I've just completed an essay for a special edition of Architectural Theory Review in honour of my friend and former colleague Arian Snodgrass, who passed away last year. Adrian introduced me to structuralism, semiotics and hermeneutics. In part of that essay I reference the writing of the literary theorist and cultural semiologist Roland Barthes (1915-1980). I'll present some of that material here in so far as it relates to the ethos of large language models.

Here is an audio of a sequence of blog posts I published in 2013. Some were written while on a trip back to Australia after an absence of 16 years. These reflections of a traveller later informed my books Mood and Mobility, Network Nature, and Derrida for Architects. My current challenge is to see what I can update in light of subsequent developments in computing, digital media and AI.

I've just emerged from an interaction with ChatGPT reviving the following posts from 2013. 148 As the mood takes you 149 Accentuate the negative 150 Howling at the moon 151 Freeways in the sky 152 Mastering the universe I was interested in my early attempt to explain what we now think of as "confirmation bias." The example I led with was of people extolling the benefits of nature settings to instil a positive mood.

Past writings, diaries, letters, course notes, publications and blog posts reveal what I was thinking 10 to 20 years ago and beyond—or do they? To read them now is to be reminded how much any authored text is steeped in the artifices of language, culture, and circumstance. There is no direct access to an original thought. Texts don't deliver an unmediated window into an author’s mind, even for the author looking back.

I'm still reviewing my blog posts from 2013. In keeping with the unstructured nature of blogging, I didn't plan sequences of posts to follow particular themes. However, a thread does emerge from this sequence of six posts -- that of attention. Soft fascination (138) introduces the theory that sustained concentration on a task induces "attention fatigue," a biologically instilled tendency in humans and other animals to moderate their innate ability to focus in order to get things done with a countervailing tendency to withdraw from the task.

I'm reprising some older blog posts from 2013 that consider vertigo, oblivion, melancholy, the motion of swings, and the emotional experience of urban spaces. Later on I drew this material together in my book Mood and Mobility (2016). 132. Swinging 133. Oblivion 134. The brain in the city 135. Well-being and geometry 136. Your inner child 137. The melancholy medium…

I'm a member of the Author's Licensing and Collecting Society (ALCS). They collect money for "secondary uses" of publications – such as photocopies, digital reproduction and educational recordings. These rights bring in only small amounts of money, so authors don't usually take the effort to collect them. Nor do they know how to do this. The ALCS gathers such funds collectively and distributes them to member authors.

I am reviewing my early blog posts Richard on the Holodeck and Shallow reading from February 2013. In the first post I noted that some people see literature (e.g. a novel by Charlotte Bronte) as a substitute for living the lives of the characters. In Hamlet on the Holodeck Janet Murray seems to suggest that reading a text is inferior to experiencing the events of the novel -- living that life.

I'm revisiting older blog posts. I'm up to the one titled "Loose ends," which reflects on the nature of origin stories in the age of the Internet. The post from 2013, mentioned horsemeat detected in hamburgers, Derrida on the desire for beginnings, and Freud on red threads in naval ropes. I have recently traced a different personal origin challenge: the story of my father's wartime service.

I'm just back from a short break in the Faroe Islands. One evening I was standing in the reception area of our hotel when I heard the faint sound overhead of robust voices in unison accompanied by the slow rhythm of what sounded like feet stamping on the floor. The next day we were at the National Museum of the Faroe Islands in Torshavn where I watched a vintage film of people of various ages holding hands in a continuous line, or perhaps a loop, and stepping sideways, two steps to the right and one step to the left.

I'm continuing the theme from the last post of looking back and looking forward. To that end I am reviewing the next set of posts authored in 2012. I start with the romance with digital technologies as they were then. When I wrote Vitruvius does steampunk in 2012, I was interested in how steampunk revelled in retro-futurism: imagining Victorian contraptions that never were, delighting in conspicuous mechanics, and offering a critique of modern technologies.

I'm looking back at old blog posts and publications. In 2012 I was also looking back to older studies, e.g. to 1994. See the 2012 post: “What’s a modem?” I'm indebted to ChatGPT for suggesting that the 2012 post served as a "time capsule." In that post I revisited our 1994 study of computer-mediated communication in architectural practice. I was struck by how quickly the terminology of “CMC” (computer-mediated communication), “telephone modems,” and “text handling” had become quaint.

