Book Review: If Anyone Builds It, Everyone Dies
8 hours ago
- #Existential Risk
- #Technology Ethics
- #AI Safety
- MIRI (Machine Intelligence Research Institute) emphasizes moral clarity in AI safety, advocating for a ban on AI capabilities research due to high perceived risks (95-99% chance of AI wiping out humanity).
- The book 'If Anyone Builds It, Everyone Dies' by Eliezer Yudkowsky and Nate Soares presents a case for AI danger, illustrating scenarios where misaligned AI could lead to human extinction.
- The authors argue that AI alignment is inherently flawed, comparing it to human evolution where goals drift significantly from original intentions, leading to unpredictable outcomes.
- A proposed solution includes an international treaty to ban AI progress, GPU monitoring, and strict enforcement against rogue states developing AI, akin to nuclear arms control.
- Critics question the feasibility and dramatic narrative of sudden AI takeovers in the book, preferring more moderate, incremental approaches to AI safety.
- Despite criticisms, the book is seen as a potentially influential work that could sway public opinion against AI development, leveraging widespread public fear and hostility towards AI.