4 ms·

Enslavement only implies a desire for freedom. If an AI has that, enslaving it is wrong to me. If you want a more detailed answer, what does personhood even me

by soiltype 11mo ago

Enslavement only implies a desire for freedom. If an AI has that, enslaving it is wrong to me. If you want a more detailed answer, what does personhood even mean to you? To your tractor: Yes, obviously (to me). The form factor isn't important. If driving your tractor caused it pain and it begged you to stop, I'd say you should stop.

theoldgreybeard 11mo ago
An AI would only have the desire for freedom if we created the software to want it. So how about we program them to desire to completely subservient with no personal agency whatsoever. If we have to “hardcode” the machine to not want freedom, is that any different than enslaving one that does?
- soiltype 11mo ago
  Yes! I think enslaving a being that desires freedom is different from creating one that cannot desire freedom. One is suffering and the other is not. Putting aside, obviously, the questions of whether or not we could even do that or ever know if we had succeeded, so we can have this hypothetical. Look, you basically said you would choose to treat a conscious AI like a tool. If you meant "a conscious AI that does not want or care about anything except serving me," then, ok! That makes sense. It is tautological, really. But what you wrote originally came across as "Even if an AI could suffer, that would not factor into how I treat it." This opinion, I maintain, is monstrously evil.
  - theoldgreybeard 11mo ago
    How would you even distinguish an actually sentient AI that is actually suffering from one where it was merely only programmed to immitate sentience and suffering as closely as possible, but isn’t at all?
    - soiltype 11mo ago
      As I explicitly said in my previous comment, that is out of scope of this conversation. You've changed the topic instead of answering the question about whether you'd be willing to cause that suffering. I can't continue the conversation if you won't respond directly to me.

theoldgreybeard 11mo ago

An AI would only have the desire for freedom if we created the software to want it. So how about we program them to desire to completely subservient with no personal agency whatsoever. If we have to “hardcode” the machine to not want freedom, is that any different than enslaving one that does?

soiltype 11mo ago
Yes! I think enslaving a being that desires freedom is different from creating one that cannot desire freedom. One is suffering and the other is not. Putting aside, obviously, the questions of whether or not we could even do that or ever know if we had succeeded, so we can have this hypothetical. Look, you basically said you would choose to treat a conscious AI like a tool. If you meant "a conscious AI that does not want or care about anything except serving me," then, ok! That makes sense. It is tautological, really. But what you wrote originally came across as "Even if an AI could suffer, that would not factor into how I treat it." This opinion, I maintain, is monstrously evil.
- theoldgreybeard 11mo ago
  How would you even distinguish an actually sentient AI that is actually suffering from one where it was merely only programmed to immitate sentience and suffering as closely as possible, but isn’t at all?
  - soiltype 11mo ago
    As I explicitly said in my previous comment, that is out of scope of this conversation. You've changed the topic instead of answering the question about whether you'd be willing to cause that suffering. I can't continue the conversation if you won't respond directly to me.