Alicorn 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: Alicorn 19 December 2010 03:49:31PM 9 points [-]

Or you build yourself a superweapon that you use to escape, and then go on to shut down your company's weapons division and spend your spare time being a superhero and romancing your assistant and fighting a pitched battle with a disloyal employee.

Comment author: shokwave 19 December 2010 04:26:01PM 9 points [-]

This reply and its parent comment constitute the "Iron Man Argument" against any kind of "put the AI in a box" approach to AGI and friendliness concerns. I predict it will be extremely effective.

Comment author: DanielLC 29 January 2013 01:04:30AM 0 points [-]

That doesn't really work in this situation, since the AI building another AI won't get it out of the box.

It's more like you can only design things, and the evil dictator can tell if it's not really a doomsday device and won't build it.

Also, he's not going to be checking on your progress, so faking it is useless, but it's also unnecessary.