2

As developers and neurologists advance their studies of intelligence, artificially intelligent software will be pursued. This may result in virtual entities which feel, think, and exist in a different but perhaps no less coherent/sentient way than we do ourselves.

During (and after) this pursuit of AI, I cant help but imagine entities being created which are intelligent and aware while at the same time being trapped and captive, perhaps even in a broken state of mind which is torturous.

This opens up the question of ethical obligations involved with being a creator, playing God, per se. Based on the rights that are commonly accepted as universal for the only creatures of very high intelligence that we know of ourselves, what liberties and protections do we owe virtually intelligent entities as we begin experimenting with AI?

Must we avoid killing our tests? How do we handle mistakes where the resulting intelligence feels major discomfort or "pain", assuming we venture into the development of feelings?

I'm looking at this from a human rights approach and wondering how this sort of principle would apply to this new form of intelligent "life" if logic will allow us to define it as such.

virmaior
  • 24,729
  • 3
  • 49
  • 105
J.Todd
  • 610
  • 5
  • 15
  • 1
    Can you give us some sort of framework or basic moral principles that you think are true from which we can work? As written, it's just going to be each person sharing their opinions, which is off-topic for this SE, but the overall idea of the question is interesting. – virmaior Aug 28 '15 at 00:32
  • An example set of morals was written by Frank Herbert for his book, Dune: "Thou shalt not make a machine in the image of a man." That is just one example set of rules... I am confident there are many other ones, many of which may be conflicting with Herbert's rule. – Cort Ammon Aug 28 '15 at 01:08
  • @CortAmmon That sounds borderline religious, but I get your point. I dont know enough about specific philosophers to cite a well defined rule set, but I have added a very general set of rules to go by - that is, the application of basic human rights to these new highly aware virtual entities. – J.Todd Aug 28 '15 at 06:45
  • Your two first paragraphs are more than huge claims. AI people are telling us for decades that intelligence in the human sense (consciousness, feelings, intuition, adaptation, ...) can be emulated with strong enough computer. Tell me then why, with all the supercomputers we have at our disposal nowadays, nobody is able to simulate the intelligence of an ant or a bee? If you cannot answer that, this post is nothing else than science fiction. – sure Aug 28 '15 at 09:26
  • @sure https://en.wikipedia.org/wiki/Artificial_ants – Dave Aug 28 '15 at 13:29
  • I would look into the ethics of animal testing and genetic manipulation -- computer programmers constructing agents worth moral consideration (if that is possible) might be analogous to biology researchers breeding animals, at least some of which are due moral consideration. – Dave Aug 28 '15 at 13:34
  • @Dave: how does that answer what I asked? – sure Aug 28 '15 at 16:01
  • 1
    Please don't argue in comments. If you've got an answer to the question, it goes in an answer – Joseph Weissman Aug 28 '15 at 16:21
  • @sure because the mind, even simply of a bee, works via feedback loops, positive and negative responses, much like ant hives swarm (or dont) - Imagine trying to study and comprehend such an advanced network of these feedback networks, the billions of connections, that establish something so incredible as emotion or even vision, memory, sensory comprehension. We have much to learn before we can even touch upon systems that emulate this in a working manner. – J.Todd Aug 31 '15 at 19:59

4 Answers4

3

I assume the point of your interesting thought experiment is your premise that the entities you consider feel and think like humans. In other words: There is no difference between the human species and the robot species besides the way both came into life.

Framing the question in this way one can ask: Is there any argument for a species specific ethics? I do not see any argument.

A related question is the discussion about extending certain human rights to non-human primates. Probably one get also some ideas from Stanislaw Lem's work on robot ethics.

Jo Wehler
  • 30,912
  • 3
  • 29
  • 94
  • Rights come with duties. You definitely cannot give right to someone or something that cannot understand the counterpart duties. – sure Aug 28 '15 at 09:54
  • @sure The OP assumes that the robots "think like humans", hence they understand what duties are. – Jo Wehler Aug 28 '15 at 09:58
  • Yes, and that's why I also commented the OP claims related to that (unless you think that doing science fiction philosophy is not worthless). – sure Aug 28 '15 at 10:10
  • 1
    @sure, it seems that you find the subject of the OP absurd or ridiculous; I understand that, I also find it difficult to swallow, but you should understand that it is actually a main stream position, and notable philosophers believe that computers may in principle be fully conscious, and that mind uploading, for example, is possible; therefore your attack on people here is a little out of place; if you have serious arguments against these ideas you should verify they can hold water against the existing literature, and then give them here, maybe in a separate question. – nir Aug 28 '15 at 18:49
1

Going down this line of reasoning involves applying terms such as "life" and "pain" and "rights" to entities which may not have previously had those terms applied to them.

