Rather wildly ambitious. But who knows.
I think what’s happening is that AI is accelerating time lines.
I would guess that because this stuff is all trial and error and they can calculate how much progress they make on average per iteration—that, further, they can calculate how many iterations on average it will take them to reach their goal—judgeing by the distance they have already come on average between iteration. Then they calculate how long it takes them on average to complete each iteration. then its just multiplication.
What AI does is enable them to more rapidly complete each experimental iteration. And they’re calculating more aggressive timelines based on the acceleration speed that AI promotes.