Growing Pains
Karen Hao has written a deep dive into the rise of Sam Altman and OpenAI. It tracks the transition from non-profit to profit. Hao covers the episode of Sam Altman being fired as a risk to the company and then being reinstated. Reflecting her approach to the global growth of AI companies, she describes the working conditions of low-wage data labellers in Kenya, the Philippines, and Latin America screening graphic content for these companies – the global, human, and environmental cost of the artificial intelligence boom. The stories she tells of workers exploited around the world “bear the brunt of all these problems; those dreams ring hollow. These dreams are of how AI will create abundance and end poverty.
Drawing from over 250 interviews with insiders, internal communications, and field reporting from around the world, Hao frames the race for Artificial General Intelligence (AGI) through the lens of modern colonialism—where tech giants extract data, computing resources, and low-wage human labour to concentrate power and wealth.
The first part of the book is about Sam Altman, the man, and the second part is about the growth of OpenAI. The story about Sam Altman has his life picked over in detail in this book. Hao goes into detail about his sister and her misfortune. This diversion struck me as basically irrelevant to the central theory about empires and AI, and more about Sam Altman’s own personality traits. Hao does try to present both the sister’s and brother’s stories and a rather disturbing allegation about child abuse, but Hao ultimately leans towards Sam Altman being wealthy and not using it to properly support his sister. I felt sorry for him, having his personality scrutinised. A man so intellectual and successful is to some degree a sociopath, so nothing new here, and nothing so damaging. He is an elite tech bro, but Musk and the Zuck probably have weirder stories to tell.
AI as a thing is only really covered early on in the book. She posits that, on the one hand, AI is “AI is thrilling: a creative aid for instantly brainstorming ideas and generating writing; a companion to chat with late at night to ward off loneliness; a tool that could perhaps one day be so effective at boosting productivity that it will increase top-line economic activity.” On the other hand, she goes on: “Generative AI models are monstrosities and their effect on impoverished communities around the world.” Which then frames the book.
AGI is a fast moving revolutionary technology with many challenges that need to be righted, but like the internet these may never happen. Referencing hallucination she refers to a lawyer who used ChatGPT to perform legal research and prepare for a court filing. “He was subsequently sanctioned, fined, and publicly humiliated after discovering too late that the chatbot had made up everything it told him……”
She uses the word “metaphor”- relating the current industrial revolution of AI to European colonialism, empires seized and extracted resources that were not their own and expolited labour. “She argues that in her view these new empires follow the same path. In particular, “OpenAI is “now leading our acceleration toward this modern-day colonial world order… and every tech giant is racing to out-scale one another.”
OpenAI’s departure from its roots as a safety-first non-profit organisation to a commercial risk-taking company is portrayed through the tensions between different teams about this. Not least how OpenAI got into bed with Microsoft to fuel that growth. The vibe is one of a hypocritical sell-out, but the facts she covers are that OpenAI had little choice if they were to grow quickly enough and protect their own IP.
The brief firing and reinstatement of Sam Altman is detailed for the farce that it was and the damage it briefly did to the company – growing pains due to weak management – the staff who left to form Claude are real challengers to OpenAI both in size and ethics.
There are serious ethical issues with the human screening of data collected. Her research into low-wage data labelers in Kenya, the Philippines, and Latin America screening graphic content, is on a par with sweat shops in the Asian sub-continent and the problematical monitoring of live content on social media. Somehow it has not hit the same level of headlines. These issues are more than just about OpenAI and potentially a wider industry sore as the exponential growth is highly competitive with little room for ethics. She states that “in an arms race all bets are off.” And how OpenAI is in its own arms race with Claude et al.
Karen Hao’s style is more journalistic than technical analysis of AI, which makes the book accessible without getting bogged down in jargon. The reporting is based on hundreds of interviews, Slack logs, and internal documents, and it shows both in the confidence of her story and in how the minutaie reads better as a body of reference than a story. Not a book for the casual interest in the growth of one of the most exciting technologies of our age, more an academic account and hypothesis of modern colonial tech bros in action.
Her epilogue is entitled “How the Empire Falls. It is not an entirely convincing one, more plausible perhaps. The logic is strong on how communities can control their own destinies with AI by managing their own LLMs, using them to protect their own societies. This ground-up approach is backed up by real examples but lacks any threat to the global “empires” she rightly argues are on a land grab; I would argue too busy to notice the minor players.
One is left with a bad taste in the mouth. The march of technological innovation has taken capitalism, in which the negatives vastly outweigh the positives. In the same way, some people do not want to know how the McDonald’s patty came to be; people don’t want to know where the answer to their “prompt” came to be.