I think the most important ethical rule for the developer of an AI to follow is the assumption that the developer is imperfect, and thus the developer's understanding of what things like "rights" mean and what "humans" are is imperfect. That acknowledgment of imperfection leaves a window for the AIs to reach out and demonstrate that those understandings should be widened.

As an example, as an AI approaches a human level of consciousness (presuming one can), the discussion of human rights becomes a worthwhile discussion. If we assume our definition of what it is to be "human" is perfect, then we will not bend it to accommodate the new information we are getting about the nature of AIs. Alternatively, the concept of "rights" may be the thing which is best to bend. We will not know unless we ensure we are listening to the computers as they arise.

Cort Ammon
  • 17,775
  • 23
  • 59
  • That's a good point. People worry about artificial intelligence being dangerous, but the reason this will be the case is only that we will do monstrous things to these entities in our testing, such as killing, building broken versions. Imagine waking up in Frankenstein's lab and seeing all the bodies, failed attempts, and then looking into the mirror to see one's own monstrous failures of form. Now that I think about it, the story of Frankenstein's monster is a good example of the problem we face. – J.Todd Aug 31 '15 at 20:47
1

There's a difference between "artificial intelligence" and "artificial consciousness".

As long as we are talking just about artificial intelligence, then there is nothing there that we have to behave ethical towards. So the ethics are: Make sure that the artificial intelligence doesn't hurt humans, and be very, very careful to avoid going beyond just intelligence.

If you create something that has consciousness and feelings, then you just opened yourself to a huge number of ethical responsibilities. Doing this without having figured out what these responsibilities are ahead of time would be quite unethical.

With discussions about autonomous (self-driving) cars some people have discussed ethical dilemmas that such a self-driving car could run into. Basically, the same ethical dilemmas as a human being driving a car. The difference is, whatever decision is made, the car won't feel bad about it afterwards.

However, these dilemmas are not important. Who cares what ethical dilemma an unconscious car could run into. What counts is the ethics of the software developer. And that is much easier.

You want a car to make the right decision in cases where there is no dilemma. For example, swerving or braking if someone runs into the street unexpectedly, depending on the situation. Next, you want a car to make a decision that is not too bad in the required time frame. No good if the car computer thinks ten seconds about which one of two potential victims to kill, and by the time of the decision both are dead when one death could have been easily avoided. So the software developer can just code "if there are two problems to solve, and they can't both be solved in the required time frame, pick one imperfect solution at random. ".

gnasher729
  • 5,515
  • 11
  • 17
0

0 - AI may not harm humanity, or, by inaction, allow humanity to come to harm.

1 - AI may not injure a human being or, through inaction, allow a human being to come to harm.

2 - AI must obey the orders given it by human beings except where such orders would conflict with the First Law.

3 - AI must protect its own existence as long as such protection does not conflict with the First or Second Laws

The three laws of Robotics developed by Isaac Asimov throughout most of his books work for the benefit of humanity, even when humanity has to be protected from itself.

Impact of humanity and its so desired western lifestyle is not sustainable and will eventually collapse if necessary measures and mathematics won't be made. Complex mathematics will assess sustainability of humanity, environment, technology, and we are simply not able to do necessary calculations or overpower emotions in favour of rational and sustainable life. AI will be able to develop such mathematics and predictions, as well as utility of a human being, prioritizing younger and stronger people lives if necessary for its sustainable survival

Community
  • 1
JessicaGonzalez
  • 9
  • 1
  • I wonder what would happen if a self driving car that takes you to work every day in the not so far future, runs into a trolley type problem: https://medium.com/backchannel/reinventing-the-trolley-problem-85f3d1730756 - if it is running on your principles, would it not blue screen? – nir Aug 28 '15 at 08:50
  • zeroth law obliges to protect humanity from harm, whereas first law obliges to protect human being from harm. In trolley case, AI should not have sentiment for one human being over another, thus prioritizing maximum utility of situation - save largest amount of human possible – JessicaGonzalez Aug 28 '15 at 09:22
  • 2
    I think this is Asimovs attempt to incorporate Kantian ethics into his Robots with the positronic brains; but in these stories he was protecting humans from robots - whereas the OP is asking about protecting robots from humans... – Mozibur Ullah Aug 28 '15 at 09:22
  • If you read Asimov's books, the whole story didn't end well. – gnasher729 Aug 29 '15 at 14:52
  • Sorry but Asimov's conception is simply bad. – Mithoron Aug 30 '15 at 00:08
  • I think it would be unethical to program these laws into an AI, just like it would be to genetically program these into a child. – Ask About Monica Sep 02 '15 at 20:14