How much is a perfect baby worth? In cash, to specifications. The parents get to say what perfect means – not just free of obvious diseases, but immune to cancer and AIDS; tall, if they want, and certainly intelligent and well-coordinated, with excellent eyesight; their choice of skin- and hair- and eye-color; and of course whichever sex, maybe even sexual preference, they prefer.

Impossible? Maybe. But some scientists are raring to try. They hope to make genetically modified humans. GM food is on the supermarket shelves, GM mice are in laboratories all over the world; people are next. And the first real step towards this could happen as soon as next year.

What is proposed is called 'germline engineering', because it affects the 'germ cells' meaning in practice the eggs and sperm with which children are made; so the children would normally pass the changes on to their children, and they to theirs. Technically, the operation would generally be performed on the very early embryo, after the sperm and egg have come together but before the cells have begun to differentiate, to make all the different kinds of cells that make up a body.

This is different from 'somatic gene therapy' which is the attempt to cure disease in a living person by fixing the genes that caused it; it is called 'somatic' because it works on the body, and 'soma' is the Greek for body – this kind of treatment is not intended to affect the patient's children. And cloning – producing offspring that is genetically virtually identical to the (single) parent – is something else again, though the techniques involved are somewhat similar. GM babies are not clones, although clones could theoretically be used to make GM babies.

Right now, the Human Genome Project is rushing towards completion. This is the enormous, and hugely expensive, attempt to identify all the human genes, whose first draft was announced (somewhat prematurely) in 2000. Surprisingly, given the level of hype that surrounds the project, we still don't know how many human genes there are. The current estimate is 30–40,000, but as recently as May, 2000, the experts at a major conference started a sweepstakes and the bets ranged from 27,462 to 153,478. (Entries cost $1; the price rose to $5 this year, and $20 in 2002, on the assumption that the pot will be paid out at the 2003 event.) Some scientists think, on theoretical grounds, that the whole effort has been over-sold and betrays a misplaced faith in the idea that there is such a thing as a normal human genome, but certainly a lot of data has been amassed. And after all that effort, and all that money, of course they want to use it.

The justification is medical. Some diseases are definitely associated with specific problems in a precise area of a particular chromosome. This is rather unusual. There have been a lot of reports (typically rather rapidly retracted) of the discovery of 'the' gene for a particular condition, either a disease or a behavior, but when you look at them closely, you almost inevitably find that they are talking about statistical correlations, which basically means that more of the people with this condition have this genetic abnormality than people who don't. Which doesn't do anything for people who have the disease without that abnormality, and only frightens those who have the abnormality but may never develop symptoms. Still, genetic diagnosis is sometimes possible, and it's fair to say, for those patients, that this gene causes this disease. In which case, if you can replace that gene with a properly-functioning one, you can cure the disease. Sounds too good to be true, doesn't it? Well, so far, it is.

Somatic gene therapy has turned out to be much more difficult than was once hoped. Even if you know which gene you want to replace, you are faced with the problem of replacing it in every relevant cell of the body (or at least most of them). And that is not easy. There have been reports of some successes, but there have been far more scandalous failures. Most notoriously, in May, 2000, the University of Pennsylvania shut down its human trials after its investigation into the death of a patient revealed numerous ethical as well as technical problems. It turns out that many other therapists had also not been notifying the National Institutes of Health (NIH) of problems as they were supposed to; after a reminder, reports of problems went from 39 in 93 studies to 691 in the same 93 studies. A Harvard scientist had actually failed to disclose that three of his first six patients died (for other reasons, but they should have been reported). The optimism of a few years ago has been wearing very thin.

Ironically, the problems with somatic gene therapy have actually increased enthusiasm for germline work – it's just so much easier to affect all the cells at that stage. But to do that, you have to diagnose the problem immediately. Which means traditional, old-fashioned sex is out of the question; you have to be using in vitro fertilization (IVF), that is, test-tube babies. Reproductive technology is now a $2bn-a-year industry, and an estimated 11,000 children a year are born after IVF treatment, so this is not implausible, in the rare cases where parents know that they are at a high risk of having a genetically damaged child. IVF also gives about a one-in-four chance of success, incidentally, so it's not terribly efficient, and the industry has been prone to scandal and abuse, but the technology does basically work. Still, it's extremely unlikely that every child conceived by the same parents would have the same disease. So, if you can test for it (and we can, using preimplantation genetic diagnosis), all you have to do is to create several embryos, as is common in IVF treatments, test them, and make sure you only implant the healthy ones. You don't have to do germline therapy at all.

Unless you want a 'better' baby. And that is exactly what some people do want. Not just science-fiction writers and blithely optimistic techno-futurist visionaries, but respectable professors at major universities. The most easily accessible advocate is Lee Silver, a professor at Princeton who wrote a best-selling book called Remaking Eden, in which he envisages people so modified that they literally could not breed with ordinary humans – they would be a different species. He calls them the 'GenRich' as opposed to the 'naturals' and predicts a society in which the GenRich are "a modern-day hereditary class of genetic aristocrats" comprising 10% of the population, while the 'naturals' – 90% of us – "work as low-paid service providers or as laborers" and don't even get educated properly, instead being taught only "the basic skills they need" to do their mindless jobs. Originally he put this scenario 350 years in the future, but by the time the paperback came out he had accepted the advice of his colleague, Professor Thomas Shenk, the chairman of Molecular Biology at Princeton, that this was far too conservative. He's now expecting to live to see at least the start of it.

