This paper describes existing views on ethical rules governing interactions between intelligent entities and suggests several new approaches based on a new understanding of underlying philosophical concepts, such as identity, consciousness, time and death.
In the future we will need to interact with our intelligent creations, with augmented humans and with complex systems that integrate humans and artificial entities together. The process of this interaction poses new challenges that can be answered neither by common-sense approach nor by the field of computer science. Many of these questions are have been discussed by philosophy in a slightly different context. I think it is interesting to look at how these questions can be answered in regard to future development in the field of AI.
Importance – we need to understand these issues now as it takes a long time for values in the society to change. When we would need this understanding there might be no time left already.
There is no sign of slowdown in the growth of computer processing speed. Just recently Intel announced its plans to produce terahertz processors (1). Next generation computing technologies will utilise nanotechnology (1). DNA and quantum computers are already on the horizon.
Many authors in this field (2, 3) agree that by 2020-2030 the processing power of standard computers will exceed the processing power of the human brain. A number of researchers (4, 5) believes that this will cause the Singularity, “a … point in the future, … where the pace of technological development becomes extremely rapid … because of advances in artificial intelligence, computer-human integration or other forms of intelligence amplification” (6). A number of scenarios has been proposed (7) according to the roles and positions of humans, posthumans and AIs.
In this paper I want to discuss some implications of further AI development.
If we consider only the biological brain and the mind of unmodified human to be natural (as opposed to artificial), most of the future intelligences, including ours will indeed be artificial, i.e. deliberately created through conventional programming, genetic algorithms, self-modification and other methods. It means that even the development of our future brains lies in the field of AI.
I consider it to be most likely that the differences between posthumans and AIs will blur very quickly. The main reasons for that are outlined below:
· Intelligence is emerging property of complex systems. The nature of intelligence is the same, at least for some categories of AI.
· Potential demand for human augmentation is large.
· Hybrid systems are possible, as indicated by well-known experiment with connecting snail neurons with a silicon chip (8)
· The neural structures of the human mind substrate can be reproduced, both on the level of individual neurons and neural networks.
While some reasons have been proposed why intelligence is not possible for computers (or any other Turning machines), such as quantum properties of the neurons (9), as long as these properties obey physical laws they probably can be reproduced outside of the human organism, although possible not in conventional digital computer form. On the other hand, the soul, as outlined in any religion, is improbable and impossible to discover, because the term is usually defined negatively (there is nothing certain about the soul, only what it is not and what it cannot be).
So we can assume that future intelligence entities will be possible in both “natural” form and artificial form, as well as anywhere in between. The process of uploading (10) is moving the mind from biological human brain into computer (any Turing machine). The mind will be possible to move between any Turing machines (11), or even to their static representation (data storage).
Some of the carriers for the mind that have been proposed are quite large, such as Jupiter brains and Galaxy brains (12). They will be, no doubt, quite powerful and we can expect much more complex structure than possible in the human brain.
Interactions will change, everything will change, but we need to know in advance. As the process of change will be gradual, we need to find answers for our stage.
Philosophical answers are useful, because the transition to such complex brains will be gradual and the very first stage are usual “normal” humans that need this kind of motivation and guidance provided by the philosophy. The problems are those of self, consciousness, identity, time, death, etc. I will briefly go through them.
All of these concepts have been developed through evolution. As it usually happens during the time of change, old concepts become ineffective and have to be replaced by more suitable ones. Of course, evolution does not know what is more suitable, so many different new concepts are tried and the most effective remain.
One of the main principles is the “gradualism”. We used to live in a world where most of object classes had well-defined boundaries. The ancient or medieval man had no problem identifying alive or intelligent objects/creatures. That does not mean there have been no grey areas, though. Consider the blacks. They were not considered entirely human for a long time, as well as infidels and savages. Over time, when they were becoming more integrated into the society (both internally and from other countries) and new objects were appearing in between (mulattos, new flavours of Christianity, assimilated natives), the cultural rules and assumptions had to change. Now we consider members of all these groups to be humans.
Similar changes can be expected in regards to AIs, augmented humans and posthumans. The problem is that the scope of the forthcoming changes is much greater than anything the humanity had experienced so far. The differences can be so huge that even the adaptation might be impossible, as there will be no society where the rules can evolve or be enforced.
