My fascination with electronic sound production began when my father brought home a Grundig TS 340 reel-to-reel stereo tape recorder. Like many other recorders it had separate heads for recording and playback, which meant that you could play a recording, add new sounds and feed the combination back through the recording head in real time.…More
Spatial affinities
In natural language models (NLMs), semantic embedding vectors capture the position of a token in multidimensional feature space. The space is derived from word proximities in a natural language corpus and is derived by the automated adjustments to weights within a neural network model. NLMs can deploy these vectors in a number of ways, including…More
Attending to the city
In the book Network Nature, I explored how people attune themselves to the natural world — or at least, how we attune to that part of our spatial experience that we are inclined to describe as “natural.” We also attune to artifice, such as urban environments. Cities present a spectrum of stimuli that shape our…More
Hiding one AI generated picture inside another
Steganography is the art of concealing one image inside another. I discussed the basics of the technique in a post: Hiding one surface inside another. It involves bit-shifting. In an 8-bit image each pixel is represented by an integer in RGB (red, green and blue). With 256 shades of RGB that provides over 16 million…More
Evaluating your AI
I’ve been implementing small scale trials of automated natural language processing routines that deploy the same methods as ChatGPT, i.e., implementations of the so-called Transformer architecture. That requires training on sequences of words in an original source document, equating each word to a semantic encoding (i.e., a long vector of 30 floating point numbers sourced…More
NLP sequences and cycles
In spite of its esoteric mathematical intricacies, automated natural language processing (NLP) as in conversational AI, draws on at least one primal everyday phenomenon. I’m referring to the concept of periodicity, i.e., cycles, periods, rhythms, repetitions, etc. Periodicity is a major principle through which we understand time, temporality, ordering, and sequencing and permeates so much…More
Robot probes city grid
Sentences are of course sequences of tokens (or words), and one of the major tasks of large language models (LLMs) is to predict the plausible continuation of sentences from some start condition, such as a prompt in a chatbot dialogue box, taking account of the previous flow of tokens within a given context window. Several…More
AI learns ABC
Neural networks (NNs) as deployed in automated language processing (NLP) are good at identifying and reconstructing patterns in data. So a NN trained on a corpus of texts can identify words that are commonly grouped according to their proximity within sentences, e.g., we wouldn’t be surprised to find words (tokens) such as “building,” “services,” “construction”…More
Chatting with multi-modal AI
ChatGPT now supports image upload, which means that I can chat about pictures, a step change in its application to architecture and design. Here’s my first such conversation, with chatGPT responses shown as quotes. The platform has information about my interests in urbanism and AI via the Custom Instructions settings. So responses reference my interests.…More
Non-NLP models
As an NLP neophyte I thought the mechanisms of automated natural language processing, such as ChatGPT, esoteric and arbitrary. At best, the general pre-training transformer (GPT) model is structured as a series of mathematical functions that apparently work in simulating human language, but bear little relation to the processes by which language is actually spoken,…More
By the same token
Automated natural language processing (NLP) of the kind deployed in conversational AI typically breaks blocks of text not just into words, but into tokens. Tokens are strings of characters extracted from a corpus of texts — the set of texts an NLP system is trained on. Tokens are more efficient than storing just whole words.…More
The neural neighbourhood
In a previous post (AI makes AI) I described an an auto-associative neural network with an example using made-up data. Here’s a more sophisticated version with actual urban data. Consider this fragment of a city map. I overlaid a 100 metre grid carving the neighbourhood into 72 cells. Within each of these cells lies a…More
My first AI language model
Here is output from my first home grown generative language model. I trained it on my own writing (19,000 words), it extracted my vocabulary and it was able to produce nonsensical but better than random text, considering the simplifications and assumptions in the language model I used. “Unveils ancient daydream charles regulatory demands interactive aggregate…More
AI makes AI
One of the benefits of (and concerns about) large language models (LLMs) such as ChatGPT and its successors is that it can write computer programs, even AI programs. The concern is that it might start to do so autonomously without human intervention, but that’s for another post. GPT writes code I took advice from ChatGPT…More
Find yourself in AI
Guest post In the vast expanse of the digital universe, AI, particularly large language models (LLMs) like ChatGPT, have found a home, influencing our lives in countless ways. A lingering question, however, is whether we are inadvertently influencing these models back. In other words, could our writings have been part of the training data for…More
What’s next
Industry, education and everyday users are making increasing use of large language models (LLMs) driven in part by the prominence of ChatGPT and other AI tools. The technology is developing at a pace. The analysis of commentators, critics and legislators also gain traction as they evaluate the implications of the technology and seek to influence…More
Meaningful, imaginative, fluid
The recent escalation of interest in text-based AI coincides with our students completing their project work, dissertations, research degrees, etc. So it’s been a useful period to test the capabilities of platforms such as ChatGPT4 as a tutor — or perhaps as co-tutor. As yet there are no constraints in place that seriously impede its…More
Training your AI
