I think that this 79-page presentation by Caltech's Anton Kapustin is both insightful and entertaining.
If you are looking for the "previous slide" button, you may achieve this action simply by clicking 78 times. Click once for the "next slide".
If you have any problems with the embedded Flash version of the talk [click for full screen] above, download Anton's PowerPoint file which you may display using a Microsoft Office viewer or an OpenOffice or a LibreOffice or a Chrome extension or Google Docs or in many other ways.
Spoilers are below.
Anton describes the relationship between mathematics and physics, mathematicians and physicists, and so on. He focuses on the noncommutative character of algebras of observables in quantum mechanics. No mathematician really believed the Feynman's path integral and no physicist was interested in the mathematics by people like Grothendieck.
However, some smart opportunists in the middle – for example, Maxim Kontsevich – were able to derive interesting results (from mathematicians' viewpoint) using the path integral methods applied to the Poisson manifolds. And it wasn't just some lame undergraduate Feynman path integral that was needed. It was the stringy path integral that may be formulated using an associative product.
Hat tip: John Preskill, Twitter