To illustrate the potential complexities of future interactions I provide one example. Note that the terminology is not agreed upon yet and words such as creature, person, entity, etc., might be interchangeable to some extent.
Imagine a future person (A) wanting to observe the black hole (13). A creates a copy (B) of himself (in a simpler case only a non-sentient, non-intelligent subprogram of A is created), sends it to the black hole in a spaceship to gather information and then in the last moment that copy (or changes from its initial state) is sent back to A. Then A can incorporate the changes into himself or he can decide to discard them. We can say that in some sense B was a separate person. He was basically killed (or committed suicide) with the assumption that later the changes will be added to A to reanimate B.
There was one person originally (A), then another (B) branched from him, then B was killed (with the above assumption) and then he exists only in the form of potential (source: Ettinger, can be considered life/existence). B can be added to A (but A does not want that) or A can be added to B (cause the mind of A is partly owned by B, as A was also B some time ago and A promised (to B and to himself) that his mind will be available to restore B by adding his changes), or some original version of A can be taken from backup and added to B).
We can be already confused by these surprising and unknown interactions between just two entities: A and B. Was that mind in the spaceship falling into black hole still part of A, or was it a separate person? The future entities will undoubtedly be very complex. They might incorporate dynamic mind branching, restoring of mind sub-states, deleting active sub-programs, etc. Additional difficulty will arise from the use of analogue computers, where physical structure will play as important role as the programs. In any entity many infinite-small changes will be happening constantly on the local scale and on the scale of the whole mind. Given that the speed of light limits the communication between different parts, the question arises what should be considered separation and what not. Are sub-programs independent?
The scope of the changes in structure of persons and interactions between them that we will go through is evident from the scenario presented above. Next I will present some existing philosophical and logical approaches to the complex systems of the future. The main questions that need to be answered are of independence of related entities and moral limits in dealing with them. What can be considered part of yourself and what is some independent entity with its own rights?
1) Vernadsky first proposed (14) the idea that all intelligent entities somehow are parts of one giant “mind” or intelligent structure, that he called “noosphere”. While it is not so today, it can technically be made so in the future. Still, even the consensus about the questions of independence among the parts does not solve the problem forever. While there remains an outside world, the potential for interaction remains, such as adding new parts or removing existing ones. In addition, some parts may suddenly feel that they are more than a part of the whole what inevitably leads to conflict.
2) It has been proposed that individuality is sacred in any form, sometimes including even inanimate objects. For example one of the five principles of "Jain dharma", a “religion heavily influenced by Buddhism and Hinduism, … [prohibits any violence towards other] … human beings, animals [and even] … vegetables” (15). The problem with this idea is that it is not a feasible one, as we shall not be able to draw a line between a vacuum cleaner, a human and a Jupiter brain. But obviously, we cannot give them all the same rights.
3) Today’s status quo can be described as biological chauvinism. Most people consider anything that does not look like human not to be intelligent. To a very large extent, this stems from ignorance of basic sciences and technologies. Problem is that we already have cyborgs (people with cardiostimulators, artificial hearts, etc.). In the near future we will add direct connection to the brain (similar to already existing hearing aids) and then make some modifications to the brain. After that, if we reprogram parts of the brain, we will have something that is not human and by the bio-chauvinism has no rights. This is against the common sense as well.
4) A functional approach can be derived from the original Turing work “Computing machinery and intelligence” (16). If an entity operates similar to a human, we can consider it intelligent and hence grant it similar rights. There are numerous problems with this approach, the most significant being that there can also be non-human intelligence (for example, extra-terrestrial intelligence or AI that was created to operate in a quite different environment with fundamentally different goal structure), which cannot be judged by this criteria.
5) One of the oldest ethical principles, the Golden Rule can also be applied. To accommodate for different goals and values, it has to be modified: “Whatsoever ye would that others do to you if ye had their goals, do ye even so to them”. But another problem remains with the priorities – should ye do good to yourself or to others? And example is when one wants to attain infinite processing power and thus infinite matter and energy. In the post-Singularity world one might be able to easily convert all the available matter into its own brain at very large speed. The Golden Rule cannot help compare the value and the importance of different entities.
