Most of the things you noted were mentioned in the article. Yes, it’s true that when we’re looking at DNA, we’re looking at a product of more then 3.5 billion years of selection and it’s going to be pretty efficient and well adapted as far as biology is concerned.

However, it’s the redundancies and the necessity of dealing with mutations that make it an inefficient design. It’s not about how many bases you choose, but how you encode the data you need and how you protect it. A good designer should’ve made a simple encoding that matches the 20 amino acids one to one and protected the genome from mutations by backups which replace altered DNA.

Funny enough though, some organisms can actually do something like that with single strand annealing. But that process is less of a backup and more of a deletion and substitution.

We have four nitrogenous bases for evolutionary reasons – they are, in fact a consequence of chemistry, and the double stranded DNA structure. Because of chemistry, we have the purine/pyrimidine base combinations, and because of the double stranded DNA structure, the unit of information becomes a base pair, rather than a base. This system requires a minimum of 4 bases to work.

If you are working with 4 bases, then the triplet code is a must. If you only use 2 bases instead of 3 out of the possible 4, you have 4^2 = 16 possible combinations. This is not enough to code for 20 amino acids, not to mention start and stop codons. So we need a triplet code, which means 4^3 = 64 combinations are available. This allows for some redundancy as well as start/stop codons, but also this is the minimum number necessary. So it’s not “inefficient” in that sense – you really can’t do the job with 2, you NEED 3. Not to mention that this redundancy offers protection against mutations which can kill you.

You could argue “well, if I were the designer, I wouldn’t go with 4 bases to begin with, I’d pick fewer”. Then you would get into much deeper details, none of which really make an argument for efficiency. The double stranded DNA structure protects against mutations – an “open” single strand is much less stable, so this is not a waste or “inefficiency”. You can’t just pick arbitrary molecules as bases either – the purine and pyrimidine bases are well suited to the task because of their high affinities with each other, because of the way they integrate with the sugar/phosphate backbone, because of how that allows the molecule to coil, so you can have a meter long molecule packed inside a tiny cell nucleus, and yet have any part of the molecule instantly available for transcription without uncoiling the whole thing.

In short, none of this is inefficient – unless by efficiency you mean “gimme ECC memory and no other constraints other than those imposed by these silicon computers and programming languages I am used to”. Those computers and languages also have their constraints, which have to do with the nature of the electronics they run on, and the history of how we developed computing hardware and software. In the same way, biology has its own constraints. It is not “inefficient” just because those two sets of constraints are different.

Just as the creationist/ID nuts go overboard with their design argument, I think some supposedly scientific people go overboard with their “there is no design – this is all a jury rigged inefficient PoS” argument. There are indeed some cases where biology has produced some strange results, that we as human engineers find kludgy. But much more often, when we call something “inefficient” the real problem is that we don’t understand the system well enough.

I have no problem comparing DNA to computers, since many people are familiar with computers. I would take the trouble to point out some of the facts I have mentioned in this reply, however, so that people realize that analogies only approximate the truth to help explain some specific facet of the comparison, and the reality is of course much more complex.