Before we ask why robots should have emotions, it is useful to ask why we have emotions. Although it is usually assumed that robots/computers will not have emotions (unless, as with Commander Data in 'Star Trek - The Next Generation' they are 'programmed in'), it increasingly seems that emotion (whether accompanied by consciousness or not) is integral to the prioritisation of goals.

It would seem - although I am open to disagreement - that all animals experience emotion, whether consciously or no. So emotions are either something which all higher (terrestrial) lifeforms have needed to evolve, or an accompaniment to something which they have apparently all evolved (whether or not emotions are essential). This suggests that emotions are extremely useful for any being which needs to perform well in a complex (mental as well as exterior) world. It is therefore unlikely that emotions are 'intrinsically human' or in principle only characteristic of organic lifeforms.

People with 'emotional deficits' have been found to be less able to choose between goals; they get mired at the stage of deciding what to do.

But while this evidence is certainly suggestive, it is no more than circumstantial. Is it possible for drive to exist without emotion?

Emotion is not just about realizing what your goals are; it's also about wanting to act on that knowledge. If you do not have frustration and longing to impel you, then perhaps you will end up doing nothing. Why try and win a mate, gain top marks, make friends or get food if you don't really care? It may be that emotions are so intertwined with drive that an active being couldn't function without feelings being wired in.

[more later - I'm not satisfied with these arguments yet.]

The fact that we are conscious may be one of the most remarkable of all.

Before I go on, I should probably explain what I mean by consciousness, because surprisingly often, philosophers and others have mistaken other things for consciousness: for example, self-awareness, 'theory of mind' (i.e. the attribution of thoughts and motives to others), and attention. Although these may often be associated with consciousness, you can trust me when I say they are not the same thing. Think about it: you can construct a tiny robot that pays attention to itself, constructs simple hypotheses about others, goes to sleep sometimes and pays attention at others, and has no consciousness whatsoever. It acts, but doesn't experience (that statement is unscientific, because I can't prove that such a robot isn't conscious. But I'm pretty sure I'm right).

So: why has evolution bothered to provide us with awareness? I can think of two possibilities. One is that consciousness has a specific role, being intrinsic to higher mental functions (such as formulating a novel course of action, for example). Another - which I favour, partly just through intuition - is that consciousness is simply an 'emergent property'; one which does not actually have a role, but which tends to arise spontaneously in organisms (whether organic or inorganic) which are appropriately organised and sufficiently complex. (It is unclear which is more important - type of organisation or degree of complexity). I suspect that although we are now roughly able to detect where consciousness is at a given time (for example, by determining the location of the ‘P-300 wave’), the answer to why consciousness arises will not become apparent for a while yet.

The mathematician Roger Penrose (outstanding in his own field) has attempted to show that Artificial Intelligence will always lack a ‘certain something’ - consciousness, free will, or somesuch thing - because of the pivotal role of quantum effects. His efforts have been widely lambasted, so my contribution to the admonitory flood will be brief. Simply this: that quantum effects act on a physical level; the functions of our carbon-based bodies can mostly likely be replicated in other media (given time); and therefore, that any interaction between quantum phenomena and our bodies can probably be replicated in other media - those of Artificial Intelligences.

In any case, I think the 'quantum escape clause' is very much overused in a variety of philosophical fields - in fact I've never seen it used well (not that I know much about quantum physics).

But I haven't shown that all intelligent machines will necessarily be conscious. I don't think that consciousness is necessarily copperfastened to intelligence; I think it could also be linked to the way a brain is constructed. For example, ant nests behave in very intelligent ways, because all the separate ants work together as a unit; but I don't think many people would argue that there's a sort of unifying consciousness hovering between the ants. If a machine doesn't have parts that communicate with each other very quickly and complexly, it might be intelligent, but it might not be conscious. Perhaps consciousness will only emerge from a brain that's internally joined up; perhaps there are other ways that consciousness is dependent on structure.