Silver is not alone. A 1998 Conference on Engineering the Human Germline was organized at UCLA by Gregory Stock and others, and attracted a distinguished panel, including the grand master of modern genetics, James Watson, co-discoverer of DNA and the founding Director of the Human Genome Project. Another was sponsored by the Extropy Institute, an extreme libertarian group that does not believe in 'forcible taxation' and advocates a 'transhumanist philosophy'. They also think the Food and Drug Administration (FDA) should be abolished – the consumer is king and should be allowed to buy any medicine they want without any bureaucrat deciding what is safe and what is not (a prospect over which snake-oil salesmen would be licking their lips).

Not everyone goes that far, but most of the people pushing for germline manipulations think the government should stay out of it. So far, it almost has, in the US. The Clinton administration banned the use of federal funds for research into human germline modification, which is a significant deterrent, but if you use your own money, you're probably in the clear; the FDA claims a right to regulate, but this is disputed. Elsewhere, the situation is different – Germany has the strongest stance against germline interventions, and the whole Council of Europe is in the process of ratifying a complete ban, as is Canada. Many other countries, including India and Japan, have banned human cloning, but technology may be moving faster than the law-makers. Which is exactly what the scientists want. "If there's a terrible misuse and people are dying, then you can pass regulations," says Watson, adding, "I think it would be complete disaster to try and get an international agreement." Meanwhile, Mark Eibert, a California attorney, is threatening to sue the state over its moratorium on human cloning, which he claims is unconstitutional.

In general, the idea being promoted is that anyone who can afford it can do anything they want. After all, they say, parents pay to give their children the best education, so why not give them the best genes?

First, because it's wrong. It is as wrong as love is right. Certainly there are people who rationalize their choice of partner – it's commonplace to talk about what excellent children they would make together – but if that is all there is to a union, there is something dreadfully inhuman about it. This is an argument that will never satisfy the self-described rationalists who advocate germline modification, so there is no point in elaborating except to say this: If the prospect appalls you, trust your instincts. And if it doesn't, are you really sure?

Legally, it is absurd to claim that society does not regulate the way its members raise children. Child labor is forbidden; education is compulsory. And the argument that we should allow human genetic modification because it will happen, legal or not, does not hold water either. Incest is banned, and incidentally was prohibited long before any scientist could define the harm caused by inbreeding; it occurs, and no one claims that as an argument for legalizing it. If society chooses to outlaw germline engineering, that is society's right.

Scientifically, there are those who say that the entire effort is futile. Something will happen if we fiddle with genes, that's for sure; what's not certain is if it will ever be what is intended. Experiments involving the introduction of human growth hormone into animals have occasionally resulted in large mice, and much more often in no surviving mice at all; or sometimes in lean but not especially large pigs. And so on. Most of the animal experiments have failed – which is no knock on the investigators involved. They have been trying to find out something obscure by performing difficult and complicated procedures, and have indeed advanced the cause of science by analyzing their own mistakes. That is one way of moving forward. Animal-rights activists complain when animals are used this way; almost everyone would object if the experimental creatures were human. To which the proponents argue blithely, if it doesn't work, it won't continue. And how many heartbroken parents and damaged children will it take before they come to that conclusion?

Some of the advocates of human genetic engineering say that what it leads to is just an extension of the society we live in. And they almost have a point. A society that doesn't mind the rich getting ever richer while the poor don't, that allows some 30 million people to have effectively no access to healthcare, that keeps increasing fees even at state colleges till good students are priced out of education, that celebrates competition at all costs and greed as a social good ... that society might think that it would be acceptable to gamble its genetic heritage on the off-chance that its rich kids might get even more advantages.

Or we could look at this prospect and say, enough. If that is where this road goes, we should change direction. It is time for some balance in our lives, as individuals and as a society. Time to spread the wealth. Time to focus health resources where they can do most good. Time to treat everyone as equal.

And we had better hurry. In September 1998, W. French Anderson and his colleagues put forward a proposal to investigate what they call 'in utero gene therapy' which blurs the distinction between somatic and germline interventions, some think quite deliberately. They hope to fix a specific genetic disorder by injecting a replacement gene into a three-month-old fetus, acknowledging that "an occasional vector particle may enter an egg or sperm, thereby resulting in germline gene transfer." Animal studies are under way, and they hoped at the time to propose "clinical protocols" in about three years. That means, plans for the treatment of people. And the risk is, that without anyone knowing exactly what is going to happen, people will be changed forever.

How much is a perfect baby worth? To an infertility specialist pulling down $625,000 a year (they are now the highest-paid of doctors), plenty. Do we really want them selling GM babies as the latest in status symbols for the very rich?

All babies are perfect. And priceless. Hold that thought.