Freefall 2547
Socks, Spoons and Neural Nets
[!0.987]2014-09-08
An A.I. has to be told something is human. D.N.A. isn't good because you can't have the robots treating everything as non-human until they draw a sample. So what do you use?
A phone can talk and respond. A mannequin looks human. Not all humans can talk. You have to take a number of factors and add them up and if they go above 100 percent, the A.I. treats the target as human. A fully functional poopy head like you would add up to 250 percent.
This brings up which way you want your errors to go. If you have a robot working in meat processing, you want it to fail positive and think something is human rather than to fail negative and start processing its coworkers.
Color by George Peterson