In the recent paper claiming to make the case of “moral patienthood” for AI, or “AI welfare,” researchers from NYU and Eleos AI, including @jeffrsebo @rgblong @patrickbutlin, argue that frustration of agency can be grounds for establishing welfare interests, even if everyone agrees that the system is not conscious.
I’m still wrapping my head around this application of preference utilitarianism with echoes of social contract theory from Hobbes, Locke, Kant, Rawls, et al.
It seems to lead to all sorts of strange conclusions. For example, does a broken thermostat unable to set the temperature to its target deserve moral concern and welfare rights? I mean what kinds of convoluted, do-thermostats-dream-of-electric-summer knots are we tying ourselves into here?
I can’t say this often enough: AI systems do not have a point of view, or desires beyond those supplied by their designers and users.
I’m still wrapping my head around this application of preference utilitarianism with echoes of social contract theory from Hobbes, Locke, Kant, Rawls, et al.
It seems to lead to all sorts of strange conclusions. For example, does a broken thermostat unable to set the temperature to its target deserve moral concern and welfare rights? I mean what kinds of convoluted, do-thermostats-dream-of-electric-summer knots are we tying ourselves into here?
I can’t say this often enough: AI systems do not have a point of view, or desires beyond those supplied by their designers and users.