With the change in processor designs and the challenges that come with concurrent and parallel programming, we need to look for new programming paradigms. The need for correctness and the thinking in deterministic algorithms have governed the way we build software. However, with the change to parallel architectures as the norm, it seems that we need to focus on either guaranteed correctness or absolute performance. In this project, we investigate together with IBM Research and the Portland State University how programming models can work in such settings and how we can bound the expected errors to acceptable levels.