So this 'Arc' protein has been in our genomes for quite some time. One of the two papers published simultaneously last month demonstrated their results on development and cognition in flies. [2,3]
The protein is the one that acts like the 'patches' on a soccer ball, where the inside is 'sticky' to RNA. And you can assemble a viral-like capsid  around that RNA, and seemingly transfer that RNA to another cell. A single RNA can produce a number of proteins over an extended period of time, while the transfer of proteins (for instance, in a lipid vesicle) is a much faster action, but much less long lasting or impactful (in terms of number of proteins in a location over time).
One very curious idea is that this is part of an as-yet undescribed mechanism for the delivery of nucleic acids into human cells. This process (of delivery) is also the missing piece to making most gene therapies useful. And in those terms, this particular discovery might be about as useful in our long-term bio-capabilities as the discovery of CRISPR's capabilities. The one really distinguishing feature is that this 'virus-like' protein is, literally, already humanized. Your immune system won't attack it - which is one of the primary concerns with using existing retroviruses or lentiviruses in a therapeutic setting.
If you want to play with the human Arc sequence and build your own delivery mechanisms, go for it:
This is also yet another way in which particular organisms have 'captured', and even completely subsumed other organisms for particular effect (see mitochondria). There are likely many more buried in our historical genomic archive.
In the octopus' case, the RNA is being altered post-compilation, in ways that are somehow critical to their neural development. And that kind of process is unusual, and seems to be one that the octopus really depends on for its 'smarts' - it's not just an evolutionary side-project.
While in the OP's case, the protein likely comes from an HIV-like infection that we've repurposed to help transmit compressed, ready-to-run code without having to transmit the large and short-lived binaries (proteins).
Kandel himself gives some wonderful lectures on this, easily found on YouTube.
It leads me to wonder how many other structures out there we have already described and unwittingly utilise, patterns re-discovered for purposes at another level of abstraction.