Empire of AI (Books column)
I finished reading a book yesterday.Empire of AI" from Karen Hao, which he bought from Foyles during a trip to London. In general, it was difficult for me to break away from reading this book and while flying to Moscow, I read it by a third - you seem to already know the main events around OpenAI, but in a coherent story they begin to look quite different. This book is not about ChatGPT as a product or how transformers are built. Rather, it’s about the early years of the GenAI revolution and a company that finds itself in a strange position: simultaneously a research lab, a startup, an ideological project, a Microsoft infrastructure client, and an organization that claims to be building technology with risks for all of humanity. I especially liked the following points.
1Dynamics of the OpenAI Foundation In retrospect, it’s easy to think that everything went straight: smart people got together, took a lot of compute, made GPT, then ChatGPT, then everything is clear. But the book clearly shows that there was no straight path. There was ambition, fear of AI concentration in big companies, conflicting mission perceptions, money, ego, research uncertainty, and a very fast-changing reality.
2Friction between Elon Musk and Sam Altman This is not just a dramatic detail for a biography. Through this conflict, one can better see how different models of power and control can be behind the same public rhetoric about “safe AI for humanity.” Who makes the decisions? Who owns the levers? Who determines what security means? In AI companies, these issues are not secondary, but architectural. Interestingly, just recently ended the trial between Max and OpenAI on the topic of conversion from a non-profit organization to a commercial one.
3OpenAI: Applied, Research and Safety For me, this is one of the main engineering layers of the book. Applied is drawn to product, users and product deployment. Research moves the frontier and lives on the logic of scientific breakthrough. Safety tries to keep the issue of consequences and risks under control. All three cultures are understandable and necessary. But in one company they create constant tension: speed versus caution, product versus research, mission versus market.
4Partnership with Microsoft The book helps to feel that GPT breakthroughs were not just the result of models and data. It’s also a story about infrastructure, finding capital, accessing compute, the ability to turn research into a working system, and the pace at which normal organizational processes begin to fall apart. Without this layer, it is easy to romanticize AI as a pure science, although in practice frontier is driven by a bundle of research, infra, product, funding, and distribution.
5Crisis with the Board of Directors I remember it well as news chaos: Altman's firing, return, employee pressure, Microsoft in the background, strange statements and almost total opacity. But in the book, this episode reads not as a sudden crash, but as the result of contradictions that have accumulated from the beginning: nonprofit mission, commercial reality, safety fears, personal power, and the cost of leadership on the frontier.
Separately, this book tells the story from the author’s point of view of Karen Hao, who graduated from MIT and worked as a long journalist covering events in the field of technology and AI. In the end, she conducted an investigation into the 300+ interview and collected a lot of materials to be able to assess what was happening behind closed doors of the company OpenAI:)
I recommend the book to anyone who wants to understand GenAI not only as a set of models, benchmarks and APIs, but as a system of people, power, infrastructure and organizational compromises. Especially useful for engineers, managers and architects: after this book, it is better to see that the frontier moves not only in the laboratory, but also in the structure of the company.
#Books #AI #OpenAI #Engineering #Management #Research