We see that existing approaches are not particularly effective in dealing with possible realities of the future. It is necessary to radically rethink many basic philosophical concepts. The single most important idea that has to evolve is the idea of self, the idea of person, entity or creature. When previously unimaginable modes of existence become possible, the idea of self and many related concepts need to be updated. Next I will evaluate the ideas of identity, its changes and the role tools play in it, consciousness, time and death and propose some changes to our current understanding of them.
The key change should be making the identity a matter of degree. In our society it became absolute, because in the past all humans we usually interacted with were different enough. This makes it ineffective when dealing with exceptions, such as Siamese twins or clones that prove different degrees of identity exist. The transition to artificial intelligence will make possible any degree of identity, from perfect copies to fundamentally different systems.
I think, in the future the value (identity, importance) of a person will be determined neither through comparison with others, as they can be very similar – parts can be copied verbatim, nor through the similarity with them or with some idealized models, but through how well is this person suited for fulfilling its goals. As people move up to self-actualisation on the Maslow’s Pyramid of Needs (17), they will associate themselves with their goals, not unsteady qualities.
While changes in abilities, programs, qualities, knowledge or memories are definitely going to become commonplace, the real change of identity will happen whenever the goals are modified.
Changes from within, such as reviewing and updating the personal goals, have to be accepted. But the situation is more complicated with changes from outside. There is nothing inherently bad about this; there are positive examples, such as changing the criminal’s mind. But in general it might be considered harmful and unethical to do this. The only way to develop the necessary set of rules is to go through evolutionary changes starting from rules and values we have today. Changing the goals of a simple robot (vacuum cleaner) is clearly acceptable. In the case of more complex entities, it will probably be a matter of degree, with such factors as good reason for change, small amount of changes, gradual changes, agreement with the entity that is being changed, low intelligence level of the original entity making the action more acceptable.
Part of our identities will consist of various tools integrated into our bodies and minds. While the situation seems clear at the modern level, the tools quickly become indistinguishable from our selves when we move to higher technology levels.
Originally people used material tools, but everything that was outside of the body was not considered part of the person. When some implants started being used, such as teeth and cardiostimulators, even artificial objects became almost human parts, as long as they are inside the body. But computer-to-brain connections, such as existing cochlear implants are in a grey area. It is not clear if they are part of a person, especially, if they are partly outside of the organism. Same situation with more modern AbioCor artificial heart (18), the question remains if the charging device with batteries, for example, that we wear outside and that is not connected to person (except electromagnetic transmission of power) a part of human?
Gradually we shall begin using thinking tools that are connected to brain, first using them as an interface, but we still shall be able to differentiate between natural (neurons) and artificial (chips) elements. It will become more complicated, as we depend more and more on our tools, like Steve Mann (19, 20), who has difficulties living without his wearable computer. And finally, when the substrate for our intelligence is the same as for our tools, there is no clear border any longer. Still new questions will be raised. What if some vital software, like image recognition, is used by several persons at the same time? What if the personal memory is shared? Can it still be considered part of a person, or will several persons be able to share a part of their identity, of their self? To make these questions more urgent, Mann is already allowing other to see with his “eyes”.
We can imagine several such examples, ranging from a small group of cloned humans with shared memory and thoughts acting together (21) to a full-scale Hive where individuals give up all their individuality.
The consciousness is not much more than meta-knowledge, when an entity is able to think about itself differently than about rest of the world. (22) There is no other difference. It is a function that can be added or removed. It is likely that ethical norms regarding consciousness will develop gradually and we will realise the differences between conscious and not entities and understand what can be done to either of them (murder, torture, sacrifice, etc.). This is a question that can only be answered based on the defined values.
Personal identity also changes over time. But personal time is just an illusion, unrelated to physical time. The whole life of the person can be stored in a limited-length book, assuming the brain is computational to a sufficient degree. It can even be printed backwards. In our example of falling into black hole this issue will be further complicated by the General Theory of Relativity, as the personal time for A and B will have significantly different speeds. The Omega Point example (23) shows that even when physical time slows and the length of the period approaches zero, our personal time can be potentially infinite.