BLIND AND SIGHTED WATCHMAKERS

'The blind watchmaker' - the title of a book by Richard Dawkins - is evolution. The 'sighted watch-maker' is ourselves.

The above arguments demonstrate (to my satisfaction, at least) that AI is in the early stages of a dramatic life-creating project. So: should we allow it to continue? What are the objections?

One argument is that it is wrong to 'play God'. But evolution, a blind and amoral process, has been 'playing God' on this planet for approximately four billion years. The result: the natural world, which has - at least, perhaps, until recently - been more characterised (among the sentient animals) by pain and struggle than by happiness. It has also, of course, resulted in the fascinating species known rather narcissistically as Homo Sapiens Sapiens ("Wise Wise Man").

Life in the wild, as any naturalist will tell you, is often appallingly cruel; our own lives are not endless merry-go-rounds of delight. However, I doubt whether many of us - bar a utilitarian or two - would argue that it would have been better for life never to arise. If blind processes can have results which we value so much, why should we forbid the creation of an entity whose faculties (creative, emotional, intellectual) could be so much deeper and more advanced than ours? Or to forbid artificial implants into the brain? As noted earlier, every schoolmaster is engaged in changing brain structure. To play the conservative God in this instance would be equivalent to going back 65 million years and deflecting the life-changing asteroid which killed some of the most intelligent (most sentient?) beings then in existence - the dromaeosaurid dinosaurs (raptors). Why do that? In order to prevent the global dominion (among large creatures) of more mentally agile creatures: mammals, the group from which Man [sic], the 'paragon of animals', has sprung. Many of the arguments being bandied about ('we must not play God', 'we must not endanger humanity by allowing robots to develop intelligence') are reactionary and destructive.

We cannot shackle development based on the objections given above; we must manage it and make sure that it does not worsen the lives of any conscious beings. There is nothing so wonderful about preserving our human form with its limitations - most if not all of our inventions have in effect been attempts to supersede these limitations. Surely it would be a terrible waste of the potential for enhanced mental experience if we were to prohibit the invasion of our skulls by 'exterior prosthetics'. Objections to changes to the human body and mind can only, I think, be valid when they deal with the psychological and social effects of cosmetic alterations.

SHOULD WE STAY HUMAN?

Eugenics, Designer Babies, and Jerking Knees

Eugenics was the twisted child of Darwinism; it was one of the theories underpinning the barbarities of the Holocaust. Its proponents thought that natural selection (the process which has created all lifeforms on Earth) was being interfered with by the forgivingness of liberalism and the modern welfare state, and that in order for humanity to become truly healthy, the dregs of society would have to be siphoned off. Eugenicists' approach was two-pronged: people of 'good stock' were to be encouraged to breed, and 'degenerates' were to be kept childless. Active policies of discrimination were integral to eugenics: people of degenerate stock (the unemployed, blacks, slum-dwellers, gays, Irish [hurrah!], gypsies, etc. - basically any group that the eugenicist was already prejudiced against) were to be sterilized or even killed. Eugenics was not confined to Hitler's Germany. Here are just a few of the other perpetrators: Sweden, China, Singapore, and the United States. It was a black chapter in the history of ideologies.

There is a superficial similarity between eugenics and some of the genetic modifications which are being mooted at present; but I think that many debates on humanity's future have been distorted by this. If you listen to a discussion of the role of genetic engineering in humanity's future, you will probably hear the talking heads bandy about terms like 'designer babies' and 'eugenics'. (A designer baby is one whose parents have chosen to engineer what it will be like: for example, they might decide that it will be blue-eyed and good at sports, or black, tall, and very intelligent.) I think it fair to say that most people are extremely wary about making any 'improvements' to the human race. After all, wouldn't such alterations return us back to the bad old days of Hitler and of the depraved Nazi doctor, Joseph Mengele?

