Mnemonic infidelity

Training a large language model (LLM) starts with dividing a very large corpus of texts into basic units, i.e., recurring tokens (such as symbols and parts of words), and calculating the relative positions of tokens in the texts. These relationships are processed in a neural network to create very large numerical vectors that represent patterns…More

Absurdum explicandum

Following my previous past on different modes of explanation, I was interested in how an audience would decide that an explanation was inadequate, did not make sense or was full of errors. Avoiding the language of truth and falsity, I asked NotebookLM: “Please mirror these five categories of explanation with examples of explanations that most…More

Explain this …

In my previous post I explored how LLMs provide explanations of the “reasoning” by which they produce their outcomes (i.e. responses, answers, decisions). At least in their current iteration, platforms such as ChatGPT are tuned to generate an explanation independently of the process by which they came up with their response in the first place.…More

Mandala and metaphor

I’m on sabbatical as I write this, and taking the opportunity to research AI in the urban context as I travel. Here I am in Central Java, at Borobudur for the first time. This trip was prompted by my early work with Adrian Snodgrass, with whom I joint authored the book Interpretation in Architecture in…More

Urban cryptography

I’ve been investigating city life through the lens of cryptography. Here are some of the claims I think I can make on behalf of an “urban cryptography.” The most obvious contribution is that cryptography (digital encryption) addresses and resolves challenges of securing data and information flows in the city. But there’s more. Cities rely on…More

Camouflaged incitement

Everyone fights, or are prepared to say they do. As I noted in my last post, Fighting words, a major charge against Trump was that he exhorted his followers to “fight like hell.” His defence lawyers countered with a lengthy sound montage instancing peace-loving politicians (I think all Democrats) invoking the word “fight” in their…More

Fighting words

“Surveying the tense crowd before him, President Trump whipped it into a frenzy, exhorting followers to ‘fight like hell [or] you’re not going to have a country anymore.’ Then he aimed them straight at the Capitol, declaring: ‘You’ll never take back our country with weakness. You have to show strength, and you have to be…More

Write me a city

I’m interested in cryptography and the city, as a way of thinking about the supposed smart city. The graphic part of cryptographic might make us think of drawing, but it has most to do with writing. Etymonoline.com describes the common abstract noun ending -graphy as a “word-forming element meaning ‘process of writing or recording’ or…More

The word on the street

You may wonder at the affinity amongst rich, famous, powerful and (mostly) white men — and hip hop. We may have expected an affinity between Barack Obama and rappers such as Jay-Z and Kanye West, but not Donald Trump. To those in the know, rap gives expression to social conditions where there’s “unemployment, violent crime…More

A life of crime

Crime is immensely popular — in fiction if not in everyday life. No victim enjoys the consequences of crime. But many of us enjoy a good mystery, and the narrative aspects of crime meet some human need, as when solving a puzzle, or watching others solve it. Burglary materialises metaphor, putting a thing in the…More