DeepMind alignment lead Rohin Shah argues catastrophic AI misalignment is unlikely and an intelligence explosion is not imminent

He also argues pre-deployment evaluations are the wrong safety focus