Pay close attention to the final few slides, they seem to outline a proposal for a set of changes that I think can be best described as "add strict variants of the various parallel/concurrent computing primitives haskell has to python (plus side effects tracking) so we can have python be natively good for HPC applications in the future"
It's worth mentioning the accompanying blog post too: <a href="http://khinsen.wordpress.com/2010/07/12/euroscipy-2010/" rel="nofollow">http://khinsen.wordpress.com/2010/07/12/euroscipy-2010/</a>