EmpiricalRequires evidence

Ord's figures already assume we respond, and the gap between them is the part that belongs to us. His estimates incorporate the possibility that we get our act together; had he assumed business as usual they would have been substantially higher. His own image: we are heading toward a Russian roulette with two bullets in the gun, he thinks we will remove one before it is time to pull the trigger, and there might be time to remove the last one if we really try. So the headline should not be the risk he expects to remain, about 1 in 6, but 2 in 6, which is one in three: the difference between a lacklustre effort by humanity and a heroic one. Part I's upper bound of 1 in 2 should read 1 in 3. The 1 in 2 nearby is Ord's AI capability timeline, not a risk estimate.

Risk landscape