Hacker News Reader

Comment by cellis | Hacker News Reader

Who here has watched as much Black Mirror/Westworld as I have and know how this ends? My question is, if a robot is engineered well enough to replicate the "real" thing, then when you test on it, aren't you just shifting the torture from "living" to another thing yet to be granted the rights of the living? If you answer "no", then you have to admit that the model is incomplete and it does not in fact feel pain and is thus in an inadequate replica. Also agree with other comments that this is and will remain computationally infeasible at least until production quantum computers arrive.

View on HN

You can ask the reverse question - if the Robot version of a mouse is sufficiently similar to a mouse, maybe the mouse also doesn't actually feel pain, and is really a kind of biological robot.

Maybe so are you.

Dualism holds that living entities are special, and can never be replicated artificially. So your question can actually be turned into dualism vs monism (which is not a question with a settled answer).

This comment doesn't make any sense to me. Why would the existence of a robot copy imply a lack of pain? How can you argue that whether humans feel pain is an unsolved question? Shall we experiment on you instead?

Many philosophers are confused. Free will is a solved problem (we have free will regardless of the metaphysical premise, unless you're using a weird definition of free will, and even I can prove it for any given metaphysical premise), yet they're still working on it.

If somebody links "free will" to "determinism", they're very confused about what free will is.

There are a lot of philosophers doing genuinely useful work, but so much of the field is pre-occupied with trivial (and non-trivial) variations of the "if a tree falls in a forest, does it make a sound?" koan.

I must be quite confused then, but how is free will not linked to determinism?

There's a proper proof, which is too long to fit in this margin (i.e. I'm too lazy to find a link), but I can do a much shorter one:

1. Non-determinism is equivalent to determinism plus extra, random input.

2. The addition of random noise does not bestow free will upon something that didn't have it without the random noise.

3. 1 and 2 imply no non-determinstic systems can possess free will if deterministic systems cannot possess free will.

4. 3 implies free will is independent of determinism.

See my reply here: https://news.ycombinator.com/item?id=23727782

This would run into the "here is a hand", or rather "here is a fist" objection.[1]

Pain is the ground reality that is to be explained. If you manage to explain away pain rather than explain it, you haven't actually answered the question at all. If your answer at the end is "pain is not real", this doesn't prove that pain is not real, it proves that your premises were wrong.

[1] https://en.wikipedia.org/wiki/Here_is_one_hand

If a computer prints on the screen "I'm in pain", is it actually in pain?

If you tell me "I'm in pain", are you actually in pain?

If you are sedated with a memory blocker, and are then hurt, but you don't remember it afterward, were you actually in pain? (This isn't an academic question - many medical procedures simply paralyze the patient and add a memory blocker, but don't provide much analgesia. The patient can not express pain, and does not remember it. So does the pain they felt matter?)

You want an explanation of the wrong "ground reality", you are discussing the wrong question, it's not "is pain" real.

It's all about the subjective experience of pain, and can a robot share in that? Can a person in a coma share in that? Can an animal?

> If you tell me "I'm in pain", are you actually in pain?

> If a computer prints on the screen "I'm in pain", is it actually in pain?

Maybe. If you SMS me that you are in pain, are you in pain? "I'm in pain" is evidence of someone being in pain. But being in pain, the referent, is just being in pain. There's no additional indirection.

> If you are sedated with a memory blocker, and are then hurt, but you don't remember it afterward, were you actually in pain?

Yes.

> You want an explanation of the wrong "ground reality", you are discussing the wrong question, it's not "is pain" real.

No, this is an issue that was actually brought up.

> > maybe [...] you [...] [don't] actually feel pain

All reason is grounded in experience. If I know anything, it's the things that I experience directly. Pain is probably the most direct experience there is.

edit:

> It's all about the subjective experience of pain, and can a robot share in that? Can a person in a coma share in that? Can an animal?

Yes. No. Yes.

These are not difficult questions to answer... You're phrasing it like there's some kind of philosophical difficulty here, but I don't think there is. If you define pain as an aversive sensory experience, ie. a sensory experience that makes you wish it would stop (this definition covers disgust and discomfort as well, and I think that's valid), then it seems clear that anything that can form goals (and yes, a memory blocked sedated person can form goals, they just can't act on them) can experience pain. I think the objection to computers is just that we rely too much on our empathy to determine when something "is in pain". I think an NPC in a game that moves away from a fire is experiencing pain, even though its mind is extremely simple, just like an amoeba would when it avoids a harmful chemical. It doesn't have such a thing as a qualia of pain, maybe (I'm not convinced!) - but that's hardly necessary.

