3 ms·

It's okay I guess if you think you have no causal impact on if any of that happens, but otherwise you better be doing something to help, even if it would only w

by Vecr 3y ago

It's okay I guess if you think you have no causal impact on if any of that happens, but otherwise you better be doing something to help, even if it would only work in a tiny part of probability space, instead of waiting to die.

throwanem 3y ago
Why waste time posting here that you could spend weaving mosquito nets, then? Surely this can't be a global maximum in utilon space.
- Vecr 3y ago
  I'm talking about VNM utility, I'm not a utilitarian or a EA.
  - throwanem 3y ago
    VNM applied in the realm of ethics is definitionally utilitarian.
    - Vecr 3y ago
      > That your utility function may be off model from the most doctrinaire possible interpretation of the theory does not seem to me the same as saying this is not an example of a theory. Sure, I don't want to get into a definitional argument. However, any VNM utility function can be the exact opposite of anything you propose as good, just by inverting the comparison order. If you have numerical outcome scores that could be as easy as my score =: 0 - your score > (Did you mean to reply to your comment rather than mine?) It's a HN thing.
      - throwanem 3y ago
        Well, yeah, that's the fundamental failure of utilitarianism: it's so abstract as to say nothing meaningful, because it explicitly delegates the entire definition of what actually is ethical to the utility function, and nothing past that is distinct from any other ethical theory. Again I grant this varies from Bentham's definition, but Bentham has been dead a long time now. Modern utilitarianism has been defined by Singer, for better or for worse - mostly worse - and Singerism is what I describe. In any case, Bentham's version still bears the flaw; 'good' can stretch with sufficient cleverness to cover all the virtue or deontology in the world. > It's a HN thing. I doubt that; this and your prior comment are the second and third time I've seen it in eleven years of active participation here. Are you using some kind of extension or script to augment the UI? If so, I strongly suspect that is causing it.
  - Vecr 3y ago
    > VNM applied in the realm of ethics is definitionally utilitarian. I assume you are thinking deontological ethics, virtue ethics, etc vs utilitarianism, but non-benthamite utility functions exist. Actually, if they didn't I assume a lot of AI doomers would be less doomy, as quite a few of them are utilitarians.
    - throwanem 3y ago
      No, I'm thinking utilitarianism. I appreciate that the standard definition includes a term about "for the greatest number". But you're still defining an ethical utility function which you then seek to maximize. That your utility function may be off model from the most doctrinaire possible interpretation of the theory does not seem to me the same as saying this is not an example of the theory. The same could be said of effective altruism, which is anyway just the prosperity-gospel version of utilitarianism. That one should be doing utilitarianism badly does not mean one is not doing utilitarianism at all. (Did you mean to reply to your comment rather than mine?)
      - Vecr 3y ago
        > Well, yeah, that's the fundamental failure of utilitarianism: it's so abstract as to say nothing meaningful, because it explicitly delegates the entire definition of what actually is ethical to the utility function, and nothing past that is distinct from any other ethical theory. I think it's honest, as it really forces you to make a distinction between good (what should be) and real (what is). "Should" goes into the utility function, "is" goes into the world model. I'd honestly recommend giving it a shot if you have free time. > I doubt that; this and your prior comment are the second and third time I've seen it in eleven years of active participation here. It's the anti-flame war/spam thing kicking in. I'm manually responding to the "wrong post" but that's pretty much fine.
        throwanem 3y ago
        Oh, I see. If you click the timestamp on a comment, that view always has the reply form. Only in the thread view is the reply link hidden for the first few minutes after a comment is posted. I'm familiar with Hume. The is-ought distinction has value, but is orthogonal to utilitarianism, if not explicitly opposed.
dsign 3y ago
Just this. And there are things we can do. Those of us who live in democracies can influence outcomes a little. It may be too early for political action, but perhaps is not too early for writing and reflecting, and for coming up with theories about how the near future may look like, and sharing them with the world. If things turn really dire, there is always that turbid matter our advanced culture forbids us to reach out for: violence and revolutions. In other words, if it is the AIs or us, be it because the AIs are outright evil or because our economic systems make them evil, then we can always choose “us” and resort to forceful or even violent corrections.

by Vecr 3y ago

throwanem 3y ago

Why waste time posting here that you could spend weaving mosquito nets, then? Surely this can't be a global maximum in utilon space.