Well, what is the difference between the medical procedures that everyone (bar a few Jehovah's Witnesses and perhaps a smattering of others) accepts today, and these ideas of designer babies and eugenics?

• At present, we accept that medicine can be used to tackle 'deficiencies' (illness, disability) which prevent people from living normal lives. But with eugenics and designer babies, the 'betterment' of the human race is also sought.
• With eugenics, people who aren't considered to 'measure up' cannot be allowed to reproduce. In the worst incarnations of eugenics, these degenerates are earmarked for murder. Note, however, that we can have designer babies without thinking that non-designer babies should be discriminated against.

It is obvious that eugenics is a repulsive doctrine. Yet I would question whether designer babies should be placed in the same category. Here, the aim is to make some mental or physical improvement; it need not involve discrimination.

I think a comparison with education is instructive. Imagine two parents. One decides to have a designer baby with a superb brain; another decides to have a normal child, and give it a superb education. A third child will have neither incredible brains nor a great education. In both the first two cases, the child's mental ability will be enhanced. The designer baby will have greater skills no matter what the training; education, by providing the other child with greater language skills and theoretical tools, will increase its practical mental ability. The third child will lag behind both. The downside of education (and technology) is closely analogous to that of genetic modification: it creates class differences between those who have access to the techniques, and those who do not. Children who leave school at fifteen are generally less likely to end up with top jobs; similarly, we can expect that if we engineer super-smart kids, they will generally take the cream of the jobs on offer.

Q: But isn't designing babies somehow worse than education? Deliberate genetic manipulation is more invasive, less natural.

A: Well, if we used naturalness as our guide to living, we would have to abandon almost everything that distinguishes Western life.

Q: What about the invasiveness?

A: Although teachers don't physically insert knowledge in their pupils' brains, they're undoubtedly engaged in changing the structure of their pupils' brains 'for the better' (think of the reaction that phrase would arouse in the context of designer babies). The effects of education are usually as permanent as those which would be brought about by manipulating the genes of an unborn child.

I am not attempting to argue that the effects of education are necessarily as wide-ranging as those which genetic manipulation (GM) might have. For example, whereas education creates differences in actual learning, GM (and given time, AI), by altering the capacity to learn, can create far more dramatic differences between haves and have nots. However, this is a difference of degree rather than of kind; it appears to me that one should either be against all such changes, or against none. If we are willing to accept one form of brain alteration, why shouldn't we accept the other? The central issue is not whether we should allow people to improve themselves mentally and physically, but that the people who remain unchanged - whether by education or GM - should not be discriminated against.

Whilst a certain amount of information can be readily transferred from an entity before its death, the vastly complex interactions which constitute that entity are harder to replicate - or to keep replacing while it is alive. Although replication is not survival for a Self - if you copy a cassette you end up with two, not one*, I refer to it because of the curious difficulty in locating the dividing line between replication (without proper integration into the organism being copied) and sustenance. I inhabit (indeed am) my body; but a number of years ago, I inhabited a body whose constituents are now far away from me. So:

* something that has puzzled many viewers of Star Trek - do they kill off the Captain Kirk who's left behind in the transporter room? I think many people would agree that the optimal strategy would be to kill both versions, using prolonged and sadistic methods.

As I argue above, even the most interesting and productive tasks (apart perhaps from things like sex and eating) will eventually be performed best by machines. So what will humans be doing if the most important decisions and discoveries are being made by machines? The meshing of humans and machines will increasingly have consequences which many see as undesirable. Horrendous though some might think it, it might eventually be possible for all humans to live their whole lives without moving (if you've seen the highly entertaining film 'The Matrix', you get the idea)*. What's more, it is conceivable that these lives could be utterly satisfying, and completely unproductive. If humans as we know them still exist, the needs for social and sexual interaction and for physical exercise could perhaps be entirely sated without budging an inch. Already, many people are becoming absorbed in a 'computer world' as an alternative to engaging with other members of their own species.

