A BIT MORE PHYSICS
Garrett Lisi's attempted "theory of everything" was based around the idea that the pattern of symmetries of the laws of physics is that of the group E8. We don't see all those symmetries in the laws we know, but that isn't necessarily a problem because of a thing called spontaneous symmetry breaking. Imagine a big tank of water. Within that tank (at least away from its edges) no direction is different from any other -- the laws of physics don't have preferred directions, and neither does a bunch of water molecules bouncing off one another at random. This symmetry corresponds to conservation of angular momentum, so e.g. movements of water within the tank should obey that law. OK; now what if we reduce the temperature and the water freezes? Ice crystals, unlike liquid water, have a regular structure and they do have a preferred direction; so some things that happen in our tank full of ice may fail to conserve angular momentum.
(Let me be careful about this. Of course the overall laws of physics still satisfy that law, so if we consider the whole contents of the tank C of AM will still hold. But if we treat the ice in the tank as "background" and consider, I dunno, electrons moving around within the ice or something, their total angular momentum may fail to be constant, because of angle-dependent interactions with the ice.)
The same kind of thing can happen with the more "abstract" kinds of symmetry: as the universe cools -- and note that the universe now is much much cooler than shortly after the big bang -- it may settle into some less-symmetrical configuration (like our tank of ice), and the asymmetries in the universe may look to use like asymmetries in the laws, even though the real fundamental laws are symmetrical.
So, anyway, Lisi's theory is probably wrong, but the mere fact that E8 is bigger than the group of symmetries we actually observe in the universe isn't in itself good reason to disbelieve it. Now let's do (or at least gesture vaguely towards) some more mathematics, with the hope of ending up with some understanding of how the group E8 relates to the lattice E8.
LIE ALGEBRAS
To describe the relationship between the Lie group E8 and the lattice E8, I need to go via another kind of mathematical object called a Lie algebra. These are in some sense simpler cousins of the Lie groups, but they're harder to describe and harder to explain why you'd care about them. Still, here goes.
Let's look at one of the very simplest Lie groups. Take the ordinary euclidean plane -- imagine an infinitely large perfectly flat piece of paper. Fix a point in it; call it O. And consider rotations around O. These form a Lie group; it's sometimes called U(1), never mind why. We can describe it geometrically: it's a circle. (Pick some point that isn't O; call it P. Describe any rotation by saying where P goes to. That point is always on the circle through P with centre O. Conversely, if we pick such a point, there's a rotation that takes P there.)
A circle is pretty simple, but you know what's even simpler? A straight line. And if you look only at very small portions of the circle, they look a lot like very small portions of a straight line. This is the fundamental idea behind calculus: all nice things look linear if you look at a small enough scale. And linear things are easier to work with, mathematically. So: suppose we have a Lie group -- that is, a continuously-varying set of symmetries of something. Take a tiny region of it near to the origin; that is, near to the trivial symmetry that does nothing. This tiny region looks like a piece of euclidean space in some number of dimensions, but there's some extra structure on it that describes the relationships between those symmetries -- what happens when you do one followed by another.
What does that extra structure look like? It turns out that it's best described by what's called a bracket operation, meaning that given two elements x,y we have another written [x,y] which you should think of as meaning xy-yx. Imagine e.g. that x,y are actually matrices. Matrix multiplication usually doesn't commute, so generally xy-yx is not zero. Here, in some sense the multiplication operation corresponds to the "do one symmetry followed by another" operation on the original Lie group; that no longer exists in our "infinitesimal" view, but the bracket describing what changes when you switch the order of the symmetries still does. Kinda.
The bracket operation obeys a number of laws like the so-called Jacobi identity [x,[y,z]] + [y,[z,x]] + [z,[x,y]] = 0. The details aren't important right now. Anyway, the idea that we end up with when we follow this path and do attend to the details is that of a thing called a "Lie algebra", which describes what small pieces of a Lie group look like, and which in general looks like a euclidean space of some dimension together with a "bracket operation".
Corresponding to the Lie group called E8 there is, therefore, a Lie algebra also called E8.