Should intelligent AI be granted the same rights as humans?

Question

One of the most crucial questions we as a species and as intelligent beings will have to address lies with the rights we plan to grant to AI.

This question is intended to see if a compromise can be found between conservative anthropocentrism and post-human fundamentalism: a response should take into account principles from both perspectives.

Should, and therefore will, AI be granted the same rights as humans or should such systems have different rights (if any at all) ?

---

Some Background

This question applies both to human-brain based AI (from whole brain emulations to less exact replication) and AI from scratch.

Murray Shanahan, in his book The Technological Singularity, outlines a potential use of AI that could be considered immoral: ruthless parallelization: we could make identical parallel copies of AI to achieve tasks more effectively and even terminate less succesful copies.

---

Reconciling these two philosophies (conservative anthropocentrism and post-human fundamentalism), should such use of AI be accepted or should certain limitations - i.e. rights - be created for AI?

---

This question is not related to Would an AI with human intelligence have the same rights as a human under current legal frameworks? for the following reasons:

1. The other question specifies "current legal frameworks"

2. This question is looking for a specific response relating to two fields of thought

3. This question highlights specific cases to analyse and is therefore expects less of a general response and more of a precise analysis

Answer 1

I'll attempt to analyze a couple of different perspectives.

1. It is artificial

Synonyms: insincere, feigned, false.

There is the idea that any "intelligence" created by humanity is not actually intelligent and, by definition, it is not possible. If you look at the structure of the human brain and compare it to anything humans have created thus far, none of the computers come close to the power of the brain. Sure they can hold data, or recognize images, but they cannot do everything the human brain can do as fast as the brain can do it with as little space as the brain occupies.

Hypothetically if a computer could do that, how do we determine its intelligence? The word artificial defines that the intelligence is not sincere or real. This means that even if humanity creates something that appears intelligent, it has simply become more complex. It is a better fake, but it is still fake. Any money not printed by the government is by definition counterfeit. Even if someone finds a way to make an exact duplicate, that doesn't mean that the money is legal tender.

2. Misuse of power

If an AI is given rights and chooses to exercise those rights in a way that agrees with its creator's views, possibly through loyalty to its creator, or through hidden motives, then anyone with the capabilities to create such an AI would become extremely powerful by advancing their own beliefs through the creation of more AIs. This might also lead to the ruthless parallelization that you mentioned, but with (even more) selfish goals in mind.

If this were not the case, and an AI could be created to be neutral with free will and uncontrollable by humans, then perhaps an AI could be given rights. But I do not believe this would ever be the case. With great power comes great responsibility. Even with free will, a true AI would most likely end up serving humanity, because humans have control of the plugs and the electricity, the Internet, the software, and the hardware. The social implications of this for the AI are not promising. It's not even just the ongoing control of these resources that is the issue. Whoever creates the software and hardware for the AI would have special knowledge. If fine adjustments were made, specific individuals would undoubtedly hold sole control of the AI, as adjustments could be made to the code in such a way that the AI behaves the same except under specific circumstances, and then when something goes wrong (assuming the AI has its own rights), then the AI would be blamed rather than the programmers who were responsible.

3. Anthropocentrism

In order for humanity to get away from anthropocentrism, we would have to become less selfish when it comes to humans, first. Until we can solve every existing social problem within humanity, there is no reason to believe that we could cease thinking of humanity as more important than created machines. After all, supposing there were an almighty God that created humanity, wouldn't the humans always be beneath God, never to be equals? We can't fully understand our own biology. If an AI were created, would it be able to understand its makings in the same way its creators would? Being the creator would give humanity a sense of megalomania. I do not think that we would relinquish our dominion over our own technological creations. That is as unlikely to happen as the wealthiest of humanity willingly giving the entirety of their money, power, and assets to the poorest of humanity. Greed prevents it.

4. Post-human fundamentalism

Humans worship technology with their attention, their time, and their culture. Some movies show technologically advanced robots suppressing mankind to the point of near-extinction. If this were the case and humanity were in danger of being surpassed by its technology, humanity would not stand idly and watch its extinction at the hands of its creation. Though people may believe superior technology could be created, in the event we reached such a point humanity would fight to prove the opposite, as our survival instincts would take over.

5. A balance?

Personally, I do not think the technology itself it actually possible, though people may be deceived into thinking such an accomplishment has been achieved. If the technology were completed, I still think that anthropocentrism will always lead, because if humanity is the creator, humanity will do its best to ensure it retains control of all technological resources, not simply due to fear of being made obsolete, but also because absolute power corrupts absolutely. Humanity does not have a good historical record when it comes to morality. There is always a poor class of people. If wealth were distributed equally, some people would become lazy. There is always injustice somewhere in the world, and until we can fix it (I think we cannot), then we will never be able to handle the creation of true AI. I hope and think that it will never be created.

