Standard SVM is a binary classifier; In SVM, you minimize the hinge loss; you focus on points that are most difficult to discriminate. Intuitively, you can think of SVM as an algorithm that finds a point A from one class that is closest to the points B and C from the second class.