Referring to Architecture X Architecture: A Dialectic (AXA from hereon in) as a weighty tome is not just metaphorical: the 272-page book clocks in at 1.12kg – 250g more if you include the equally chunky slipcase. It also aims to be heavy in the other sense: with an uber-cool design; cornucopia of 450 full colour and monochrome images; seven theoretically dense texts (I glance down inadvertently from the screen to the page at this point and the words “photography’s so-called indexical relationship to reality” catch my eye); and standout quotes from big thinkers that punctuate the flow. Apart from the deep orange of the endpapers and a handful of highlighted pages, the white type of the text sits on matt black coated stock. The whole artefact has been meticulously printed in China and beautifully produced by Hong Kong based Oscar Riera Ojeda Publishers.




If, like me, you flick through the picture-pages before engaging with any text, the cascades of images appear to make little sense. You’re supposed to read the essays first, the reason why a very clever person chose to put them at the front. There’s a Foreword, followed by overview essays, then two extensive speculative projects resulting from the imbrication of architecture and AI, each with its own short intro and a plethora of images. The ensuing Practice section features recent designs from the producers of AXA, Californian firm Metropolitan Architectural Practice, otherwise known as MAP – all designed with the help of AI, but more real-world focused than the preceding experiments. Then, there is a coda essay by critic, curator, and educator Aaron Betsky, the title of which – From the Manifest Destiny of Dissolution to the Ether of Unknowability – is worth quoting in full. It’s probably snarky and superficial to say this language gives a flavour of the narrative parts of the book, though there’s also a little truth in it.

The authors, quite rightly, take their intellectual provenance seriously, although their introductory acknowledgements confusingly credit a “network” in which, somehow, Octavia E. Butler and Mark Fisher rub shoulders with Trevor ‘Plagen’ (sic) and McKenzie Wark. Aware that it’s only page ten and I’m already carping, I pull myself together and read the Foreword, by academic and filmmaker Kum-Kum Bhavnani, who warns about “the profound and sometimes disconcerting racial implications” of the intersection between architecture and so-called artificial intelligence. Bhavnani evokes a number of ethical considerations in her brief introduction, including the ecological impact of data centres, Silicon Valley’s overenthusiastic appropriation of the concept of manifest destiny, and the regenerative possibilities inherent in ephemeral urbanism (though one might question the use of Burning Man as an exemplar of the latter phenomenon – the Nevada desert festival having succumbed to ultra-gentrification, with wealthy attendees landing in private jets along with a small army of flunkies and personal chefs).

Why not instead evoke the origin story of the Situationists, in which a temporary encampment of Piedmontese gypsies in Alba, Italy, inspired the definitive mobile city, Constant’s New Babylon? Constant just worked with his own human intelligence and a bricolage of scrap materials, still managing to build an enduring vision of mobile anti-capitalist urbanism that continued to evolve over the course of nearly two decades.


 

MAP Studio, the practice’s experimental arm, has been using AI visualisation tools like Midjourney, Dall-E, and Stable Diffusion since 2022. The book’s useful introductory essay by Kyle Steinfeld describes their eruption into architectural visual culture in paradigm-shifting terms – this software arriving “suddenly, severely, and seemingly out of nowhere” – and also contextualises them in a short history of AI since the 1950s.

Despite, however, the credit given in the acknowledgements to Venetian professor of the social history of science, Matteo Pasquinelli, little of the critical history of AI documented in his book The Eye of the Master has found its way into Steinfeld’s account. In his writing, Pasquinelli offers all the historical evidence needed to reach the conclusion that AI should never have been labelled ‘intelligent’, and that the use of the term Artificial Intelligence is essentially a marketing play.

The studio appears to use AI tools to push the boundaries of architectural imagination, describing their role as that of a “synthetic collaborator”, rather than a mere “computational tool”. They collaborate in an iterative process between human and machine, pushing the resultant imagery into previously unknown and unexplored territories, though one of the disappointments of this volume is that this process is never really articulated.




It would have been fascinating to see exactly how speculative imagery developed from prompts and interactions, and for the book to have visually and textually mapped the steps in this creative journey. I think there’s a simple explanation for this lacuna, apart from – perhaps – professional, proprietorial concerns about giving their game away: the entire project is in thrall to the portfolio as a presentational form. Images are labelled, but without commentary. If they are meant to tell their own stories, then they are notably quiet about it – the viewer is challenged to make sense of the aesthetic offer.

Take one of the two major speculative visualisation projects represented here, A Topography of Chance, which Christiane Robbins, one of the book’s authors and Director of MAP, describes as an “ontological provocation”. The process pushes, apparently, beyond being a series of aesthetic experiments into “ontological investigations into the changed nature of architectural thought itself” – though throughout AXA there is no clear explanation of the relationship between such high theory and the visual end-result, which can make the theoretical elaborations seem pretentious or gratuitous and the images random.



It is telling that there is precious little architectural theory per se in the book. The theorists it refers to are better known for their broader views on modernism and postmodernism, and whose spatial ruminations are, whilst interesting, a small subset of their fields of investigation: Baudrillard, Benjamin, Deleuze, Barthes, Fisher, Virillo. Architectural theory has found it almost impossibly hard to find a way out of, and forward from, the philosophical dead-end of the new materialisms that succeeded postmodern theory. The authors of AXA are perhaps aware of this glitch, and to their credit evoke concerns about racial bias in AI, the tech industry’s brainless reprise of New Frontier ideologies, the profound ecological damage wrought by data centres, and so forth – but it feels like they go nowhere, another impasse.




Looking at the book’s finals section, Practice, one can see why. The site-specific could-theoretically-be-built projects it features would end up, if carried forward into the physical world, as multi-million dollar constructions, a fact that points to a core contradiction in contemporary architectural practice. Many architects genuinely, I suspect, recognise and would like to address the multiple systemic crises the world currently faces, but the majority of their designs are impossible to realise without substantial, if not huge, amounts of money. The profession is, by default, in hock to capital. Which I believe is why its theoretical foundations are stuck in the mire of disembodied, dematerialised post-postmodernism.

Once you start incorporating considerations of race, ethics, class, social and economic inequalities, power, and autonomy into your theories, designs, and built projects, you’re going to come head-to-head with capital. The vast majority of architects cannot, literally, afford to do that. And even if they could, Large Language Models would be of little use, if any: they don’t have the necessary values, ethics, or – surprise, surprise – humanity. And when they are, superficially, taught these qualities, they’re invariably the wrong ones. I, for one, wouldn’t want to live in, or anywhere near, a house that had been specified by Palantir.

 

Architecture X Architecture: A Dialectic is an elegant performance and by looking through the lens of MAP’s work, a crisp snapshot of where the profession has got to in its adoption of AI. It ends up articulating, perhaps inadvertently, the divergence and cognitive dissonance between the blunt realities of doing architecture, the contradictions of thinking it, and the actual utility of a grossly over-hyped technology. In that, the book is interesting, thought-provoking, and intellectually useful. The authors have genuinely tried to represent AI-assisted visualisation as more than slop, and in that the book succeeds: the resultant imagery is, for the most part, precise and intelligible.

But the extraordinary speculative structures rendered are sited in desert landscapes that remain, to some degree, uncanny valleys: they still just don’t feel right. Should MAP have simply kept these visions to themselves? After all, the process seems to be working well for the practice itself as a component of their creative development.