The training data for a GPT model such as ChatGPT4 consists of many (hundreds of billions!) tokens harvested in sequence from various online sources, such as Wikipedia amongst many others. This source data is processed as a continuous stream independent of document boundaries. Tokenize The initial training task is to tokenize the entire corpus, identifying…More
More on automated recollection
Conversational AI platforms such as ChatGPT generate predictions for what should come next after a user inputs a question, statement, paragraph, or other text prompt. In early text prediction software, a simplistic language model might calculate the most likely word to follow a given input like “door,” based on pre-calculated statistical analysis of word co-occurrences…More
The human in AI
As critics contemplate the artificial in artificial intelligence, it’s worth considering the extent to which human agency plays a role in this technology, particularly in LLMs (large language models) and conversational AI. These technologies are invented, developed and improved by human beings, and they are fuelled by gigabytes of human generated texts. That much is…More
Why a neural network forgets
Conversational AI, such as ChatGPT, has limited capacity to recall the content of earlier conversations. OpenAI does not disclose all the details of its operations, but scholars estimate that ChatGPT4 can process and recall up to 10,000 words in a single session or thread. That’s a substantial improvement on earlier models, but it doesn’t ensure…More
AI textsurfing
Before digital text markup I would highlight key words in a difficult document with a coloured highlighter pen. Students with a more methodological inclination developed this into an art, with colour coding and supplementary markings and marginal comments. (I see that Staedtler imply that their pen users are textsurfers.) Now my markup practice is fairly…More
Can I use AI in academic writing?
I’ve been coaching masters dissertation students as they complete their final projects. I’m interested in large language models (LLMs) and their applications. At the moment, it’s easy to include the ChatGPT platform as a participant in one-on-one discussions with students. The subject matter of their projects is digital media, so familiarity with the applications, strengths,…More
Attending to more than one thing at a time
In an earlier post (Attention Scores) I considered how automated natural language processing (NLP) models attempt to simulate the way a listener or reader will focus on key words and groups of words in a sentence to decide how to respond to the sentence. I won’t repeat the calculation here. But recall that the automated…More
Confidential documents and conversational AI
Confidentiality is key in any profession, especially as it related to client-consultant relationships. I’m hard pressed to find confidentiality foregrounded in architectural codes of practice, but it is crucial in law and financial services. The Handbook of the Financial Conduct Authority, for example, states that a financial advisor (a “skilled person”) “may not pass on…More
AI and the Cryptographic City
My book Cryptographic City: Decoding the Smart Metropolis (MITPress) came out in May 2023. Here’s a summary taken from the MIT Press website. Cryptography is not new to the city; in fact, it is essential to its functioning. For as long as cities have existed, communications have circulated, often in full sight, but with their…More
I took my AI on a holiday
Here’s what we talked about, unedited and unabridged. It’s very upbeat! Illustrations are from my holiday in Mauritius June 2023. Me: Like many idyllic islands in the tropics, Mauritius has a dark past. ChatGPT: Indeed, like many places around the world, Mauritius has a complex history that includes both positive and negative aspects. While the…More
Welcome to the Apocalypse
Not everyone is averse to the prospect of a global AI-induced apocalypse. Catastrophizing circumstances and events caries a certain appeal to some, in particular those who identify with the status of a powerless underclass. Let social, political and economic systems fall! Let AI take over! I think here of those who identify as dispossessed, who…More
Generalised AI as existential threat
Many specialised AI applications are acceptably proficient in identifying people, animals and other objects in pictures, in searching databases, winning at chess and in many other areas invisible to most users, such as controlling factory production lines, navigating aerial vehicles, surveillance, medical imaging, diagnosis, and aspects of smart city infrastructures. That’s “narrow” AI. But general artificial…More
Attention scores
One of the important techniques in the new wave of highly successful generative natural language (NLP) models is the use of attention scores in neural network (NN) training. Here I continue the investigation I started in previous posts into how NLP works. To recap: a word in NLP models is represented as a point in…More
AI Armagedon
Should AI research be shut down? A group called the Future of Life Institute warns against the recent developments in generative AI platforms, prompted especially by advances in natural language processing, e.g. chatGPT. The open letter follows their similar warning of 2015. Both letters have several high-profile signatories who are involved in AI development. The…More
Words in order
Neural network researchers invented several methods that store and make inferences about the order of words in a sentence. The main method I will present here provides one of the components that undergirds the recent impressive performance of natural language processing (NLP) models known as transformer models. The method also resonates with my prior investigations…More
AI eschatophobia
It seems that everyone is talking about (or with) ChatGPT. The platform’s convincing conversational acuity and ability to synthesise disparate conceptual threads provides a vivid demonstration of AI’s potential. I’ve now read several accounts online where scholars, programmers, writers, musicians and artists use ChatGPT or similar as a creative companion to explore ideas — comparable…More
Fine tune your AI