As I continue to trawl through earlier blog posts I see that one of my 2012 posts followed a one week sojourn in Iceland. Iceland's traditional narratives and mythologies draw on the activities of a pantheon of heroes, one of whom is the trickster god Loki. In that post, and subsequent publications, I referenced Lewis Hyde, Carl Jung, Sigmund Freud and Jaques Lacan who remind us that the 

I'm in a cottage in the countryside. There's an attic, a shed, a greenhouse, old gardening tools and furniture. Such legacy paraphernalia and economic necessity invite strategies of repair and re-use -- practices that spill into the intellectual sphere. I'm re-using earlier blog posts. These posts were published between March and May 2012, beginning with reflections on architectural authenticity, drawing from Louis Kahn, Heidegger, and Angelus Silesius to question whether buildings might exhibit the kind of agency that translates into “desires” (what buildings want) and whether our professional and intellectual tendency towards “thick thinking” can allow for simply…

I have been blogging about text-to-speech and voice cloning apps. I've also been turning compilations of my posts into audio files using Speechify suitable for podcasting. (See previous posts.) It seems the TTS (text-to-speech) tools I am using do not accommodate TTS HTML or other reliable means of introducing intonations and pauses into the readings. I'm particularly exercised by the lack of a decent pause between my posts in the audio of such compilations.

I am reviewing my early blog posts on technology, media and culture. The compilation here includes weekly posts between 9 April and 31 December 2011. The posts make reference to "actual events" that year. These include: Apple’s iPad 2 release; the impact of social media in the January "Arab Spring"; Osama Bin Laden’s death; the death of Kim-Jong Il; Barack Obama and the raid on Abbottabad; the proliferation of Social Media (esp.

This audio file (generated by Speechify's synthetic voice) contains the first 41 regular weekly blog posts I published starting in 2010. I began the blog after the publication of my book, The Tuning of Place. I have just called on ChatGPT to set the context for the original content. My prompt: "This content was first published in 2010. Position these blog posts in historical context.

Silent personal reading emerged in the 18th and 19th centuries, coinciding with the wider circulation of printed books and pamphlets, according to literacy scholars. Before this period, reading was typically performed aloud, even in solitude. Text-to-speech (TTS) software seems to be returning us to that read-out-loud practice. Our texts can be read to us by smartphones and other devices, at variable speeds and with different voices and intonations -- and while we are multitasking or on the move.

The term "clonecasting" often refers to copying an actor's persona, presentation style, appearance, and voice to create visual and audio media content. Audiences might think they are seeing and hearing a particular actor in a film, but the actor's presence is fabricated from digital models. Famous cases involve the reconstruction of the deceased actor Peter Cushing in Rogue One: A Star Wars Story…

My first blog post appeared in 2010 followed the publication of my book The Tuning of Place. I titled the post "Tuning as ..." The Christmas eve that followed I produce a post "Silent Night." Here is an excerpt from Silent Night in audio format. It runs for 2 minutes: That's not me speaking, but a clone.

I’ve been busy pruning an overgrown laurel hedge. Apparently, the previous owner cultivated the entire leafy barrier from a single twig he had surreptitiously snipped from a hedge in a garden centre. His careful propagation from that twig—growing, dividing, and replanting—constitutes cloning. A quick AI-assisted search reveals that the verb “to clone” derives from the ancient Greek klōn, meaning “twig.”

A recent headline in the Higher Education section of The Guardian said “Thousands of UK university students caught cheating using AI.” I could see that coming! It is as if some headlines (taglines, tweets and chyrons) are ready and waiting for a report, evidence base, study or authorised opinion to make them real. To help test the ease of "predictive headlining" I have tried to generate a menu of such headlines ready and on call for the moment.

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 of associations between tokens. These vectors function as…

You know that a technology has gone mainstream when it starts to affect the consumer side of the retail trade. I first heard the term "AI shopping" on the BBC Radio consumer programme You and Yours: "Now, a growing body of evidence shows around a third of us are using AI to find products, plan holidays and get the best deals.

Large language models (LLMs) are called "large" as they are trained on very large volumes of text data, about 570 Gb for ChatGPT 4.0's base model (prior to fine tuning). It sounds like a lot, but my iPhone 16 Pro has a storage capacity of about 500 Gb. Other smartphone models have twice that. So a smartphone could easily store the training text-data for a base level LLM.