[en] For forty years, plasma turbulence theory has been based on notions of 'mixing length', 'diffusion', and other ideas from mean field theory. However, there is now a plethora of evidence, from both simulations and experiments, that plasma turbulence is strongly intermittent and that transport is bursty, and composed of a multiplicity of 'events' of differing scales. In this case, 'predictive modelling via transport coefficients' is an oxymoron. Thus, a fundamentally probabilistic treatment is necessary, i.e. one should aim to calculate probability distribution function (pdf) of flux, rather than an anomalous diffusion coefficient. Of particular interest in this approach is the tail of the pdf, which describes large 'events' or avalanches. In this paper, we discuss recent progress in formulating a probabilistic theory of drift wave turbulence and transport. We also report on progress in understanding the more general question of structure formation, particularly in the context of sheared flow. (author)