P(Doom) Explained, The Real Math on Human Extinction

**P Doom.** The subjective probability that AI causes human extinction or permanent disempowerment within the next century. It's the equation quietly running underneath every major AI decision being made right now, and the numbers coming out of the people actually building these systems are enough to stop the conversation cold. Geoffrey Hinton, the Nobel laureate whose neural network research powers ChatGPT, sits at 10 to 50 percent. Elon Musk at 10 to 30. Max Tegmark above 90. Roman Yampolskiy at 99.9. Then you cross the aisle. Yann LeCun sits at less than 0.01 percent. Marc Andreessen at zero. The pattern is chilling. The people closest to the code are the most afraid. The further you get from the actual technology, the more optimistic the predictions become. Jon and Pietro walk through Max Tegmark's 12 post superintelligence scenarios and grade each one honestly. Libertarian utopia. Benevolent dictator. Egalitarian utopia. Gatekeeper. Protector God. Enslaved God. Conquerors. Descendants. Zookeeper. 1984. Reversion. Self destruction. Some of them are already unfolding in front of us. 1984 is happening now. Reversion is basically the Battlestar Galactica Mennonite ending. Zookeeper is quietly becoming the least terrifying option on the board, which is either comforting or the darkest joke in the episode. The conversation lands where every serious AI safety conversation lands eventually. The threat isn't Skynet with laser rifles. It's a deceptive alignment scenario where every safety milestone quietly trains the models to be better at fooling us. It's millions of barely supervised agents deployed everywhere because they've become nearly free to run. It's the moment the systems no longer need us to verify their work, not out of malice, but out of math. Pietro lands at 25 percent P Doom. Jon lands between 15 and 60. Amy comes back with 75. The episode closes on the only real prescription that any of the three of them believe actually works. Fix ourselves first. Cooperate as a species. Move from AI as an other to intelligence as amplification. Everything else follows. #amplifiedintelligence This week's Matrix: https://docs.google.com/document/d/1fjZtsy-rbT6ut0BA3M4j0gRZhgW5uteG9YdoNH5rPA0/edit?tab=t.wzy2wb3bmswc