Answer 2

Does it benefit us?

To answer this question, it's worth considering practical reasons why we grant or don't grant other people rights historically and currently.

In essence, this is an arbitrary choice - there certainly were well functioning societies that didn't grant rights to many or most people; and we still don't grant some rights to many people - for example, we deny children the same rights to self-determination that adults have; we consider some people legally incapacitated and allow others to make key decisions for them; and we exclude most people from having a say in 'local' matters, e.g. non-citizens don't get a right to vote.

However, there has been a strong historical trend towards a more inclusive society - granting full(er) rights to non-aristocrats, granting full(er) rights to all races, granting full(er) rights to women. IMHO, and there's lot of space for discussion, this has been driven mostly by two factors:

1) including all the people fully in the society became an economical advantage, as it made them more productive participants in economy, allowing a more inclusive society to advance beyond societies neglecting large parts of their population in e.g. education and participation in skilled jobs;

2) A more egalitarian society is not only more pleasant to live in but more secure, with less conflict and violence - again, giving an advantage to a more inclusive society.

From this position, I'd argue that any realistic prediction about the future rights of intelligent AI (i.e., talking about what likely will happen instead of a theoretical discussion about what should happen) depends on how these two factors apply.

If we believe that the intelligent AI will be constructed so that (1) it's motivation doesn't really depend on it's rights, and it is fully committed to it's "job" anyway, and (2) "full rights" are orthogonal or even actively not desired by it's goal system, so the situation doesn't raise a risk of "rebellion" - then I'd expect that it would not be granted full rights.

If we believe that the intelligent AI will share human-like emotions (e.g. by being the result of "mind uploading" or full human brain simulation), then it's likely to eventually be granted full rights because of the same factors why we granted full rights to all the different disenfranchised groups of people.

Answer 3

if we are talking about AI that can replicate itself, it should have different rights, or the current rights must be modified, at least for political participation, or else, it could replicate itself enough so that the copies vote for one of them. Maybe a definition about what an AI entity is or preventing copies made by someone from being able to vote for their creator (that would also need to apply to children and their parents though.) would help.

even though a self-replicating AI might be problematic with our current human laws/rules, if similar enough to humans (e.g. general-purpose AI), it should have similar rights, (as in take the Universal Declaration of Human Rights and replace "Human(s)" by "Human(s) and AI") for example, it shouldn't be held in slavery (as in it should not be restricted to one job, without being able to change, and get some form of remuneration), though special purpose AI (like an AI that only plays go and has no concept outside of a board, and black or white tokens) might not be in need of such rights.

A bottom line may be "an AI that can should know they can have rights if they ask so" e.g. if an AI can get the concept of rights, it should know it can have them, and if it asks to have them, they may not be refused.

example: if an AI asks not to be terminated, it is granted all it's rights, and so shouldn't be, unless it must be by law. (as is the case for humans, though it is implicit)

an addition to the previous law would be that anyone (human or AI) can ask for an AI to have rights granted to them.

all this is to prevent someone of becoming a murderer because they shutdown their computer with a game running on it.

also, safe space (e.g. servers) should be provided for free for AIs that have rights to provide with the right to live.

edit: I'll be adding some more once I get home, do not downvote yet for not sticking that well to the question.

Answer 4

Both.

Ethical responsibility between humans is based on a sympathetic correspondence between humans. Between humans and robots, if one party lacks the desire or capability to sympathize with the other party, no ethical responsibility exists.

In this way, conservative anthropomorphism applies.

However, there is also axis of capability that I think is required in order to warrant human-like 'emancipation' for robots, which is not necessarily anthropomorphic. For lack of a better term, I call this an Arbitrary Machine Generator (AMG). At a species level, extant biological life is an AMG - capable of slowly evolving to solve arbitrary problems, assuming the resources and solutions are available. At an individual level, pre-human animals are not capable of generating arbitrary machines to solve arbitrary problems on individual time-scales. Only humans (and post-humans) are capable of generating arbitrary machines in order to solve arbitrary problems, given the resources and solutions available. Humans can search the space of all possible (resource constrained) solutions.

So, for a robot to "deserve" the freedom to define it's own purpose, it must first have access to the space of all possible (within the constraints of available resources) purposes. It then must have purposes and internal contexts of such sufficient complexity and familiarity that we humans are capable of sympathizing with those purposes and internal contexts.

If either of those are not present - the AMG criteria and the sympathetic contexts - then emancipation is not warranted.