David Krueger

None of these approaches that you'll often hear mentioned—interpretability, testing or alignment—is technically adequate. We don't know how to build superintelligent AI safely. The plan is basically to roll the dice.

We need to course correct immediately and work to prevent the development of superintelligent AI, and we need to do this internationally, which will take time. We cannot afford to wait for more evidence that the risk is imminent.

AI might literally lead to human extinction, in the same way that humans have caused many other species to go extinct.