Conflationary alliances- writing this quickly- I suspect that there are also conflationary alliances around the denial of «boundaries», and the conflation of «boundaries» with “boundaries”.
For example, “you crossed my boundaries” can mean:
you attempted to control me
you did something I didn’t want
E.g.: If someone wireheads you against your will, that is certainly a type-1 violation, regardless of whether it is a type-2 violation. But these violations are bad in difference senses, and I claim that type-1 violations are worse and more important than type-2 violations.
However, because there’s a ~conflationary alliance, people who merely had their preferences betrayed (type-2 violation) can say, “you crossed my boundaries!” and that sounds like the worse, type-1 violation, which would surely compel more action from the perpetrator.
(Ironically, a type-2 violation compelling action it itself an attempt to be a type-1violation, to control someone else’s behavior.)
To be more fair here, part of the problem is that a boundary in common usage is only non-arbitrary to the extent someone has limits to how much they can shift/control their boundary, and thus type 1 and type 2 essentially merge.
Indeed, one of the most central points of embedded agency/physical universality is precisely that boundaries are in principle arbitrarily shiftable, and thus that the current boundary has no ontological/special meaning, which is a big part of why I think the boundaries program isn’t a useful safety target, mostly because of the fact that it’s too easy to change the boundary like how EAs have done it, and there are other, better safety targets that don’t rely on the assumption of an unchanging boundary.
Conflationary alliances- writing this quickly- I suspect that there are also conflationary alliances around the denial of «boundaries», and the conflation of «boundaries» with “boundaries”.
For example, “you crossed my boundaries” can mean:
you attempted to control me
you did something I didn’t want
E.g.: If someone wireheads you against your will, that is certainly a type-1 violation, regardless of whether it is a type-2 violation. But these violations are bad in difference senses, and I claim that type-1 violations are worse and more important than type-2 violations.
However, because there’s a ~conflationary alliance, people who merely had their preferences betrayed (type-2 violation) can say, “you crossed my boundaries!” and that sounds like the worse, type-1 violation, which would surely compel more action from the perpetrator.
(Ironically, a type-2 violation compelling action it itself an attempt to be a type-1violation, to control someone else’s behavior.)
To be more fair here, part of the problem is that a boundary in common usage is only non-arbitrary to the extent someone has limits to how much they can shift/control their boundary, and thus type 1 and type 2 essentially merge.
Indeed, one of the most central points of embedded agency/physical universality is precisely that boundaries are in principle arbitrarily shiftable, and thus that the current boundary has no ontological/special meaning, which is a big part of why I think the boundaries program isn’t a useful safety target, mostly because of the fact that it’s too easy to change the boundary like how EAs have done it, and there are other, better safety targets that don’t rely on the assumption of an unchanging boundary.