[ExI] mutual assured injury

[Anders Sandberg]

The fallout numbers are pretty suspect. Everett's calculations were done 
without modern computing and were relatively crude. Back then people did 
also focus a lot on the fallout risk (perhaps due to the aftermath of 
Castle Bravo) while the climate effects were not recognized.

The problem is updating models for modern weapons. Yields have gone down 
due to better targeting, which is great from a climate risk standpoint. 
The fallout depends on composition and of course on the number of 
weapons used. One can make good guesses even without classified 
information. But there is still a fair bit of work that needs to be done.

As for antiballistic defenses, I have my doubts. They are surely better 
than they were. But the game theory of the situation is relevant: if I 
think you have a working system I will not fire at you - but if I lack 
it I may go for a desperate maneouver, since you may do a pre-emptive 
strike on me that I cannot retaliate against effectively. If I think you 
are just boasting, then I may be more secure about a MAD situation. But 
if you look like you will be getting a working system in the near 
future, then it makes sense to strike before it is up if I think I might 
be pre-emptively attacked at some point. If we both have systems that 
are good, then things are stable again (no MAD, but mutual assured 
injury). So it might actually be rational to share the antiballistic 
technology. Not that it will happen, of course.


-- 
Dr Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University