* It is worth bearing in mind, though, that this is merely an extrapolation of current trends. Trends often peter out; things move off in a completely different direction. For example, the emergence and importance of airplanes and computers were both (to my knowledge) largely unforeseen, except by a few lone thinkers. In the case I'm discussing, mobile computing (mobile phones, WAP and the like) could ensure that our species doesn't collapse into an abyss of sloth.

I will now ask a question which, I am sure, will seem outrageous to some people. Why should we value 'real' relationships - relationships with other humans - over relationships with other (non-human) machines? In the short term, the answer is clear - machines do not cater to all the needs of humans. They cannot love us, we cannot raise them as children; because they are (note tense) not conscious, a relationship is meaningless. This says less about what is (in an absolute sense) right and wrong than about the makeup of humans; we need specific types of relationship with other feeling entities. Thus far, only humans have been capable of providing such relationships; but this situation is changing. The role of machines is increasing in areas such as the following: sexual imagery and toys; lifelike animations of people; music and art; and, slowly, conversation. What is required for human happiness is certain types of sensory and physical stimuli. Can all of these stimuli be provided by machines? And will they? I think that eventually, all of them can. But what I am not sure about is whether people will accept relationships which are only apparently complete - perhaps more 'complete' than most of today's relationships, which, as you know, involve compromise between people with divergent emotional and intellectual needs.

Say machines provide you with a virtual reality in which you can choose as many or as few (constantly willing) sexual partners as you wish. Each of these virtual people (or whatever sort of entity you choose to interact with) can engage you in constantly stimulating conversation. They will shutup if you want them to, or if you prefer feisty companions, that can also be arranged. If you wish, you can lead armies to glorious victories against the forces of darkness every day, or spend your time knitting stupendous virtual sweaters. If you can't think of an intellectual or emotional challenge to keep you occupied, the machine can do so. If your chemistry predisposes you to periodic dissatisfaction, that can be changed if you like, or if the machine deems it suitable. Suppose that you can do all this, and that you know it to be partly illusory. The intellectual interaction could be with a conscious machine, but it might essentially be patronising you - keeping your inferior mind occupied and entertained. All the best thinking, even the most intense feelings and artistic experiences, would be the preserve of machines. How far would you reject such a lifestyle, and spend your 'quality time' with humans (often tiresome beings, but beings like yourself)? I suspect that many people would spend less time with their own kind than they might think; but perhaps there will be a persistent tendency to avoid a life in which we are only spoonfed by our vastly smarter mentors and the diversions which they (and perhaps some humans) construct for us.

Although I agree with Warwick that machines will eventually become vastly smarter than us, I think the above quote is rather flawed. Certainly, the human race will no longer persist as we know it; it will probably no longer be the dominant force. But to assume that they will subjugate us is to take guesswork a step too far. How can we know that a far more intelligent creature than humans will not take our welfare into account? Or that morality will not be either built into their brains, or simply an intrinsic part of high intelligence? Even humans, who are neither particularly intelligent nor outstandingly moral, often treat other species with a lot of respect; and these positive behaviours are on the increase. (Unfortunately, however, our destructive power is also on the increase; so the more technologically advanced we become, the more moral we need to be).

I have tried to work out how machines will treat us in the future. One observation I would make is that Warwick's 'slave worker' scenario is probably far-fetched. Creatures of such physical and manual ability should probably not need humans to carry out tasks for them. But in general, I found - and was correctly informed, mainly by Kenny Lynch - that my investigation was too much to take on. For now, I think we just can't know; even though I'm intuitively optimistic about it, there are too many ifs and maybes to make a convincing case. I feel pretty sure that their intelligence will vastly outstrip ours; but as for how they will act when they become the most powerful creatures in the world, I think it's just a matter of wait and see.

