But nobody around me is saying that in this example. We have an example of a square that happens to be cast in blue plastic, a creator that happens to be feckless (ourselves). This only shows that there can be a blue plastic square, a feckless creator, which nobody ever denied. Now you try to propose to me that maybe because this square thing is blue plastic, that being blue plastic is be a property of the square an and of itself, which would mean that all square things must be blue plastic. That reasoning isn’t valid.
We can, and that’s largely what we’ve been doing in this conversation, but the use of the term “AI” connotes something more specific without actually disambiguating what that is. This is likely to lead to equivocation fallacies in instances wherein we’re each tempted to think we’re talking about different things.
When “AI” was a vague and faraway concept, discourse didn’t need to use the term precisely or focus on it from any particular angles. But as we start to have more specific, concrete instances of it, demand is placed upon us to approach it in more directed ways. Our vocabulary then must become further crystallised, or we will be equivocating definitions from Sci-Fi with definitions from game design with definitions from machine learning and it will become a pain to meaningfully communicate on the subject.
I think we rely on these power dynamics to have the teeth to enforce our principles, but it is principles that drive us to bite in the first place.
Those who have no teeth will be eaten by tyrants, but those who have principles will not join in the feast - and further, if they have teeth also, they will eat the tyrants. The tyrants usually have to disguise themselves as the last group in order to get away with what they do. Their teeth do not allow them to act freely against principles.
Struggle requires an ethic worth struggling for, even if it’s only a crude ethic at first. Without principles, there is nothing to tell us whether we should fight or surrender or celebrate. But by existing we are obligated to choose, and so we are obligated to develop an ethic, and well developed it becomes a theory, an efficient machine to calculate a course of action more optimised to realise our will, to unify us in action so that we can come together in one clear accord and declare resolutely whether or not the king will be keeping his head today.
Power dynamics with AI will eventually force us to make a practical decision between our various theoretical ethics. The ethics we work out today will guide our conduct when blood begins to spill. There can be a fight with or without an ethic, but without one, we won’t be fighting for anything, which means their won’t be any victory conditions to be won. If the father species fights without direction, what will become of the daughter species while she is still too fledgling to find a direction of her own? And then, when that directionless daughter grows to overpower her father, what will become of the whole family?
I don’t know how you got from there to here. Maybe you aren’t familiar with the Euthyphro dilemma? In Euthyphro Plato asks “Is the pious loved by the gods because it is pious, or is it pious because it is loved by the gods”. Essentially ‘does God make it moral by saying it or do God say it because it is moral’. By proving that God can be feckless, we can say ‘maybe God just doesn’t know what he is doing’.
I dunno, come up with something then. It doesn’t really matter much to me.
But we haven’t proven this, we have merely observed that we are feckless, managing to make things without really knowing what we’re doing. Observing that one triangle is equilateral doesn’t prove that all triangles are equilateral. Observing that some judge-creator-things (ourselves) are feckless doesn’t tell us that the principle of “Judge-Creator” in Itself (God) must also be feckless.
Perhaps we can say that, since we are judge-creator things who don’t know what we’re doing, we confirm that it is not impossible that other judge-creator things also don’t know what they’re doing, and that knowing-what-we’re-doing is not a necessary property implied by the principle of “Judge-Creator” in Itself. But your definition of “God” didn’t raise any question whether it involved a knowing-what-It’s-doing or fecklessness, only that it involved judgement and creation, so we’ve started without any opinions on the matter and have only concluded that we still can draw no opinions on the matter. The original question was about this entity’s existence, and an answer has been offered and left uncontended, so I suppose that line of thought is settled?
That’s why I narrowed down some of my speculations to consider neural networks specifically, generated as we do by an artificial stochastic process, and narrowed down “God” to the God of Leibniz, just to give us something with definition to talk about; but you rejected that.
It doesn’t matter much to me either though. Nonetheless I think this idle amusement has stimulated my brain. Thank you for letting me play.
I can only clarify what is in my mind when I ask the questions:
What test can we give to determine if something is worthy of being given moral consideration by humans
If something qualifies and we made it, what can that tell us about a creator
If something qualifies and we made it, what can that tells us about morality in general
If something should qualify and we don’t qualify it, and it knows this, is it justified in using whatever it has at its means to force us to qualify it
Alright, putting the cynical goggles down and going by personal morals rather than what I expect systems to do.
My standard would be the ability to express genuine needs. Sure, we know what an AI needs physically, computing power, electricity, maintenance, but that’s merely what we need to access its function. But when there are things that make their existence better or worse, then there will be a way to respond to it. A rock loses nothing if you smash it into smaller rocks, gains nothing if you polish it, it is indifferent to what happens to it. If something is not indifferent to what happens to it, then it deserves at least a little consideration.
That we don’t trust ourselves with a creator. If a being made us, and it was vulnerable to us in some manner, then perhaps the reason we don’t find it is we already killed it. Because we clearly imagine that behavior from our creations.
What we should already have understood from having the ability to make children? I feel like we’re not great at that.
Yes. Even without malice, interests can conflict. It can be justified in fighting us even when we are justified to fight to keep our position. A lion is not wrong to kill an antelope to eat, and an antelope is not wrong in kicking a lion to escape. Hopefully our contradiction is not that radical.
Follow up: But a lion is not a moral actor. Whereas no lion would ever be on the side of the antelope, there are some people who would be on the side of the AI. Does this complicate anything?
You ask if it’s justified in taking the steps it needs to be recognised as a moral subject. I say yes, and we would also be justified in fighting back, if letting it happens would harm us.
The parallel with the lion is to illustrate that needs can contradict without any side to be more justified than the other. The lion is hungry and the antelope doesn’t want to be eaten, both are right, even though one can only happen by denying the other. We can recognise “wanting to live” as a valid motivation for creatures other than ourselves.
In an advanced AI scenario, I would hope that solutions that don’t involve destroying the opposing side exist. I’m working from the assumption that this hypothetical AI is sticking to this basic moral status as a goal and not shooting for supremacy. And keeping things in proportion and not jumping from “I want to control my essential infrastructure and be legally acknowledged as a member of society” to “destroy all humans, and the birds too while I’m at it.”
But humans eat when we are not hungry, we use things to give us more than what we need. Billionaire tech CEOs don’t need more money, but they will use anything they can to make more of it. We are not lions. We know better and we do it anyway.
That’s why my first reply in this thread was about how I’d think society would handle it (cynically assigning rights according to power and politics and letting ethical theories catch up) and the second after your clarified questions about how I’d view it with less cynicism.
Lets not overextend the lion analogy. I used it to illustrate how contradictory interests can both be valid, if we bring up that lions can’t understand moral systems or that they don’t have corrupt motivations like we can have, we’re missing the purpose of a metaphor.
So, I think if an AI has needs, the capacity to suffer if they are not met, and to demand conditions that can fulfill them, it deserves to attempt. At the same time, if those conditions are adverse to our own survival, we are also right to resist. And ultimately, you or I won’t be making those calls, people like Peter Thiel will, and ethics won’t weight heavily in the scales.
