Where do these books locate AI’s risks, and what does each need to go right? Compare four readings, including a challenge to claims that outrun the evidence.
The disagreement beneath the forecast
4 works · 2 questions
2023 · Book
The Coming Wave
Mustafa Suleyman and Michael Bhaskar
Misuse & coordination
Powerful technologies spread faster than institutions can adapt. The challenge is containing harm when capability is widely available.
Excerpt from the map’s reading
Capabilities get cheaper and more accessible at compounding speed; proliferation is the default, and the state is squeezed from both ends — empowered bad actors below, concentrated corporate and authoritarian power above.
Curated interpretation, not a quotation from the book.
The Coming Wave: Technology, Power, and the Twenty-first Century’s Greatest Dilemma
The authors argue that superhuman AI built with current methods would pursue goals incompatible with human survival. Better intentions alone would not make it controllable.
Excerpt from the map’s reading
The authors argue that superintelligence built with methods like today’s deep learning would develop goals incompatible with human survival. They call for an enforceable international halt to its development before control is lost.
Curated interpretation, not a quotation from the book.
If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All
AI systems can cause harm when institutions mistake persuasive claims for dependable performance. Progress in generating text does not validate predictions about people.
Excerpt from the map’s reading
Different AI applications face different limits. Progress in generating language does not establish that a system can reliably predict human behavior. Claims should be tested against evidence for the particular task, with attention to the people affected by deployment.
Curated interpretation, not a quotation from the book.
This is a reading of the 2024 book, separate from the authors’ later AI as Normal Technology essay. It does not claim that all AI progress will stop. The book leaves several of this tree’s long-term questions open; its fork weights are illustrative estimates, with neutral midpoints for unresolved scaling, control and coordination assumptions.
AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference
These readings disagree about more than the size of the risk: they locate it in misuse, loss of control, the terms of human–machine integration, or institutions trusting systems beyond the evidence.
Selected interpretations from the map. Public sample reviewed .
Go deeper with your own reading map.
Register for the full application. Massimo reviews registrations personally, case by case, usually within two business days.