I will throw out R just because it should be thrown out... there, out there, I mean.

I know, it's a terrible language, but it is a great collection of capabilities. Quintessential example of love/hate.

He mentions one: Julia. It's still young, so not as many libraries available. But it's a very fast language designed for scientific computing.

PDL (Perl Data Language) offers a great alternative - http://pdl.perl.org

I use R, Mathematica, Haskell, Julia, or my own CAS that I built.

What do you think of Mathematica? I've been toying with the idea of using it for sketching out ideas related to my computer graphics work, but it's a big investment. :/

I found it's utility primarily is dynamically controlled visualizations, for example, when I wanted to see how residues were distributed in the C^2 plane for a certain class of complex functions.

I'm not certain what sort of graphics work you are considering, but if it involves a lot of nonlinearity or higher level functions, it may be a good fit.

Julia is targeting the scientific computing community, and I'd love to see it succeed as it is a very well thought out language.

The language is in active development, although it is still early days with library support currently lacking (namely, a good plotting library).

Although there's a very sophisticated set of bindings that makes it almost trivial to call Julia code from python and vice verse, making it possible to use matplotlib instead of waiting for a solution written in Julia. Take a look at this [1] (skip to the ten minute mark for the PyCall stuff) presentation from this year's scipy conference.

[1] http://www.youtube.com/watch?v=Eb8CMuNKdJ0

Thanks, I wasn't aware of the PyCall library. This provides an incredible amount of interoperability between the two languages.