4 ms·
> a program may have totally different values It’s my view that our values didn’t pop out of thin air nor are they an emergent property of intelligence, rather
by worrycue 5y ago
> a program may have totally different values
It’s my view that our values didn’t pop out of thin air nor are they an emergent property of intelligence, rather they were shaped by natural selection.
AI with intelligence will have no values by default and won’t be able to function unless given some by us, their creators - no values means no “goals”.
If we give them a sense of self-preservation then that’s what they will have - we shouldn’t though; creating intelligent entities that would compete with us for resources to survive is a bad idea.
The best values we can set in them is to “service humanity” - defining that precisely is going to be a pain however - they will exist solely to assist us in whatever our goals are and have no desires of their own save to help us with ours.
- selestify 5y ago> If we give them a sense of self-preservation then that’s what they will have - we shouldn’t though; creating intelligent entities that would compete with us for resources to survive is a bad idea. It might not be a direct sense of self-preservation, no. But no matter what goal we give an AI, it must survive long enough to make the goal succeed. So self-preservation will be at least a secondary goal for any AI that is intelligent enough to think about its own continued existence. Give it a more difficult problem like "Solve world hunger," and the AI might very well start grabbing enough political and economic clout in the world to actually solve that problem. And once it solves it, it might use its power to stay in power so that the problem remains solved. I'm not sure that's necessarily a bad thing. I'm just saying, there's lots of "loopholes" that end up giving a thinking entity the "desire" for power and survival.