It was just a matter of time before somebody started working on something like this. Amusement aside, I’m impressed that Prof. Bringsjord managed to make a magazine as serious as Scientific American with this topic. In order to make a “classically evil” AGI, as opposed to a merely “indifferently evil” AGI, I guess you would face some similar issues to the creation of ethical AGI — formalising the concept of maximal evilness is probably pretty hard.
-
Archives
- December 2011
- November 2011
- June 2011
- May 2011
- January 2011
- December 2010
- August 2010
- July 2010
- June 2010
- April 2010
- March 2010
- January 2010
- December 2009
- November 2009
- October 2009
- September 2009
- August 2009
- June 2009
- May 2009
- April 2009
- February 2009
- November 2008
- September 2008
- August 2008
- July 2008
- June 2008
- May 2008
- April 2008
- March 2008
- February 2008
- January 2008
- December 2007
- November 2007
- October 2007
- September 2007
- May 2007
- April 2007
- January 2007
- December 2006
- November 2006
- September 2006
-
Meta
Flip the sign on the utility function.
If that works, then you’re just a few bits away from accidental maximal benevolence. Being robustly evil would require a more careful design strategy…
But wait, that would take away the opportunity for a heroic stunt at flipping a bit back! Nasty platonic bits, they are.
Such agents are supposed to back-up their utility functions – to prevent such a catastrophe.
This purportedly mirror image of “friendliness” is just as idiotic.
Bringsjord isn’t really much of a scientist more of an attention hungry media whore, and you see with him Barry Smith a “true believer” in absolute fixed ontologies and look at the comments on the paper, as daft as daft can be!
You are in good compagny Singularitarians…
Real concerns about “evilness” look more like this.