My first blog posting appeared 25 August 2010 titled “Tuning as …” It followed the publication of my book The Tuning of Place. In the short posting I proposed: “Tuning might well be the metaphor of our age. ” I was thinking of smartphones and their role in moderating our relationships with environments. The term “tuning,”…More
Architecture in multidimensional feature space
In a previous post (Predicting proximity), I reviewed the NLP (natural language processing) operation of calculating the relationships between words in a corpus of texts. So the word “architectural” is closer to the word “urban” than “architectural” is to “culinary.” Very close word proximity could mean that one word can be substituted for another in…More
The invention of language
Automated natural language processing (NLP) technology is impressive, though I need to remind myself of its limits. It is easy to elide thoughts about its linguistic capabilities with a sense that it is on the way to mastery of language, and therefore human intelligence. In what follows I will follow the line that NLP is…More
Imperfect patterns
Training a neural network (NN) involves automatically adjusting numerical weights and thresholds (biases) to account for all input and output pairs presented to the NN. After training on these input-output patterns the NN should reproduce the appropriate output pattern when presented with any one of the input patterns. Note that it is not patterns that…More
Just one neuron
I revisited our earlier (1990) article on neural networks “Spatial applications of neural networks in computer-aided design.” Neural networks were novel in architecture and CAD. What follows is an update of the part of that article in starting to explain how neural networks function. Neural network layers Neural network (NN) models store information as numbers…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
Automatic pattern completion
In 1990 Arthur Postmus and I published an article about “spatial applications” of artificial neural networks (ANNs). In a more recent article, Gabriele Mirra and Alberto Pugnale at the University of Melbourne developed up-to-date applications of AI in design. They generously cited our article (amongst articles by others). I concur with their assessment of the…More
Who’s listening?
In every mind there is a higher function watching or listening to the inflow of sense data and monitoring its own thought processes. In his seminal book on the philosophical challenge of consciousness, Daniel Dennett argues convincingly against this proposition. He directs his criticism against those who presuppose that somewhere, conveniently hidden in the obscure…More
The next word
“We are so in sync we finish each other’s … sandwiches.” That’s a variation on a joke from several sources: The Simpsons, Arrested Development, The Good Place, etc. In some social situations where I can’t think of anything appropriate to say, I just start a sentence anyway, not knowing what comes next or how it…More
Predicting proximity
Tokens NLP (natural language processing) systems don’t necessarily consider only words as the basic components of prediction. They may break words into smaller units roughly corresponding to syllables, punctuation marks and even single characters. The general term for these textual units is “tokens.” An NLP system will include in its algorithmic workflow the construction of…More
Nothing beyond text
In explaining the political philosopher Jean-Jacques Rousseau’s (1712-1778) autobiographical book Confessions, the continental philosopher Jacques Derrida observed that there is nothing to support Rousseau’s claims about his relationships with his family members, other than what the text discloses. Derrida generalises this observation to the controversial proposition that any attempts to ground spoken or written assertions…More
Attention is everything
Attention is a key element in cognition. At our most thoughtful we direct attention to features in our environment that are most important to us at that moment. Attention can wander, of course, we daydream, and we can pay attention to inexistent things, memories and objects of the imagination. A lecturer will come to the…More
Learning to transduce
Human beings at their most rational are able to generalise from examples. If you stand under the shower head before turning on the tap then it is likely you will be dowsed with cold water before it gets to a temperature you are happy with. That if-then rule is a generalisation borne of a few…More
My tutor is an AI
In constructing this post I’ve been cross checking some of my explanations about neural networks with OpenAI’s chatGPT-3, which is a highly responsive chatbot available at the OpenAI.com website (free at present). I’m most familiar with spreadsheets. So that’s where I’ll start. You can think of a neural network as a matrix or grid (a…More
Intentional systems
In his book Consciousness Explained, the philosopher and cognitive scientist Daniel Dennett says that a study of hallucination, “will lead us to the beginnings of a theory — an empirical, scientifically respectable theory — of human consciousness” [4]. I’ve explored what some philosophers say about hallucination in previous posts and tried to relate that to…More
Chatting with an AI about urban inexistence
“Intentional inexistence” is a philosophical term adopted by the nineteenth century philosopher Franz Brentano (1838-1917) to indicate the commonplace human capacity to imagine things. We may suppose that inexistence refers to things that don’t exist (nonexistent things) or that are under demolition, but that presumes too much — or too little. As explained by Linda…More
Cryptographic City
The description of my next book is out with the cover design on the MIT Press website. The graphic designer says that the cover is in code, though I have yet to decipher it. Cryptography’s essential role in the functioning of the city, viewed against the backdrop of modern digital life. Cryptography is not new…More
Cryptographic index
To introduce cryptography as a theme in my Media and Culture class I provided each student with a customised string of code containing the name of an AI-themed film. The key to decode the secret message was the student’s unique 7 digit university identification number. If the encrypted string is uaroulojz and the student key…More