Re:John Carmack on the Joe Rogan Experience [video]
by realbarack on Thursday August 29, 2019 @11:35AM
(#20830413)
It isn't hard to give an AI a goal but it is hard to do so safely. As a toy example we could design an AI who treated say, reducing carbon emissions as it's goal just as you treat eating and sleeping as yours. The issue is that the sub goals to accomplish that top-level goal might contain things we didn't account for, say destroying carbon-emitting technology and/or the people that use it.
[ Parent
| on HN ]
Re:John Carmack on the Joe Rogan Experience [video]
by etherealG on Monday September 02, 2019 @11:11AM
(#20859545)
Human's have many basic goals that are very dangerous when isolated in that way. It seems to me that nature didn't care (and of course, can't care) if it was dangerous at all when coming up with intelligence. Maybe we shouldn't either if we want to succeed with replicating it.
Worrying about some apocalypse seems counterproductive to me.