One thing I don't get is why when there are 4 bases in DNA, each base only represents 1 binary digit. The way I see it, with 4 bases, there is the potential for each base representing 2 binary digits. e.g. Adenine = 00; Thymine = 01; Cytosine = 10; Guanine = 11. As it is, it appears that both Adenine and Cytosine (a and c) represent 0 and Thymine and Guanine (T and G) both represent 1. In my mind, using each base to represent two binary digits should double the storage potential of DNA. Of course, there is probably a good reason for them not doing it this way; better minds than mine have clearly been working on this, they would no doubt have considered it. Perhaps it is as a protection against DNA mutation? (less chance of a mutation actually mattering; e.g. if an a base is mutated, there is a 1 in 3 chance it will still have the same value (by becoming a c base instead); if each base represented two digits, if any base is mutated, there is a 100% chance the value represented will be changed).
Also, seeing as only one strand actually matters (the other is just complimentary to it i.e. all 1s become 0s and all 0s become 1s) , and DNA is double stranded, it strikes me that it could be more efficient to use RNA, which is single stranded. Would (in my mind) half the mass of the storage medium. Again, I am sure there are numerous valid reasons for them not doing it this way.
Anyway, regardless of what I mentioned above, this is a rather interesting development. Although the responses in this thread make me doubt somewhat that people fully understand what this means; it's not a case of storing data inside a human beings DNA (although the article does mention that that is possible on a short term scale), it's about synthesising a substance that is essentially the same as DNA and using it as a storage medium, reading it with DNA sequencing hardware. As has probably been said, it's quite slow compared to hard drives, especially solid state and flash drives. It'd not be a case of storing movies or whatever on it, the read times would be far too slow. It'd be more useful as an archiving mechanism, allowing storage of an immense amount of data in a physically minute space, and for a very long time. Stuff that doesn't need accessed on a daily basis, but could potentially need referenced some day.
EDIT: As a side note, the OP mentions that the human genome (3 billion base pairs) can, with modern technology, be sequenced in a few hours. As a reference point, you should know that the amount of data stored by the Harvard scientists was about 2064150 times more. Now I guess you shouldn't have to sequence the entire DNA sample to find the data you need (the first 16 bases should give the "address" of that strand, letting you know if it's the right strand or not, and then you can move onto the next instead of doing all the subsequent bases), but it could still take a while I think. Although perhaps I have made some error in my calculations/thought process.
"Do a science"