Infinite descent is a particular method of applying of mathematical induction. The basic form of the argument is to show that whenever you have an example of some property, that you can find a smaller example. Under the right circumstances, this allows you to infer that no example exists.
One common example of the right circumstances is when 'size' is measured by natural numbers.
The term 'infinite descent' comes from one way of justifying the method -- if an example exists, you can recursively construct an infinite sequence of new examples, each smaller than the previous one. This is a contradiction if there can only possibly be finitely many examples smaller than the original.
You often see infinite descent arguments written in a different form. (I'll consider sizes measured by natural numbers, for simplicity)
1. Assume there is at least one example of some property.
2. Then, there must exist a smallest example.
3. Construct a new example smaller than that one.
4. By contradiction, infer that there are no examples of that property.