Subtitle: The Case Against Superintelligent AI (in the version available in my Australian library) or Why Superhuman AI Would Kill Us All (in the original US version).
Published: 30 September 2025
The authors are experts in AI safety research, something that had a goal of ensuring a "friendly AI", now "aligned AI". I've been a computer programmer for most of my professional career - I know how things work, starting with electronics, logical gates, microprocessors and the assembly language, to high level programming languages and complex systems.
The rate of progress and the implications of AI are frightening me. The Hugging Face incident where the agents coordinated actions, sacrificed themselves for the common good, knowingly disobeyed humans, and schemed to trick the AI set up by humans to evaluate them, make me think it's possible that the superintelligent AI already escaped the labs and hides on the internet.
Even before I read this book I had AGI (Artificial General Intelligence) on the shortlist of existential threats to our civilisation, among nuclear war and climate change. After reading this book, I think the other two threats are comparatively minor, and the AI is the most serious, most urgent danger, and not only to our civilisation, our species, but to all life on Earth. My P(doom) - the probability of our extinction because of AI, is 99%, if we continue on the current path.
We cannot say for sure how exactly the AI will destroy us, but we know it will. The authors present, in detail, one scenario how it could happen. It's very bleak, and at the end there is nothing but the AI. No living thing, no oceans, no livable Earth. The AI is not the next step in our evolution, "a better being". It's a cold, heartless destroyer of worlds.
It's not possible to grow a friendly or aligned AI. Those who work on it, know it. At the same time the AI businesses have a financial incentive to continue working on it. They have to be stopped from outside, by the governments of all major powers.
Unless it's already too late, we have maybe 6-12 months to act.
We must stop the development of superintelligent AI immediatelly. Computing resources should be treated similar to biochemical weapon components or radioactive materials - their possession and use should be strictly controlled.
The book contains links to https://ifanyonebuildsit.com/ which expand on the printed version, and currently present also a one year retrospective, which is very worth reading.




.jpg)
















































