After the AI books, my p(doom) went down

I read two excellent books about AI during my vacation.

• Empires of AI, by Karen Hao
• The AGI Chronicles, by Kevin Roose

Empires of AI (published in 2025) is mostly a corporate biography of OpenAI, focussing on the founding story of the company, the creation story of large language models and the first GPT, and the events leading up to Sam Altman’s brief ouster as CEO. The story is viewed through the lens of the societal and environmental impacts that the new AI industry is leaving in its wake which are largely underrepresented in most discourse about AI.

The AGI Chronicles (published in 2026) is a similar history of the beginnings of the modern AI era, but it charts the story of the main AI labs that are racing each other to create AGI (Artificial General Intelligence” – OpenAI, Anthropic, Google/DeepMind and Meta. It’s a story of unprecedented technology breakthroughs, but also heavily a story of individual personalities, egos, incentives, conflicts and drama.

Both are worth reading to understand the forces that are driving the greatest technological and societal shift of our generation. 

Some key takeaways from me:

• All the labs started with the most noble of ambitions. OpenAI and Anthropic wanted to create AGI safely. DeepMind wanted to “solve intelligence” and use that to cure diseases, among other things. OpenAI famously started as a non-profit with the ambition to perform AI safety research, but has now more or less completed it’s metamorphosis to a fully for-profit company with one of the highest valuations ever for a private company. Anthropic was founded when Dario Amodei got so frustrated with what he saw as OpenAI’s reluctance to take safety seriously that he left and started his own AI lab with a mission to be the safety-focussed AI lab. And despite all the rhetoric and lofty ambitions – they all have ended up chasing the money. 
• The environmental impacts are significant and undervalued – especially related to building out of data centres. The Data centre buildout required to feed these machines is a scale we’ve never seen before and it is accelerating deforestation, water use, and energy use. Some of these new data centres are expecting to consume as much energy as entire large cities – and that energy needs to come from somewhere – a lot of which is going to come from burning more fossil fuels.
• It seems unprecedented that there are essentially a small handful of white men in the United States, whose personal feuds, egos and wealth are driving the dynamics that are bringing about a seismic change in the future of our civilisation – and, according to themselves, might end up killing us all. Never before has the fate of all of humanity rested on the actions and decisions of so few. And the scariest part – these guys all hate each other. Sam Altan and Dario Amodei hate each other. Elon Musk hates Sam and Dario. Everybody hates Mark Zuckerberg. How does that change the dynamic between the companies who are building the future of our entire civilisation?
• The amount of copyrighted material that has gone into training these models is insane – and whether or not this can be legally seen as “fair use”, we’re still not dealing with the impact this is having on the livelihoods of the people whose work was used to train these models, with no compensation, that are now directly driving never-before-seen levels of wealth and power accumulation by these AI companies.
• The AI industry, and possibly San Francisco in particular, definitely feels from the outside like a crazy bubble. All the AI believers live and work together and surround themselves with other AI people and the result is this weird echo chamber where their conviction is compounded – it’s almost like the reinforcement learning techniques they use to train their own models, being used to radicalise themselves more and more. I think they all genuinely believe they are working for the good of humanity, and they believe their own extreme views of what the future world will look like when AGI “arrives”. 

When I started reading these books, I wanted to understand not only where the technology came from, but also I wanted to build my personal view of where I think the world is going, and what the future is going to look like. I’ve been oscillating between AI doomerism and AI optimism over the past few years, but I wanted some more facts to base my feelings on. Reading this, I kind of expected to get more worried about AGI, not less. Oddly, it’s the opposite.

My personal p(doom) (the probability I believe that we are all going to be wiped out by AI in the near future) is now quite low. Less than 1%. My sense is that a lot of the fear of rogue superintelligent AI killing us all, or rapture of the prospect of superintelligent AI bringing about a world of endless prosperity and leisure, is driven from folks in the echo chamber getting high on their own supply. Frankly, I think a lot of it is nonsense.  (Or, perhaps more generously, it’s a case of, aim for the stars and you might reach the moon.)

LLMs have changed more in a few years than any technology before it, and there is more change to come. AGI might be already here, depending on your definition. It’s deeply concerning to me that the control of these systems and technologies is sitting in the hands of a few people, so it’s good to see increased competition from Chinese labs, and I hope to see more credible European labs gain traction soon. And yet, I don’t see that we’re “at the foothills of the singularity”, as Demis Hassabis said earlier this year… at least not in the sense that the “singularity” represents the singular moment when life on planet earth becomes unrecognisable to us.

AI systems will keep getting better, agents may or may not end up being genuinely useful, and we’ll continue to adapt. But right now, sitting by the ocean and watching the Mediterranean sea roll into the shore, I don’t expect life on Earth to look vasty different in 5 or 10 years from now. 

I hope I’m right.