Dima Krasheninnikov warns about extinction and loss of control

Dima Krasheninnikov personally judged a future out-of-control AI plausibly at least 10% likely to kill everyone. He considered permanent human disempowerment more likely still. The post gives no time horizon for the probability. It is his personal forecast. Dima Krasheninnikov 

Krasheninnikov’s profile identifies him as an Anthropic AI safety researcher. His research includes interpretability, control and security. Dima Krasheninnikov Dima Krasheninnikov