There is the Singularity Institute for Artificial Intelligence -- http://singinst.org/

They are tackling the problem of how to make AI "friendly" -- http://wiki.lesswrong.com/wiki/Friendly_artificial_intellige...