love of high modernism. With these, one could project the narration of one’s life in a kaleidoscope of styles and voices, just as Joyce had done in Ulysses: a cursory description of an annoying planning workshop or visit to a supermarket could be recounted in the language of high fantasy or science fiction. Use and abuse of these textual plugins became popular in certain literary circles, all the more because of the occasional thunderings against this digital prostitution of the art and craft of writing which came from the stodgier journals: although their output was, if anything, too facile and polished to really be groundbreaking as generated textual art, their use signalled that one was not above a certain populist bravado.

As is only natural, after a few years such collaborations seemed painfully dated, and the professional writers abandoned the field to those amateurs who enjoyed running an autoblog which gathered up and retold the output of their various encounters and days in the manner of, for example, a noir detective story, or an epic battle across frozen tundra, or a stylish psychodrama.

Their remained the matter of what became known as “the puzzles”. Certain scholars who had shifted from collaborating with the urDay plugins to analysing their outputs claimed that motifs and images seemed to be following patterns which, though elusive, could neither be attributed to the social media inputs, nor to the literary models used to generate the various styles. (The use of functorial analysis allowed this to be done with a degree of confidence.) For example, a week-long sequence from a university student’s autoblog, which alternated between a somewhat archaic translation of Sei Shonagon’s Pillow Book and Patti Smith’s memoirs, showed a striking affinity with certain of Pound’s Pisan Cantos, a work which neither the student nor her chosen electronic amanuenses had any connection. An archaeologist’s field notes, transformed into an elaborate science-fantasy scenario, spontaneously revealed a correspondence between certain ruins on the shores of the Persian Gulf and the galactic coordinates of active pulsars. Once one began looking for such patterns, it was said, they began to emerge everywhere, and perhaps it was this sense of ubiquity which explained the somewhat tepid response with which these demonstrations were greeted. While happy enough, at least in some circles, to let the false leads and teasing traps laid down by a legitimate genius like Joyce keep them busy for centuries to come, literary scholars saw the apparently limitless sea of neural “puzzles” as nothing more than an epiphenomenon of their computational origin, as uninteresting to them as the technical details of the programming languages used to create their word processors or functorial analysers.

Eventually, the “puzzles” became the hunting ground of that even more prolific realm of amateurism, the conspiracy theorists, to be added to their never-ending roster of patterns and coincidences, world without end.

Deepdreams Dive Two: the MIT Places Neural Net

Here are some more sample images run on the same randomised coloured fog as the last dive. These are done with a different neural net, based on the MIT Places database. They lack the biomorphic horrors of the default net, but they have a kind of weird beauty.


3a. Low-level features, and a tendency to diffract colour


3b. These look curiously suggestive of the spiral patterns on ancient Celtic artefacts.


4a. You can really see how the Places net gets cues from colour here: as you might expect, the green parts seem to want to turn into parks.


4b. This layer is just crazy for chairs. And windfarms.


4c. Pagodas as far as the eye can see.


4d. The wind farms are back, and the pagodas are starting to look like stupas.


4e. It’s strange how at the higher levels, this model starts to go wonky. There are anphitheatres forming in the lower left.


5a. Oddly specific details here: a watertower and a fountain.


5b. This is the only layer in this net that I find very unsettling. There are suggestions of distorted faces, and something like a pine-forest in the middle.