The rational approach to life and death takes into account the existence of information that can be used to replicate the entity (24). The potential of replication in some sense means that that entity is still alive. Paradoxically that means that we cannot know for sure if something is alive or not. If we use the behavioural approach (similar to Turing test) and consider an entity alive when we can interact with it or get some output from it, another problem arises. Then even in the case an entity is conscious and is thinking and therefore is subjectively alive (25), it is not clear (in the quantum sense, not just unknown) for the outside world, whether it is alive or not, as it depends on some future output.
In Clark’s “2010: The Year We Make Contact” (26) the AI researcher Dr. Chandra says “Whether we are based on carbon or silicon makes no fundamental difference. We should each be treated with respect.” Ironically, just like we are unable today to explain the thinking process using programming languages, we are unable to rationally explain the subtleties of interactions between intelligent beings without resorting to intuitive human concepts such as respect.
In general, humans are very bad at solving complex issues that involve lots of subtleties – examples, such as cloning, GM, and whatever is today’s hot topic, are abound. The “political” process is extremely ineffective and prone to errors. We need something better to find answers to the hot topics of tomorrow outlined above. And it is vital that we start thinking about the implications of further AI development as soon as it is possible.
1. Intel Advanced Research In Integrated Silicon To Fuel Convergence Of Computing And Communications Over Next Decade, 2002, http://www.intel.com/pressroom/archive/releases/20020912corp.htm
2. Ray Kurzweil, The Law of Accelerating Returns, 2001, http://www.kurzweilai.net/articles/art0134.html?m=1
3. Michael Negnivitsky, Artificial Intelligence: A Guide to Intelligent Systems, 2002
4.
Vernor
Vinge, The coming technological Singularity - how to survive in the Post-Human
era, 1993
http://www.globalideasbank.org/SD/SD-107.HTML
5. Eliezer S. Yudkowsky, Staring into the Singularity, 2001, http://www.sysopmind.com/singularity.html
6. Nick Bostrom et al., The Transhumanist FAQ, 1999, http://www.transhumanism.org/resources/faq.html
7. Hugo de Garis, The Artilect War, 1999, http://www.cs.usu.edu/~degaris/artilectwar.html
8. Zeck, G. & Fromherz, P. Noninvasive neuroelectronic interfacing with synaptically connected snail neurons immobilized on a semiconductor chip. Proc. Natl Acad. Sci. USA 98, 10457-10462 (2001)
9.
Stuart
Hameroff, Roger Penrose, Orchestrated Objective Reduction of Quantum Coherence
in Brain Microtubules: The "Orch OR" Model for Consciousness, 1996
http://www.consciousness.arizona.edu/hameroff/Pen-Ham/Orch_OR_Model/The%20Orch%20OR%20Paper.htm
10.
Robin
Hanson, If Uploads Come First, 1994
http://www.kurzweilai.net/meme/frame.html?main=/articles/art0209.html?m%3D7
11. Turing, On Computable Numbers, 1936
12.
Robert
Bradbury, Jupiter & Matrioshka Brains History & References, 2001
http://www.aeiveos.com/~bradbury/JupiterBrains/
13. Source unknown. A short sci-fi story.
14. Vernadsky V.I., Zhivoe veschestvo, M., 1978
15. shadows, Everything2, http://www.everything2.com/index.pl?node=ahimsa
16. Alan Turing, Computing machinery and intelligence, 1950
17. Abraham Maslow, Motivation and Personality, 2nd ed., Harper & Row, 1970
18. AbioMed website, http://www.abiomed.com/
19. WearCam.Org, http://wearcam.org
20. Prof. Steve Mann’s page, http://www.eecg.toronto.edu/~mann/
21. Source unknown. A short sci-fi story by some Russian author, approx. 1970-1980
22. Douglas R. Hofstadter, Gödel, Escher, Bach: an Eternal Golden Braid, 1979
23. Frank J.Tipler's Web Page, http://www.math.tulane.edu/~tipler/
24. Robert Ettinger, Prospect of Immortality, 1964, http://www.cryonics.org/book1.html
25. Rene Descarte
26. Arthur C. Clarke, Peter Hyams, 2010: The Year We Make Contact, 1984