This is why Octave is part of GNU, and why it's GPLed.
As for Scilab, they don't have quite the same aim as Octave. They don't aim to implement the same language so you can take your Matlab scripts away from Matlab, without change. They have a compatibility layer that does it partly, but there is still a need for implementing the same language. Imagine if the only way to run Javascript or Python was to sign away your rights to some company.
Freemat is a bit more comparable in aim to Octave, but it has a smaller contributor pool.
And why NumPy/SciPy will overtake Octave, if it hasn't already.
As long as people need to run Matlab scripts without Matlab, there will be a need for Octave.
A lot of people can get by just fine without FFTW or SuiteSparse (they aren't dependencies of everything written in the language, so don't belong in the standard library), and GMP is LGPL so not quite as problematic. But people do want to use a programming language to deploy applications, including proprietary commercial ones. Doing that in R or Octave isn't really possible because of the GPL, and that hurts the prospects of commercial backing (as does being pretty bad languages for developing software, but that's a separate matter).
Scilab is already open source, and GCC has a Fortran backend, so I don't see how those are relevant
"our entire purpose for working on Octave is to make sure people don't have to use Matlab."
That's a really negative way to put it. Isn't Scilab a viable Matlab alternative as well. Are you hoping to block people from using that as well? What about Intel's fortan compiler...
The point is, it's much more approachable to describe things in positive terms rather than negative when all you are essentially doing is offering a product in a not-ruled-by-monopoly market.