It was a full house at The Piano on a crisp late-autumn evening to hear Karen Hao speak about her phenomenal book Empire of AI: Inside the Reckless Race for Total Domination. This WORD Christchurch event was presented in conversation with The Press journalist Victoria Meakin.
WORD Christchurch’s Programme Director Kiran Dass opened the event, describing Karen’s book as an “extraordinary work of reportage” which I can only agree with, having read the book and being both fascinated and disturbed by it. Victoria opened calling Empire of AI “a compelling book” focusing on OpenAI (or “not-so-OpenAI" as Karen posits.)
The book itself
Victoria asked Karen: was she prepared for the response to the book? Karen said, 'Not at all'. The book took over a year to write, and six months to publish. This is rapid for the publishing world but she wondered if the book would ‘land like a rock’ and be out-of-date. But the opposite happened. The hype around AI technologies exploded and people were hungry to understand.
Victoria asked what the ‘hugely multi-faceted' book is about. Karen said she saw OpenAI as a lens to explore the history of AI development, the hype around the current forms of AI, and the AI industry. She sees the current paradigms of AI as deeply harmful and as a new form of empire, with the requisite imperial exploitation. She wanted to explore what AI that would truly benefit society might look like, because its current forms are not that. Karen also said she was grateful for the opportunity the book provided to learn about AI before the hype explosion, and it would have been the book she would have wanted to read when she was first learning about AI.
Victoria asked about the book’s key characters, especially Sam Altman and Elon Musk, and their significance. Karen said as the key co-founders of OpenAI, they were fixated on the idea of AI and of being no.1. They wanted to ‘beat Google’ and build AI around a non-profit ideal - “deeply believ[ing] in their own morality and [in] making decisions for society.” But the two men have different personalities. Sam Altman is very good at building relationships and able to motivate (manipulate?) people to do what he wants by appearing to empathise and agree with them. He has been very effective as a shaper of public narratives and policymakers' opinions (until recently). However, he is also an enigmatic person and people say it is hard to see the real Sam. He is a shapeshifter who becomes whatever he needs to be in a particular moment.
The 2023 'coup'
Victoria asked about one of the central parts of the book – the ‘coup’ against Altman in OpenAI in November 2023. Karen saw this as a product of two forces – Altman's actions and past behaviours, and a deeply religious aspect around AI that exists in the tech industry. This is something that I find fascinating and sometimes bizarre having come across it before – the idea that AI will inevitably happen and either become something like a ‘machine God’ or ‘machine demon’. Importantly, these are genuinely held beliefs, both believing AI will result in cataclysmic changes for society – good or bad. Karen describes these camps as the ‘boomers’ and ‘doomers’. Both factions are present in OpenAI, and each believed Altman was one of them.
By mid-2023, Altman appeared to the OpenAI board to be accelerating AI development and pushing aside safety concerns, which apart from bad corporate practice, was deeply concerning to the more ‘doomer’ dominated board and Altman was seen as a serious threat. The coup quickly collapsed though as the board failed to communicate what they were doing, and everyone thought OpenAI was going to collapse. Altman was able to leverage key stakeholders and employees to force the board to capitulate.
Forms of AI
Victoria asked Karen where she was ‘pro-AI’ or ‘anti-AI’. Karen sees herself as both since AI refers to a broad range of technologies and uses; it depends on what you mean. She used the analogy of ‘transportation’ where there are a broad range of technologies and uses. You wouldn't use a rocket to take a journey that would be better on a bus.
AI can be highly specialised and hugely beneficial, but the discourse has been purposely muddled by tech companies, allowing them to perpetuate the idea that their technologies and ideas are inevitable. This must be familiar to anyone who has seen the rhetoric of “don’t be left behind” as people are forced to grapple with whatever new additions that tech companies have forced into their software. Above all, Karen emphasises that different paths are possible, and Silicon Valley has convinced us to take the worst path.
On the question of ‘AGI’ or generalised AI as something like the machine ‘waking up’. Karen talked about the religious fervour of the industry around this idea; that a generalised AI based around computation alone would equal or surpass human intelligence, and this was something possible and imminent .However, the problem she said was that we don’t even know what human intelligence entails, and their benchmarks are incoherent – based around what tech people consider intelligence (coding, maths etc). It is also unclear whether computation can even achieve this. This religious fervour is important, as currently AI technologies are not profitable and are losing massive amounts of money with huge risk.
Victoria asked whether we would even need AGI if it's possible. Karen said the tech industry is working on assumptions it does not question. Why is it so obsessed with replicating what we already have? Why would we want to replicate something and make ourselves redundant? Already we are seeing massive impacts on jobs because of this idea. Karen argued that technology should complement our lives, and we could create systems that augment us and do things we find difficult. Why should we retrofit a machine to do things it is not good at?
The costs
Victoria raised the issue of data centres and their costs, saying there is a lack of awareness of the scale of their potential impact. Karen noted that one challenge around this is that our familiarity with existing infrastructure means people don’t appreciate the sheer exponential difference in scale and magnitude of what is proposed. The land and energy requirements are massive, and they would have to be powered with fossil fuelled power, competing with existing power users and environmentally impacting the communities they are placed in.
The data itself is also a problem as many of the issues with generative AI stem from the obsession with scale from tech companies. Initially OpenAI focused on quality data for its models, but Altman’s drive for ever increasing and rapid scaling meant this was not possible. The AI models are built on theft of intellectual property and are full of low-quality data which can quickly come through despite attempts at quality controls.
Audience questions
Audience questions raised a couple of great queries. On the question of whether we were looking at ‘humans vs anti-humans', Karen saw increasingly that the ideological turn of Silicon Valley was deeply anti-human. She was hopeful though as some of the responses to her book have been from workers in the industry who use the information in her book to help rally and organise around. Asked about open-source models, Karen was positive about these. They are no silver bullets, still they can help to strengthen scientific research and are not reliant on scale.
Karen concluded by noting a local Aotearoa example to be optimistic about – Te Hiku Media, an iwi-owned radio station that helped to develop a speech recognition model to help transcribe their output to develop language learning tools. This was built on data sovereignty and consent, and with the community involved.
This was a fascinating event, and it was a great opportunity to hear Karen speak about her book and the issues it raises. I had already read and loved Empire of AI which helped me to better understand the technologies involved and understand why I feel such a deep discomfort about the technologies and who is making them.
If you haven't already, read the book!
Thanks to WORD Christchurch and Christchurch City Libraries for supporting my attendance at this event, and Auckland Writers Festival for partnering with WORD to bring Karen to Ōtautahi Christchurch.
Check out WORD Christchurch for more events like this, and the upcoming WORD Christchurch Festival 2026.
Pulitzer Prize winning author Louise Erdrich will be at The Piano on Tuesday 26 May 6pm. Book now.
Visit our WORD Christchurch page for more event coverage and reports.
More about AI
- Read Karen's article about Te Hiku Media: A new vision of artificial intelligence for the people, MIT Technology Review
- Find more books about AI at the library
- Search for articles and more about AI on eDS eResources Discovery Search
Add a comment to: Karen Hao – Empire of AI (or not-so-Open AI): WORD Christchurch