The "Friendly AI Critical Failure Table" was originally posted by Eliezer in 2003; when I ran across a mention of it somewhere, I realized that a lot of people probably haven't seen this old classic yet.

(For people not familiar with the concept, "critical failure tables" are found in some tabletop role-playing games. If you are trying to do something and fail really badly (a "critical failure"), you may be required to consult a table that contains a list of catastrophic outcomes that your screw-up might cause, rolling some dice to determine which one of them you did cause.)

New to LessWrong?

New Comment
2 comments, sorted by Click to highlight new comments since: Today at 8:07 AM
Specifically, the AI is an incorrigible practical joker.

Of course

The GURPSFriendlyAI page linked from the message, on the Wayback Machine: http://web.archive.org/web/20140907002238/http://sl4.org/wiki/GurpsFriendlyAI

The AI Critical Failure Table linked from there (renumbered and slightly expanded from the version in the message): http://web.archive.org/web/20130828205405/http://sl4.org/wiki/FriendlyAICriticalFailureTable