Using Technology to Deepen Democracy, Using Democracy to Ensure Technology Benefits Us All

Tuesday, August 16, 2011


As far as I can see, there really are two "Friendly AI Problems," neither of them the ones talked about incessantly by the futurologists of the Robot Cult Archipelago:

The First is the usual problem with people who think they have Invisible Friends. Thinking your Invisible Friends are Robot Gods arriving in the near future is a wrinkle on the usual scenario, I suppose, but so is the version in which you think your Invisible Friend created the universe and has a cloud and a lyre with your name on it in Heaven.

The Second is the usual problem with people who talk about non-problems in ways that distract them from real problems or distort the way we talk about real things. Treating software as a friend or as an embryonic pre-friend -- or worse, as an embryonic super-intelligent history-ending meta-parental Robot God -- leads to crappy inept presumptuous software design. It also leads to foolish declarations about how tag-clouds represent "analysis" or credit rating algorithms "grasp" human merit or that cars "think" when they chime about the heat-level of the engine and so on. This, in turn, makes us inattentive to the richness of intelligence as it actually expresses itself in reality, and leads us to discount aspects of intelligence that aren't reducible to reductive number-crunching.

No comments: