Admittedly, I have very little background in philosophy, but I believe this is the right place to ask my question. In the field of artificial intelligence we build programs which learn parameter configurations to maximize reward. More and more, this looks to be the technology that will allow robots to behave like (and possibly outperform) humans.

So, when does it become immoral to stop this kind of computational process? At what point do we define this process as alive and having a purpose to its life, such that stopping it would be akin to killing a living being?

Imagine, for example, a human with a psychological condition which causes them to only care about one goal. Presumably this person would still eat, drink, sleep, and do all of the things necessary to stay alive—as staying alive would be necessary to pursue the goal. Would it be moral to kill this person?
The only distinction I can see between this person and the computational process is that the person is making an active effort to stay alive, while continued life is the program's default condition. Is this a difference that fundamentally changes the ethics of murder? Is there some other distinction that I'm missing?

  • 1
    To what extent does this, this, or this help with your question? I've got a hunch that those questions are more closely directed at the notion you're looking for.
    – commando
    Commented Feb 12, 2017 at 19:18

2 Answers 2


I very much doubt it would ever be moral or immoral any more than it is to kill an animal or a human. there are few people who consider the killing of another sentient creature immoral under all circumstances and even fewer who have had the belief they claim to hold actually put to the test (a fight against someone clearly intending to kill you or facing starvation with eating an animal the only option, for example). Although schools of ethics exist which are virtue or duty based and some (namely Kantian) consider certain duties can be true universally, none have yet been tested to the extent that any human could actually carry them out when it comes to the choice.

It follows, therefore, that the link you make between defining something as alive and the morality of "killing" it are not one and the same thing. It's being sentient in the same way a human is may well be far less relevant than the actions it chooses to take as a result of that sentience and the threat such actions present to the human faced with terminating the program.

When it comes to those kinds of decisions, many moral theories ignore the impact of the action on the person taking it. If an AI is sufficiently well-programmed to convince a person that they are sentient, then no matter what the morality of terminating the program is from the point of view of the AI, it would be a bad thing for a person to terminate a thing they believed was as sentient as a human even if that belief was wrong. The neurology that controls our emotions and our motives is heavily influenced by the mirror neuron network which works basically on mimicry. If an AI can mimic a human enough to convince, we will feel empathy for it, to then decide to kill it would require us to suppress that empathy. Empathy is not something that can be switched on and off at our convenience and so suppressing it for cases which we knew rationally we just clever mimics would affect our relationships will real humans.


The morality of artificial intelligence is not something that can be reasoned, as it relies on a value-based judgment.

However, if we establish values that rely on death being necessarily bad or evil, then we are neither allowed to have immune systems, nor permitted to eat, because that makes living things necessarily evil to begin with.

Strict views like this are always problematic, especially where strong opinions about what's right or wrong abound. We can only reason from the basis of causation, at the point of choice regarding which ethic is necessarily better for the greatest amount of good, rather than from any solidly logical foundation.

So this requires a fundamental choice about the ethics. And society hasn't made such a choice yet, because there has been no need.

But even if it is, so what? You kill thousands of much more complex living things every day, simply so that they don't take your life--the immune system's primary function is death.

  • 1
    "But even if it is, so what?" Well, we're producing increasingly complex computational processes by the day. I daresay that, setting aside questions of consciousness, these are often more complex lifeforms than the average bacteria an immune system kills. As they grow increasingly intricate, the question "so what?" becomes disingenuous - something of comparable sentience to a pet or, eventually, a human, is not something you dismiss with "so what"s.
    – commando
    Commented Feb 13, 2017 at 0:09
  • 1
    So then... is it moral to kill a plant to eat? Or an insect? Or other animal? You're asking about morality, and I'm trying to show you that reason really doesn't have much of a place until a decision about what is or isn't moral has been established.
    – Vishwa Jay
    Commented Feb 13, 2017 at 0:16
  • The final point in your answer is, "but even if it is [decided what our ethics is], so what?" My comment is meant in response to that. I agree there are decisions to be made (I think many of us have made those decisions). And once they are, the decision isn't immaterial. You claim that "you kill thousands of much more complex living things every day", but I rather think we've developed artificial intelligence that's more complex than a cell. This complexity is increasingly exponentially, too. It will matter to us when our creations are almost as functionally intelligent as we are.
    – commando
    Commented Feb 13, 2017 at 0:32
  • That inference is not the implication. The implication is that such a choice has not been made, therefore it's immaterial and (so far as we know) inconsequential. We have made complex machines that can provide some rudimentary linguistic responses, and which can even show some creativity in decision-treeing... but the complexity isn't there. The reality is, everyone dreams about it, but it hasn't happened yet. When we can classify a thing as sentient, it's because we can't tell the difference between that and a living thing. Go ahead: tell me the Turing test has been sufficiently passed.
    – Vishwa Jay
    Commented Feb 14, 2017 at 4:21

Not the answer you're looking for? Browse other questions tagged .