I haven't actually tried this new Ruby library but numpy/scipy and even more so Matlab are widely used in the ML/AI community. You can find tons of sample code, books, blogs etc that use these tools so it's a lot easier to get started. Numpy is also pretty fast so you may actually get away with pushing your exploratory code into production depending on your scale.

Also, the graphing capabilities of Numpy or Octave can be extremely helpful in exploring your data and figuring out what your code is doing.