the ethics of rare AI errors: mistakes that kill and need for truly human

Artificial intelligence must be 99% reliable, but it's the remaining 1% that can kill.In a world where AIs dialogue, diagnose, reassure, guide, evaluate and sometimes decide in our place, one truth remains invisible to the general public: it's not the frequent errors that threaten our safety... but the rare ones.These mathematical micro-drives, statistically invisible, can nonetheless lead to a misdiagnosis, a bad medical recommendation, a biased judicial decision, an ignored suicide risk or a catastrophic financial rating.With implacable clarity, Camille VANARD reveals the hidden face of modern AI systems: their inability to handle extreme cases, structural blindness to weak signals, silent biases that destroy lives, and the total absence of global legal accountability.Through unprecedented analyses, rigorous models, shattering case histories (medicine, psychiatry, justice, finance, education, robotics) and the revolutionary concept of rare errors, this book finally offers a new reading grid for understanding real risks.

Pagina's: 248, Paperback, Our Knowledge Publishing