CONSCIOUS BEINGS IN INORGANIC MEDIA:
Anthropomorphism, Anthropocentrism and the Constriction of Ethics

"These [software entities] are not models of life, but independent instances of life . . . Because everything we know about life is based on a single example, our every conception of life is highly parochial. For most people, it is impossible to separate our concept of life from the material in which our one example is embedded. Organic, wet, material, etc., are integral parts of our conception of what life is"

It is not only with regard to other animals that humans are biased; although advocates of eugenics are often seen as (and quite often are) racist weirdos with aspirations to Godlike status, scarcely a whimper is aroused by the efforts of the AI community to realize its primary goal: computer programs which aren't mere idiot savants, but can reason intelligently about a wide range of things, be wise, and all the rest of those things which humans are so far best at (on Earth, at least).

If the AI people do achieve this - and I think it will eventually happen - I would guess that the resulting intelligences stand a good chance of being conscious* - probably far more so than ourselves. If you think about your own consciousness, it is (like mine) fairly limited most of the time; the novelist Milan Kundera was right when he spoke about the 'lightness of being' (although I'm not sure whether he should have called it unbearable). Something that strikes me as I go about my life is how unthinking and automatized I am a lot of the time. Obviously, this kind of living - where you're not all that conscious of anything much - is the result of instincts, and of well-memorized behaviour sequences such as the one for tying shoelaces. During emotional or exciting experiences, my consciousness is at its most activated; but most of the time, my experience of the world is a fairly low-level thing. I can easily imagine an AI having a far more intense, awake life than humans have.

* This is not a scientific statement, because we don't know how consciousness arises. If it's mainly due to how the brains of 'higher' animals are organised, then perhaps it will not arise in machines/computers. But I suspect - without much evidence - that complexity is also very important, and that many different types of structure could give rise to consciousness.

Ray writes that

"The umbrella of Artificial Life is broad, and covers three principal approaches: . . . . in hardware (e.g., robotics, nanotechnology), in software (e.g., replicating and evolving computer programs), in wetware (e.g., replicating and evolving organic molecules, nucleic acids or others)."¹

¹ From 'An evolutionary approach to synthetic biology', Artificial Life 1, 1994.

This reminds me of some mistakes people commonly make when thinking (or intuiting) about life:

When they see two equally complex lifeforms, one of 'wetware' and the other of hardware or software, they are more likely to consider the wetware as life. The bias is especially strong with regard to software, because it is even more different than robots are from the lifeforms that we are accustomed to. Robots can look and move a little like animals; but computer programs flit around like the brainwaves we never see. But it is the software - in brains, in robots, and in computer programs - which is the real seat of intelligence.

Imagine a computer program which is as intelligent as a human, in the same ways (an unlikely prospect, but bear with me for argument's sake). These computer programs are being made to serve humans. Now imagine a brainless foetus which is being grown so that its organs can be harvested by doctors and implanted into sick people. My guess is that many people would be more put off by the latter scenario, in spite of the fact that it is the foetus which is mindless and therefore undeserving of ethical consideration.

Often, our ethical behaviour is determined by how similar an entity is to ourselves. This is already a problem with regard to how we treat different animals: for example, I would guess that most people would be more distressed by witnessing a cat being tortured - and yowling with pain etc. - than by witnessing the same treatment meted out to an octopus, which - so I have heard - is about as intelligent. At present, new lifeforms are arising, whose physical and behavioural characteristics will generally be even more dissimilar from those of humans.

As the programs we create grow more intelligent, they will increasingly fall within the ambit of ethics; we must be careful to avoid mistreating those which possess emotions and consciousness (of course, it isn't possible to mistreat those which don't), and to watch out for the bias against dissimilar lifeforms. However, this may be a relatively short-term problem, as the management of life could slip out of human hands, and be (wholly or partially) taken over by more intelligent machines. Nevertheless, it would be a good idea to keep a close eye on our creations, and to think about ways of determining whether they might be conscious. The Turing test?

[more later]

Thanks to Kenny Lynch and J. Arlen Pruitt for their help and advice on this page.