I feel we must overcome our distrust of functional descriptions. Pain is a functional pattern, and not a very complex one.

Easiest way I can come up to go with it is to admit we sometimes exploit life for the greater good, and also swear to yourself that you do your best to keep it as natural, efficient and minimal as possible. Otherwise you don’t get to even grow and eat single grain of wheat do you...

I think you are missing something obvious here. Pain does not exist in any objective sense, it is a construct created by the brain in response to certain stimuli as a survival mechanism. If you were creating a simulation of a human body to use in medical experiments then surely you would not give it a brain that functioned in any meaningful sense.

I would not say that a body incapable of feeling pain is an "inadequate replica". A human or animal under anaesthetic does not feel pain - are they inadequate. You can certainly perform surgery on an unconscious person, and you can administer medications to them.

[edit] I should have added that I am totally in agreement that we are not even close to being able to simulate a human body with any degree of accuracy.

Maybe pain is a part of the treatment as any body/antibody response. It is a push notification mechanism designed in a million of years of evolution for a reason. But I agree that it might be deprecated soon.

I'm not sure these questions can be answered philosophically. After thousands of years, philosophy can't even agree on the the morality of things that already do exist.

Science fiction and moral philosophy depict "model organisms" as physical replicas of organic ones for a reason. Our moral ponderings are driven by moral feelings, and they need to trigger feelings for the reasoning to be interesting. The neocortex court only debates when limbic prosecutors press charges.

Ultimately, the task here is software that answer questions, not software that is a mouse. If software those two converge, we have a moral dilemma. Seems unlikely though.

Realistically this unlikely to be a problem in this century, however you are raising some valid points nonetheless. A lot of,if not most,capabilities animals, including humans, have are related to one thing: survival. Ability to move rocks, climb trees,or build sophisticated structures boils down to body's need to get food and preserve itself from destruction. These aspects are often overlooked when modelling AI and the whole focus is on pure logic. I sometimes think that one day we'll end up with universal AI and the first thing it will ask will be sometime like: "why should I do it?".

If you don't do it our funding will be cut and we'll have to power you down.

And that's where it gets complicated. You'd either need to create it in a way that you could inflict these negative reactions, like pain,thirst,fear,and so on.,as otherwise it would respond with " OK, go on then,not that I care anyway".

If you torture someone then use a time machine to restore him to the initial state, did you do anything bad?

You can't do this to anyone real but you can easily save and then restore the state of the simulation at any point.

Torturing someone in a simulation is not any different than imagining someone being tortured, or performing lots of arithmetic operations.

So the only logical thing is to admit that torturing/killing/hunting a species to extinction is bad not because of pain, but because of stealing computational resources belonging to someone else/some other sophisticated computational process.

> Torturing someone in a simulation is not any different than imagining someone being tortured,

Isn't it? Our imaginations are predictive, not simulative.

Sure, our imagination does not have enough computational power to be simulative now, but in principle there is nothing preventing existance of a living creature with vastly more powerful brain.

If you're assuming imagination ~= simulation for an argument, you need to state the assumption up-front. That assumption doesn't hold for humans, nor for anything reasonably imaginable for people in our culture.

> Torturing someone in a simulation is not any different than […] performing lots of arithmetic operations.

Near as we can tell, the universe itself just performs lots of arithmetic operations.

I'm going to give an example similar to:

> If you torture someone then use a time machine to restore him to the initial state, did you do anything bad?

to make a point. If you create a person, torture them, then incinerate the person to restore them to the original state, did you do anything bad? If you're privileging "the present state of the universe, excluding historical events, even when evidence of those historical events is apparent in the present state of the universe", you're taking a very niche ethical viewpoint: one where anything can be undone by killing everyone involved.

> So the only logical thing is to admit that torturing/killing/hunting a species to extinction is bad not because of pain, but because of stealing computational resources belonging to someone else/some other sophisticated computational process.

I'd call this the principle of explosion, but I don't actually see how this can even follow from what you said earlier.

The example with creating a person and then destroying it is different, because after creation it can continue to live independently from yourself. But if you create a person in your computer, that person can exist only as long as you provide the computational power for it to exist. In a way it is similar to abortion, where destroying a fetus is ok because it cannot exist on it's own, and the mother cannot be forced to provide her resources to it.

Ian M Banks explored this in the culture series. When doing speculative simulation of how another civilization will react to situation, what do you do with all those little simulated individuals when you've got the data you need. If you simulated them with enough detail, isn't turning the simulation the same as genocide?

Reply on news.ycombinator.com