Currently it's one week till this years NIPS deadline and there still is a lot of work to do, as always. I just wanted to quickly point you to a post I've written on our homepage mloss.org, where I try to make the point that you could actually do useful things using a modern scripting language like python or ruby for building machine learning toolboxes. At least, you'll have more powerful modeling possibilities than in matlab.
And then, jruby-1.1.2 has been released. Beside other things they have managed to reduce the startup time from about 1.2s (on my machine) to 0.4s, which is pretty great. I'm planning to write an article about how useful the jruby/java integration is, in particular for machine learning, but that will be after NIPS, I guess... .