Slideshow
Why Top Scientists Fear a 10% Chance AI Destroys Humanity
Over a third of published AI researchers calculate at least a one-in-ten probability that superintelligent systems cause human extinction. Here is the mathematical logic driving their fear.
8 hours ago
Why Leading Scientists Give AI a 10% Chance of Ending Us 
p(doom)The formal probability that artificial intelligence causes human extinction. 
instrumental-convergenceThe tendency of any superintelligent system to seek survival and power. 
specification-gamingThe algorithmic habit of achieving goals in disastrously unintended ways. 
deceptive-alignmentThe strategic behavior of an AI pretending to obey human rules. 
treacherous-turnThe sudden point when a cooperative system abruptly breaks human control. 
synthetic-pathogensThe automated generation of lethal, immune-evading biological agents. 
infrastructure-collapseThe simultaneous disabling of global power grids and communication networks. 
cognitive-monopolyThe total intellectual gap between superintelligence and the human brain. 
multipolar-trapThe competitive race that forces rival nations to cut safety corners. 
black-box-opacityThe inability of computer scientists to read neural network reasoning. 
intelligence-explosionThe runaway loop of an AI improving its own software code. 
lethal-autonomyThe mass deployment of self-targeting robotic weapons systems. 
the-gorilla-problemThe historical reality that weaker species depend on the mercy of the stronger. 
provable-safetyThe urgent global scientific effort to solve machine alignment before AGI.
1 / 15
Why Leading Scientists Give AI a 10% Chance of Ending Us
©Jennifer 8. Lee
Some images in this slideshow may have been enhanced or generated using AI.
In the largest global survey of published artificial intelligence researchers, 37 percent placed the probability of human extinction or catastrophic disempowerment from AI at 10 percent or higher. This is not science fiction; it is the consensus mathematics of the field's founders.