To add to what's been said about translation errors in the softcover edition, a very large portion of the equations in section V have been carried over incorrectly. Gauss' convention of writing indices in superscript instead of subscript seems to have confused the translator to no end, and indices on variables multiplied together were combined like exponents wherever possible. For instance, what was "t_i t_i" (variable t with an index of i, multiplied by itself) in the original Latin was converted to "t_2i" (variable t with an index of 2i). In the worst example, about HALF of the equations in articles 200-201 are incorrect as given. I haven't seen the revised edition, but I would hope this was corrected.
Apart from this, there are plenty of other, scattered typos, but these are for the most part easy to catch and tolerable. The original Latin version is available for free online on the Gottingen University library website, and it may be worthwhile to keep it handy to double check the equations.
As for the book itself, I cannot possibly recommend it highly enough. It was the first serious math book I ever read, and it inspired me to take up math as a profession. While the actual material covered in D.A. may be a bit outdated, it is a great window into the mind of one of the greatest mathematicians in history. It is a masterpiece, and worth the time (and $40) of any serious math student or professional.