1. Not good enough for YOU maybe being slightly better at solving a class of problems or a better memory (on almost any metric) is a measurable improvement and qualifies.

2. Irrelevant (just like #3 and #4) because we're not talking about identical cloning.

3. We domesticate on iteration and result, not strict design.

4. Reliability is a lot like saying we get results, just not 100% of the time.

We probably already have altered humans in circulation. The Israelis aren't waiting around for approval, they are just doing it. I wouldn't be surprised by someone discovering the US Military has embarked on a genetic program (Bourne Identity style). Even the Germans were doing it in WW2, so I could see it coming to light in my lifetime. It wont be commodified in the US in 50 years, but swiss facilities might be an option.

You're referring to Nazies, so are you suggesting simply sterilization of "stupid" people? That gets you most bang for the buck and could have been done since 60's with ease.

God, let's leave the idiot child of human engineering (eugenics) out of this.

I can almost forgive the early twentieth century for its idiocy, but this is a computer science heavy board and evolution is a hill-climbing algorithm. Reducing variation in the population (eg, killing/sterilizing "stupid" people) just clusters your population more tightly, which makes it prone to getting stuck on local maximums.

It's almost acceptable if you lived before DNA was understood and either thought that the evolutionary landscape was perfectly sloped upward or our genome involved continuous "factors" that could be "concentrated", but nowadays we know that DNA comes in discrete chunks (genes) and that the edit distance between two good versions of a gene can be greater than one (eg, you might need three mutations to get from good gene A to better gene B, and it passes through less-good genes C & D to get there).

Eugenics is entirely a short-term gain: you entrench your population solidly in a local maximum, which gives you a higher average than before, but prevents you from ever reaching a higher non-local maximum.

(And that's assuming that your environment is stable and your selection function was well-chosen.)

How about "not good enough to justify the monetary cost and risk of failure"?

Our current methods of genetic engineering involve things like "use a really tiny needle to inject a few hundred copies of a gene into a cell; hope a couple of them are inserted into DNA in such a way as they're expressed and none of them are inserted into the DNA in such a way that they stop other important genes from working".

This is fine if you want to engineer, like, wheat. You do this to a bunch of wheat, you plant it, you grow it, you discard the plants that don't grow right or just don't have the target change, the ones you keep you breed the old fashioned way until you have a stable plant line expressing the desired gene.

Applying a method like this to humans means you end up with a lot of sickly or dead or malformed babies, a lot of normal but rather expensive babies, and some number of babies on which the treatment was successful and the target genes were correctly inserted.

Assuming either you can filter out the failures at an early enough stage that no one cares, or you are such horrible people you're cool "filtering them out" as 1, 2, 3 year olds, or you're cool with creating birth defects but nice enough to provide economic support and health care to them for their entire lives as long as you also get some number of smart babies...

So you've created some adorable abominations against god and man!

Well, do the genes you inserted correctly do what you want them to?

Our DNA does not contain lines like "max_items_in_shorterm_memory = 7" that you can just change to a 15 or a 50. It codes for proteins, molecules of funny shapes that either serve as building blocks for our cells directly or catalyze reactions or do any number of other things, and it codes for in which cells these proteins are manufactured and in what quantity. Presumably you want your modification to be primarily expressed in brain cells and not in muscle or kidney cells (it doesn't do much good to improve brainpower if you also induce renal failure), let's have that be a gimme.

So how do you know that your modification is working? When we create things like bt corn or vitamin A wheat, you can just look at the corn plants or wheat seeds and see if they are producing lots of bt or vitamin A like you want. You could probably perform a few brain biopsies to determine if the target protein is being produced in the target quantities in the brain, but that probably induces a little brain damage while also not actually answering the question: if you were wrong in your guess, your genetic modification may work correctly (produce the right protein in the right parts of the body in the right quantity at the right time with no deleterious effects) and still not improve memory or spatial reasoning or conciseness or whatever else you care about.

So you've got to raise these babies until they're ten or twenty or thirty years old, studying them all the way, and try to guess whether they are smarter than they would be without the modifications. You need statistically meaningful sample sizes, with control groups of similar but unmodified genetics raised identically. Time-consuming and expensive!

Time-consuming and expensive kind of sums up the whole process. It doesn't make sense for individuals (you can't just say, "make my baby smarter"); it doesn't make sense for governments (there are cheaper ways to either make an entire population smarter or get your hands on a handful of geniuses).