The Book of Silicon | 2328

I think much of contemporary AI and Internet criticism fits into three categories, two of which are based on short stories by Luis Borges.

Full Show Notes:

Support the show! 
Subscribe to the zine
Watch on Youtube

Permanently moved is a personal podcast 301 seconds in length, written and recorded by @thejaymo

The Book of Silicon

I think much of contemporary AI and Internet criticism fits into three categories, two of which I based on short stories by Luis Borges – both are about the fear of endless media.

The Book of Sand

The first story is called The Book of Sand

A mysterious narrator recounts an encounter with a tall Scottish Bible-seller, who shows him an ancient book, originally purchased from an Untouchable in India. The book, marked with the title “Holy Writ” and the word “Bombay,” is also known as “The Book of Sand”. 

The narrator exchanges a portion of his pension and some other valuables for the book. He conceals it on a bookshelf behind his copy of One Thousand and One Nights.

The book, written in an unfamiliar language and interspersed with illustrations, appears to be limitless, with new pages seemingly sprouting from the front and back covers as one reads.

Throughout the summer, the storyteller becomes fixated on the book, meticulously studying and cataloguing its illustrations, while avoiding the outdoors to prevent its potential theft. 

Slowly it dawns on him that the book’s title symbolises its infinite nature, its pages like endless grains of sand.

Eventually, he comes to terms with the book’s monstrous nature and contemplates incinerating it. But he fears the endless smoke from it multiplying pages might suffocate the world. Instead, he decides to abandon the book in the basement of the National Library, believing that “the best place to hide a leaf is in a forest.”

Library of Babel

The other Borges story about the horror of infinite media is called Library of Babel. 

The world of the Library is depicted as a vast series of connected hexagonal rooms. Each room features an entrance, basic human necessities, and bookshelves occupying four walls. 

The books on the shelves are arranged randomly, without logic or reason. The contents of all the books in all the rooms encompass every possible combination of 25 fundamental characters. 22 letters, a period, a comma, and a space. As a result, most of the books are nonsensical.

The library contains every conceivable book. Including all existing and potential future works, along with variations and errors. The narrator’s logic follows that it must also contain valuable information, such as future predictions, and biographies of everyone who has and has yet to lived. However, they also imply that as any text can be interpreted in countless ways, all the infinite texts also have infinite meanings.

Despite or perhaps because of the abundance of information, all the books are meaningless. Plunging the librarians into a state of despair – often leading to suicidal thoughts. 

This overwhelming situation also gives rise to peculiar behaviours or classes of librarians. Some, known as “Purifiers,” destroy books they consider nonsensical. Others search for the “Crimson Hexagon” and its mythical illustrated books. Others still hold the belief that among the infinite books, there must be one that perfectly indexes the library’s contents. Some even believe in a messianic figure, the “Man of the Book,” who has supposedly read the perfect index, and wander the library in search of him.

The Infinite Library is a fitting metaphor for Machine Learning Models. These models have complex data structures filled with vast amounts of compressed information. Just as it’s challenging to find meaning in the Library, extracting useful information from these models can be tough. Unlike the library, LLMs are curated. First, weights constrain the model’s potential outputs. Second, specific prompts guide the retrieval of information. Both refine the vast possibilities of the model by altering the hyper dimensional structure of latent space.

It’s common to see LLMs dismissed as only predicting the next word. But what the next word is depends on where you are in the model. Make a move from one place to another in latent space and you’ll get a different set of next words.

The idea that we prompt an AI model and wander around its hyper library breaks down however when you realise that a LLM is both the library and the books.

If the Library of Babel reflects LLMs internal mechanics, then the Book of Sand represents concerns over their influence in the real world.

It’s the worry that we risk an internet full of endless nonsense and impossible images. Link farms and SEO-bait pages already already rife with gibberish, are now further summarised by AI algorithms that rank them at the top of search results. It’s as if we’re watching the Book of Sand burn in real-time, filled with the endless smoke of the burning Book of Sands pages. Perhaps in a few years it will be difficult to see the leaves for the forest.

What is happening?!

My third category diverges from AI.

Instead it addresses my own almost Lovecraftian horror of social media. Setting aside machine-generated content, social media’s algorithms have for the past decade prompted us to combine and recombine in mostly meaningless ways the outputs of creative lives, emotions, and thoughts. This relentless cycle of content creation mirrors apprehensions surrounding LLM output.

We aren’t librarians wandering Borges’ infinite hexagonal chambers but its scribes. The machine has been prompting us to answer the question: What is happening?! in its vertical rooms called feed for a decade.

Next time you read a piece on AI, ask yourself is this an Infinite Library piece, about the Book of Sand or a What is Happening story.


Prefer Email? πŸ“¨

Subscribe to receive my Weeknotes + Podcast releases directly to your inbox. Stay updated without overload!

Or subscribe to my physical zine mailing list from Β£5 a month

Leave a Comment πŸ’¬

Click to Expand

One response to “The Book of Silicon | 2328”

  1. […] The Book of Silicon […]

  2. […] notion that the internet will become flooded with machine-generated sludge is something I’ve written about before. It’s always felt like a distant storm on the horizon. But nothing prepared me for how soon […]

Leave a Reply

Your email address will not be published. Required fields are marked *