In this talk I want to explain as simply as I can the reason why some stochastic partial differential equations, like e.g. KPZ, need a procedure called renormalisation. This consists in a modification of the equation by adding new non-linearities that typically contain diverging constants. 

This is strongly related to a similar phenomenon in quantum field theory and its explanation in the context of SPDEs is probably more intuitive. 

Unfortunately a complete description of the procedure requires an algebraic machinery that most probabilists prefer to avoid. However the beauty of this technique resides also in the deep interplay between analysis, algebra and probability.