First of all a Lie group is a group with elements in a differentiable manifold for which the group multiplication and the inversion of elements are differentiable mappings of the group to itself.
For simplicity let's consider matrix groups like SU(N), SO(N), etc. These are the common cases relevant for physics. Now you can define the Lie algebra of the group as the tangent space at the group identity. This is a vector space with matrices as elements (e.g., for SU(N) the Lie algebra consists of all antihermitean traceless matrices, in physics written as [itex]\mathrm{i} \alpha_k T^k[/itex] with [itex]T^k[/itex] hermitean).
Now let [itex]G[/itex] be the group and [itex]\mathcal{L}G[/itex]. Then it's easy to show that for each [itex]g \in G[/itex] the mappings
[tex]\text{Ad}_g: \; \mathcal{L}G \rightarrow \mathcal{L}G, \; \text{Ad}_g(t)=g t g^{-1}[/tex]
are a well-defined linear mapping which respects the Lie-algebra product (given by commutators of the Lie-algebra matrices in our case), and obeys the group-representation property,
[tex]\text{Ad}_{g_1} \circ \text{Ad}_{g_2}=\text{Ad}_{g_1 g_2},[/tex]
i.e., it's a linear representation of the Lie group on the Lie algebra, which is called the adjoint representation.
The nice thing is that it's constructed wholly from the Lie-group structure itself without introducing any other kind of elements.
In physics it appears in gauge symmetries, where the gauge fields are defined as affine connections of the group's Lie algebra to define covariant derivatives. Thus, under gauge transformations the gauge fields transform in the adjoint representation of the gauge group.