[ExI] Robust and beneficial AI

[William Flynn Wallace]

​I have no idea because I can't out-think something a million or a billion
times smarter than me.

 John K Clark​


​Then what do you do when the AI comes up with a solution you don't
understand?  Implement it?  What if it doesn't work?  Nothing is perfect.

What then?  I think humans have to understand what the AI is doing even if
we can't do it as fast or hold that many variables in our heads.​
​Or Murphy's Law can implement itself and we have big disasters because of
AI errors.​

Humans have to be the metacognition section.  Someone has to decide the
Oracle is wrong, or what new variables it needs to consider, and so on.
GIGO.

bill w
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20150118/2b9007ff/attachment.html>