Re:Ideology Impairs Sound Reasoning
by leereeves on Tuesday January 15, 2019 @06:32AM
(#18910397)
> So a rational robot/human will never open the door.
If the door leads from outside to inside, there's also a small chance the robot might die if it doesn't go inside (a passing car might hit it, anti-AI vandals might destroy it, etc). It's probably safer inside. So perhaps we could say that a rational robot/human will never go outside.
Interestingly, some people do exhibit this behavior (agoraphobia) and it's generally considered anything but rational.
The problem for the AI can be fixed, without a belief system, with a lower cost for danger. Infinite cost is too high because it makes probability irrelevant. A finite but very high "cost of death" would still avoid likely death without becoming paralyzed by the possibility of unlikely events.
We sometimes see people deciding that a cause is worth risking, even knowingly giving, their life for, proving that even humans don't weigh the cost of death above all other activities, so it's hard to see why a nonsentient AI should do so.
Re:Ideology Impairs Sound Reasoning
by danmaz74 on Tuesday January 15, 2019 @06:38AM
(#18910439)
My opinion is that we could think about an infinite amount of factors to consider if we had to do a completely rational analysis for each choice we make. In practice, we can't, and I think nobody could.
If the door leads from outside to inside, there's also a small chance the robot might die if it doesn't go inside (a passing car might hit it, anti-AI vandals might destroy it, etc). It's probably safer inside. So perhaps we could say that a rational robot/human will never go outside.
Interestingly, some people do exhibit this behavior (agoraphobia) and it's generally considered anything but rational.
The problem for the AI can be fixed, without a belief system, with a lower cost for danger. Infinite cost is too high because it makes probability irrelevant. A finite but very high "cost of death" would still avoid likely death without becoming paralyzed by the possibility of unlikely events.
We sometimes see people deciding that a cause is worth risking, even knowingly giving, their life for, proving that even humans don't weigh the cost of death above all other activities, so it's hard to see why a nonsentient AI should do so.