As a physical phenomenon, earthquakes must be predictable to a certain degree. However, the problem is difficult, because the source volume inside the earth is inaccessible to direct observation and because the most important parameter, the stress level, cannot be measured directly. Also, seismology is such a young science that the cause of earthquakes was discovered in the 1960s only. Advanced seismograph networks as well as modern techniques to measure crustal deformations, such as the Global Positioning System (GPS) and the Synthetic Aperture Radar Interferometry technique (InSAR), have come on line only recently, and only in Japan are they deployed with the densities necessary for significant advances in the understanding of the rupture initiation process. In addition, no real program for earthquake prediction research exists in the United States, largely because funding agencies and peer reviewers shy away from a field in which unprofessional, but motivated individuals are active. Although claims of successful predictions are often not justified, a few correct predictions have been made. Most of these had time-windows of years, but some were accurate to days and allowed preparatory actions. To make significant progress, we must learn how to conduct rigorous science in a field where amateurs cannot be discouraged to venture. Leadership is necessary to raise the funding to an adequate level and to involve the best minds in this promising, potentially extremely rewarding, but controversial research topic.