Vecr 3y ago
I'm talking about VNM utility, I'm not a utilitarian or a EA.
- throwanem 3y ago
  VNM applied in the realm of ethics is definitionally utilitarian.
  - Vecr 3y ago
    > That your utility function may be off model from the most doctrinaire possible interpretation of the theory does not seem to me the same as saying this is not an example of a theory. Sure, I don't want to get into a definitional argument. However, any VNM utility function can be the exact opposite of anything you propose as good, just by inverting the comparison order. If you have numerical outcome scores that could be as easy as my score =: 0 - your score > (Did you mean to reply to your comment rather than mine?) It's a HN thing.
    - throwanem 3y ago
      Well, yeah, that's the fundamental failure of utilitarianism: it's so abstract as to say nothing meaningful, because it explicitly delegates the entire definition of what actually is ethical to the utility function, and nothing past that is distinct from any other ethical theory. Again I grant this varies from Bentham's definition, but Bentham has been dead a long time now. Modern utilitarianism has been defined by Singer, for better or for worse - mostly worse - and Singerism is what I describe. In any case, Bentham's version still bears the flaw; 'good' can stretch with sufficient cleverness to cover all the virtue or deontology in the world. > It's a HN thing. I doubt that; this and your prior comment are the second and third time I've seen it in eleven years of active participation here. Are you using some kind of extension or script to augment the UI? If so, I strongly suspect that is causing it.
- Vecr 3y ago
  > VNM applied in the realm of ethics is definitionally utilitarian. I assume you are thinking deontological ethics, virtue ethics, etc vs utilitarianism, but non-benthamite utility functions exist. Actually, if they didn't I assume a lot of AI doomers would be less doomy, as quite a few of them are utilitarians.
  - throwanem 3y ago
    No, I'm thinking utilitarianism. I appreciate that the standard definition includes a term about "for the greatest number". But you're still defining an ethical utility function which you then seek to maximize. That your utility function may be off model from the most doctrinaire possible interpretation of the theory does not seem to me the same as saying this is not an example of the theory. The same could be said of effective altruism, which is anyway just the prosperity-gospel version of utilitarianism. That one should be doing utilitarianism badly does not mean one is not doing utilitarianism at all. (Did you mean to reply to your comment rather than mine?)
    - Vecr 3y ago
      > Well, yeah, that's the fundamental failure of utilitarianism: it's so abstract as to say nothing meaningful, because it explicitly delegates the entire definition of what actually is ethical to the utility function, and nothing past that is distinct from any other ethical theory. I think it's honest, as it really forces you to make a distinction between good (what should be) and real (what is). "Should" goes into the utility function, "is" goes into the world model. I'd honestly recommend giving it a shot if you have free time. > I doubt that; this and your prior comment are the second and third time I've seen it in eleven years of active participation here. It's the anti-flame war/spam thing kicking in. I'm manually responding to the "wrong post" but that's pretty much fine.
      - throwanem 3y ago
        Oh, I see. If you click the timestamp on a comment, that view always has the reply form. Only in the thread view is the reply link hidden for the first few minutes after a comment is posted. I'm familiar with Hume. The is-ought distinction has value, but is orthogonal to utilitarianism, if not explicitly opposed.

dsign 3y ago

Just this. And there are things we can do. Those of us who live in democracies can influence outcomes a little. It may be too early for political action, but perhaps is not too early for writing and reflecting, and for coming up with theories about how the near future may look like, and sharing them with the world. If things turn really dire, there is always that turbid matter our advanced culture forbids us to reach out for: violence and revolutions. In other words, if it is the AIs or us, be it because the AIs are outright evil or because our economic systems make them evil, then we can always choose “us” and resort to forceful or even violent corrections.