Discussion about this post

User's avatar
Herbie Bradley's avatar

I investigated this direction for a while earlier this year and came away convinced that unilateral ASI -> <10% chance of DSA. Simply because DSA is very very hard: almost all hypothesized military capabilites that the ASI would provide would not be enough to reliably fully undermine second strike, develop weapons capable of decisive victory, etc. Becoming better than the best human hackers seems very very hard, drone swarms have potential defenses, etc.

I think countries are most likely to be sceptical of the idea of DSA and would not expect a rival country attaining ASI as anywhere near justification for a pre-emptive strike.

Expand full comment
Max Räuker's avatar

Thanks for sharing your thoughts, looking forward to reading more.

Probably similar to Herbie, I suspect you are overrating the degree to which capabilities approaching ASI will make other governments feel threatened to a similar degree as classical weapons, like ICBMs.

I expect even ASI will mostly be visible as a benign commercial product, and that relevant cyberoffense and military strategy applications will for a while start fairly invisible and slowly boil the frog.

I suspect it will seem to adversaries more like the increase of using computers for military purposes, and less like the development of nukes.

(Maybe this changes if the U.S. / China / Russia literally develop gigantic armies of Terminators).

Expand full comment
1 more comment...

No posts