Paul Graham is arguably our most eloquent statesmen regarding populist thinking about programming language design and implementation. In his essay on "The Hundred-Year Language", he ranges on a number of topics on the trail of thinking about a programming language from 100 years in the future. Java is an evolutionary dead in. Parallel programming won't really matter (but parallelism will). Perl actually has some good ideas.
There are two really great features of this essay. First, despite being a top notch Common Lisp hacker, he never comes across as a bitter Lisp burnout. Second, the discussion inspires thinking about a language 100 years in the future because it might actually be useful today. This essay should be required reading for any aspiring programming language designer.