The sequencing of the whole animal genomes such as human and mouse has allowed quantitation of a long-established concept: that our DNA is full of repetititive sequences (Link here ; scroll down to subsection titled "Repeat content of the human genome.")This DNA does not generally code for any protein, and has been referred to as "junk DNA." In the case of humans, much of this DNA represents copies of "transposable elements" which are pieces of DNA flanked by short specialized sequences which can be "transposed," i.e. cut and pasted, copied and shuffled. These transposable elements also sometimes pick up other stretches of RNA and take them along to a new location. This might be how GLUD2 originated, for example.
A recent paper by Liu et al. shows that these transposable element derived sequences have gotten more and more enriched among primates of the human lineage. That is, the human genome is much larger than the lemur genome, and probably even expanded relative to the chimpanzee's. In the end, the human genome is at least 30% transposable element-derived sequences! That is, the shuffling and copying has been ongoing over the evolutionary interval that led to our ancestors.
Since picking a piece of DNA out of its context and putting it somewhere else is analagous to pulling out and replacing parts in a machine, it is thought that most of this activity is parasitic at best to the host genome. (In fact some human cancers may arise from somatic recombination events near these transposable elements-the link is unfortunately behind a subscription wall. ) Nevertheless, it appears that genome transposable elements have been not only tolerated, but selected for, during human evolution. Not junk at all! But for now, the selective advantages gained can only be guessed.