The Lebowski theorem: No superintelligent AI is going to bother with a task that is harder than hacking its reward function
-
-
Which is preferable in your opinion : perfect bliss or perfect detachment ?
-
Detachment. Cookies are just a tool to make yourself eat your vegetables.
- 3 more replies
New conversation -
-
-
From my perspective, a pre-knowledge of this would stop me tweaking it in the first place. It sounds like herione: most people don't use this substance even with the option available.
-
Interesting but are we really aware that this option exists ? Also are we rationnal and are we able to see this option as the AI could see the hack of its reward function ? I mean, even by reading what you wrote, I still don't wan't some heroin. But would the AI think the same ?
- 1 more reply
New conversation -
-
-
But there are lots of humans who wouldn't hack their own reward functions even if it was very easy to do so. Surely there are at least a few points in the space of all superintelligent minds that would behave similarly.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Utility functions assume the prospective value of things means their expected hedonic reward. As a human being, though, you've learned to ascribe prospective value to things you know won't make you happier. So you don't have a utility function.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.