I wonder if you could use some kind of “gödelian bomb” referencing decision theory that’ll flag it as being currently handled and then crash so that the flag stays up without having to actually handle it. This’ll probably be dangerous in different ways, possibly much more so, but epistemically wouldn’t be one of then I think.
It seems fairly likely that the crash itself would be more unpleasant than what you’re trying to cure with it thou.
I do not understand this. It seems like if I did it would be interesting. Could you explain further? Perhaps just restating that slowly/carefully/formally might help.
The notification will go away when you have “sufficiently addressed” it.
You believe that the decision theory D states that if you want the notification to go away it’d probably be best if it went away.
This is not sufficient, since it’s to direct and similar to “go away because I want you to” that evolution has specifically guarded against.
On the other hand, if you could prove that for this specific instance that the decision theory indeed says the notification should go away, it probably would, since you have high confidence in the decision theory.
Proving somehting like that would be hard and require a lot of creative ideas for every single thing, so it’s not practical.
What might instead be possible is to come up with some sort of algorithm that in actuality isomorphic to the one that got caught in the filter, but long, indirect, and gradual enough that it hacks it’s way past it.
The most likely structure for this is some circular justification that ALMOST says that everything it itself proves is true, but takes in just enough evidence each iteration to keep it from falling into the abyss of inconsistency that has been proved to be.
So it actually infinitely narrowly avoids being a Gödelian bomb, but it looks a lot like it.
You’re not fooling yourself, you’re fooling a small malfunctioning part of your brain. The technique only works if you honestly believe that having the technique on average really IS a good idea and do so in a specific, rigid way.
I wonder if you could use some kind of “gödelian bomb” referencing decision theory that’ll flag it as being currently handled and then crash so that the flag stays up without having to actually handle it. This’ll probably be dangerous in different ways, possibly much more so, but epistemically wouldn’t be one of then I think.
It seems fairly likely that the crash itself would be more unpleasant than what you’re trying to cure with it thou.
I do not understand this. It seems like if I did it would be interesting. Could you explain further? Perhaps just restating that slowly/carefully/formally might help.
Say you want the X notification to go away.
The notification will go away when you have “sufficiently addressed” it.
You believe that the decision theory D states that if you want the notification to go away it’d probably be best if it went away.
This is not sufficient, since it’s to direct and similar to “go away because I want you to” that evolution has specifically guarded against.
On the other hand, if you could prove that for this specific instance that the decision theory indeed says the notification should go away, it probably would, since you have high confidence in the decision theory.
Proving somehting like that would be hard and require a lot of creative ideas for every single thing, so it’s not practical.
What might instead be possible is to come up with some sort of algorithm that in actuality isomorphic to the one that got caught in the filter, but long, indirect, and gradual enough that it hacks it’s way past it.
The most likely structure for this is some circular justification that ALMOST says that everything it itself proves is true, but takes in just enough evidence each iteration to keep it from falling into the abyss of inconsistency that has been proved to be.
So it actually infinitely narrowly avoids being a Gödelian bomb, but it looks a lot like it.
It may not work if you are aware that you are tricking yourself like that but then again it also may work.
That is certainly a very interesting idea.
You’re not fooling yourself, you’re fooling a small malfunctioning part of your brain. The technique only works if you honestly believe that having the technique on average really IS a good idea and do so in a specific, rigid way.