paulfchristiano comments on Cryptographic Boxes for Unfriendly AI - Less Wrong

24 Post author: paulfchristiano 18 December 2010 08:28AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (155)

You are viewing a single comment's thread. Show more comments above.

Comment author: paulfchristiano 19 December 2010 02:05:26AM 3 points [-]

A provably correct implementation of any reasonable encryption scheme requires automatic verification tools which are way beyond our current ability. I strongly suspect that we will solve this problem long before AGI though, unless we happen to stumble upon AGI rather by accident.

I think that you could implement the scheme I described with reasonable confidence using modern practices. Implementing encryption alone is much easier than building an entire secure system. Most flaws in practice come from implementation issues in the system surrounding the cryptography, not the cryptography itself. Here the surrounding system is extremely simple.

You also have a large advantage because the design of the system already protects you from almost all side channel attacks, which represent the overwhelming majority of legitimate breaks in reasonable cryptosystems.