Talking about genetics in terms of bytes is a bit misleading.
One computer bit is equivalent to two of these: http://img.sparknotes.com/figures/7/749a4182b7527e44d289a612e420f40c/nucleotide.gif
Also, much DNA is not used, and then much DNA is used for multiple
purposes (by means of introns/exrons and protein domains).
I liked the PS3 advert :)
In the article, I think he addressed some of the points you raise: but CGAT *is* the system, though I guess in theory you could get odd results because your "bits" are such complex things, there might be interactions you didn't account for...
And yeah, he talks about all the junk DNA and what not... a simplistic "how many megabytes is the human genome" would be misleading both because it ignores all the junk there, as well as how sophisticated and multi-use the non-junk can be. But I think he pays at least lip service to that.