The number “two” may appear to be very special in theoretical physics, but maybe it has had it’s day…
By this we mean that much of physics is described in terms of “binary objects”: Lie brackets, commutators, metrics, rank two curvature tensors, quadratic Lagrangians, two dimensional world sheets of strings, Poisson structures, symplectic two forms, Laplacians and I am sure the list goes on. However, it has become increasingly clear in recent years that “higher objects” play an important role in theoretical physics as well as modern geometry.
For example, it has become increasingly clear that n-aray generalisations of Lie algebras play a role in physics. The Sh Lie algebras of Stasheff and the (not completely unrelated) n-Lie algebras of Filipov are great examples here. In one form or another, they can be found behind the BV-antifield formalism, Zwiebach’s closed string field theory, Kontsevich’s deformation quantisation of Poisson manifolds, Nambu’s generalised mechanics and the Bagger–Lambert–Gustavsson (BLG) description of multiple
stacked M2 branes.
The last one has been of interest to me lately.
So, M-theory was introduced by Witten in 1995 as a non-perturbative unification of the various superstring theories. Here, the fundamental objects are not strings but extended membranes of dimension 2 and 5, the so called M2 and M5 branes. Since then progress has been slow. No-one really knows what M-theory is and there is no proper understanding of the dynamics of interacting branes.
Then in Bagger & Lambert  in 2006 and independently Gustavsson  in 2007 construct and effective action for the low energy description of a stack of two M2 branes. The novel feature here is that 3-Lie algebras play a role here. A 3-Lie algebra should be thought of as a “Lie algebra” but with a tribracket not a bibracket. Details should not worry us.
The theory has the fields take their values in a 3-Lie algebra and their is a novel gauge symmetry. However, the original BLG-model can be recast as a conventional gauge theory, the ABJM theory . So it starts to look that maybe 3-Lie algebras are some weird artificial artefact of M2 branes.
(There are many, many papers on the arXiv dealing with modifying the original BLG model. This usually can be understood in terms of 3-Lie algebra. I won’t say any more right now.)
But then very recently, Lambert & Papageorgakis  provided evidence that the effective description on M5 branes would also require 3-Lie algebras. However, they have not yet produced an action, which would be essential if the more or less standard methods of quantisation were to be applied.
This is fascinating. M-theory seems to be deeply tied to the theory of n-aray algebras, and in particular 3-aray algebras. There are may open questions here, both from a physics and mathematics point of view. In all it looks like n-aray algebra are here to stay.
 Ofer Aharony, Oren Bergman, Daniel Louis Jafferis, and Juan Maldacena. N=6 superconformal Chern-Simons-matter theories, M2-branes and their gravity duals. JHEP, 10:091, 2008.
 Jonathan Bagger and Neil Lambert. Modeling multiple M2’s. Phys. Rev., D75:045020, 2007.
 Andreas Gustavsson. Algebraic structures on parallel M2-branes. Nucl. Phys., B811:66–76, 2009.
 Neil Lambert and Constantinos Papageorgakis. Nonabelian (2,0) Tensor Multiplets and 3-algebras. 2010,