Hacker News Reader

Comment by naravara | Hacker News Reader

>Deontological ethics doesn't tell me to take good care of numbers in a computer, unless you adopt a set of rules that says otherwise purely in the name of being contrary for this post alone. (In which case I'd also like you to analyze the ethics of the numbers you changed simply in the act of posting your contrary reply.) Consequentialism doesn't tell me to be concerned about the value of numbers in a computer; I would have higher regard for the health and well being of an individual bacterium than a pedestrian in GTA.

It runs afoul of some formulations of Virtue ethics, where the idea is to cultivate certain proclivities and habits/patterns of thinking and behaving that make you “virtuous.”

In this case, the argument is that diminishing your capacity for empathy/compassion and gratifying violent or anti-social impulses are bad for you.

View on HN

Virtue ethics, deontological ethics and consequentialism can ultimately be shown equivalent, like three imaginary computers shown to be turing-equivalent. Here's a really rough sketch: (the idea being that you can take any theory of ethics in one group and write it in the language of any other)

To implement virtue in duty: "You have a duty to act the actions of a virtuous person." To implement consequentialism in duty: "You have a duty to make a best-effort estimation of the outcomes, and choose the best."

To implement duty in virtue: "A virtuous person is one who knows these rules and lives by them." To implement consequentialism in virtue: "A virtuous person is one who always evaluates the possible outcomes, and chooses the best."

To implement duty in consequentialism: "Define 'good outcome' such that any good outcome is one where all of these rules were followed." To implement virtue in consequentialism: "The only truly valuable thing is personal virtue. Therefore, a 'good outcome' is one where you gain as much virtue-trait-stuff as possible."

Just like it failed to deliver so far for humans, it would fail for the AI world.

All three approaches only work if one accepts them, and then goes on to implement them all the time, for all the decisions made. Not possible, hence evil in the world. Hence conservatism is correct. I have been wrong all my life. Any philosophical means to achieve a better world that assumes good humans being good all the time, or even most of the time, is doomed to fail.

>and then goes on to implement them all the time, for all the decisions made. Not possible, hence evil in the world.

This is literally the problem virtue ethics was formulated to solve. It's less about the ideal decision and more about what virtues and habits you have cultivated in yourself.

Not really. Utilitarians who try to find utilitarianism in other ethical schools can certainly do it, but it only makes sense to them as utilitarians. Deontologists generally don’t accept their formulations because their first principles are different, and that’s actually the key thing that people care about in discussions about ethics: the values that motivate them.

For most situations, all ethical schools of thought wind up with the same answer (unless you’re an Ayn Rand devotee). Those aren’t the interesting parts of ethical theory though, the places where they disagree are the margins where the actual theoretical work is being done.

Reply on news.ycombinator.com