Exquisitely researched page-turner essential reading for those most unlikely to read it
Empire of AI: Inside the Reckless Race for Total Domination by Karen Hao
Allen Lane, London, 2025; 482 pages.
Reviewed by Dr John Kannenberg
Almost a decade ago, the online museum world made collective pledges to decolonise their collections and curatorial practices, to dedicate themselves to social justice issues and “sustainability”, or climate-friendly practices. The very much welcome and necessary #museumsarenotneutral was a movement initiated by American scholars La Tanya Autry and Mike Murawski that acknowledged the inherently political nature of museums – institutions born of colonialism and empire – inviting them to challenge the dark past of their origins and find new relevance as temples of inclusion, understanding, and critical thinking.
I guess that’s all out the window now since 2026’s MuseumWeek is “celebrating” the use of “AI” in museums – most egregiously by generating their very own deepfakes and pseudo-Impressionist illustrations for their social media posts using AI rather than human creativity and craft – things museums were created to preserve in the first place.
“AI” itself is a contentious, inaccurate term: it’s sloppy and imprecise, because what is now currently being marketed as “Artificial Intelligence” is primarily one very specific type of what’s better referred to as machine learning and pattern recognition technology. The AI being hyped at present is mostly what’s now known as Generative AI: AI that has scraped – or “been trained on” – copyrighted human creative outputs like visual artwork and writing (usually swiped wholesale from the public internet, which is overwhelmingly biased towards English-language sources and rampant with complicated biases like racism, sexism, homophobia, etc.), in order to be able to spit back computer-generated “content” such as illustrations, research papers, deepfakes, and Spotify music – not actual artwork, but digital approximations of creative “outputs” built from vast online archives that have been mined without permission (not to mention that Generative AI produces a large percentage of “hallucinations” i.e. bullshit simply because it’s programmed to give a confident answer, even if it can’t find one).
It’s this type of AI that’s being pushed by companies like OpenAI, Microsoft, Meta, and Google, companies who are locked in a desperate race to invent “Artificial General Intelligence” (“AGI” for short): computer technology that is or will be smarter and more capable than human beings. Except no one can actually define what AGI will be precisely, and no one can agree on what the criteria is to determine when the technology actually gets there (“we’ll know it when we see it” is what most Silicon Valley AI pundits tend to say in their defense). All they know is, they think they need to get to AGI as soon as possible, and they think the best and only way to get there is through expansion and scaling: more data centres, more exploited content cataloguing workers in the Global South, more precious minerals mined by children in the Congo, more drinking water to cool their high-powered GPUs.
These are the issues covered brilliantly by technology journalist Karen Hao’s international bestseller Empire of AI: Inside the Reckless Race for Total Domination, recently released in paperback for the first time. Ostensibly, it’s a history of OpenAI, the company responsible for ChatGPT, led by CEO Sam “It also takes a lot of energy to train a human” Altman. Hao’s book uses the drama surrounding the OpenAI board’s 2023 attempt to fire Altman as its emotional lynchpin, structuring the book like a contemporary Netflix binge-fest by beginning at the moment Altman finds out he’s been fired (over Zoom, no less), then flashing back for the rest of the book until eventually revealing the specific circumstances that led to the firing, then ending the book shortly after Altman was reinstated only five days later. In that respect, the book is a gripping page-turner of all-too-human drama, and I found it extremely difficult to put down once I started it.
Except Altman’s story isn’t actually the main point of the book; as Hao alerts readers in her Author’s Note, Altman’s ouster is used as a framework for explaining the mindset of Silicon Valley – specifically, the tech bros currently shoving “AI” down the world’s collective throat – in order to warn the world that the race for AGI is a deeply colonialist, imperialist project that is gutting the Earth’s natural resources and exploiting workers across the Global South, and will do overwhelming damage to us well before it actually functions well enough to have any chance of actually “replacing humans” in the Global North on a large scale.
Hao’s book is not an anti-AI screed; as she says herself:
The critiques that I lay out [here] are not by any means meant to dismiss AI in its entirety. What I reject is the dangerous notion that broad benefit from AI can only be derived from–indeed, will ever emerge from–a vision for the technology that requires the complete capitulation of our privacy, our agency, and our world, including the value of our labor and art, toward an ultimately imperial centralisation project.
Through extensive fieldwork in South America and Africa, Hao introduces readers to gig workers across the Global South who are economically exploited into taking slave wage content moderation jobs, forced to endure reading and watching the most horrible examples of physical and sexual violence, to the point that it destroys their mental health, just so that ChatGPT can have a slightly lesser chance of generating offensive content. We meet local activists fighting for the preservation of their access to clean drinking water as Google, Microsoft, Meta, and other tech companies pursue underhanded deals with financially desperate governments to build mega-sized AI gigawatt “campuses” (i.e. data centres) that depend on fresh water for cooling, co-opt local power supplies, and generate unbearable noise pollution for local residents.
Hao’s book drives home a concept that seems extremely difficult for the average (or even generally well-meaning and intelligent) person to grasp, the idea that “AI” isn’t actually some magical technology that falls naturally down to us from a “Cloud”: in reality, it’s dependent upon a vast physical and human infrastructure the same as any other technology has ever been throughout history, an infrastructure whose devastating impact on the natural world will help bring us to a global water shortage in what some scientists believe to be less than four years from now (the United Nations has already produced a report declaring that Earth entered a “global water bankruptcy” even before the rise AI’s mega-datacenters).
Beyond a shadow of a doubt, the “AI” being celebrated this MuseumWeek is one of the most fundamental moral issues of our age, even before it gets to the point where it causes massive job losses by replacing humans. AI is, at this very moment, killing people and our planet on a vast scale; to uncritically embrace this technology is to sleepwalk into a global disaster that has very little to do with some sort of Terminator/Skynet-like conspiracy theory. As Hao demonstrates in the book, there are many positive uses for AI within the realm of preserving human culture (her detailing of the Te Hiku project in New Zealand to preserve Māori language is particularly persuasive). These are the uses of AI that museums should absolutely be celebrating, not the vapid parlor tricks of Generative AI.
The types of AI we should question and resist most vehemently are those that are often described as “just another tool” to help our “productivity”, i.e. the Generative AI that tries makes humans less intelligent by thinking and creating for us, rather than one that helps us to think and create ourselves. Generative AI is demonstrably not “just another tool”: a hammer doesn’t pollute a town’s water supply, a camera doesn’t have to steal from every picture ever taken in order to capture an image, and a telephone doesn’t need to exploit gig workers in Kenya. Generative AI is existentially unlike any tool ever created before, and Karen Hao has provided here a deeply researched, extremely well-written warning sign that every one of us needs to consider as we make decisions on how much to embrace these new apparatuses of empire.
Museum of Portable Sound Rating:
🎧🎧🎧🎧🎧/5
Pros: • Excellent, exciting, well-researched, must-read
Cons: • Too long for the casual readers who desperately need to read it
The Museum of Portable Sound 