In my opinion, earthquakes can be predicted, but most of the time the prediction will fail because not enough is yet understood about earthquakes to be able to confidently do so.
The Chinese once predicted an earthquake in 1975 based on animal behavior observations. They evacuated the area and many lives were saved. However, the next year, a powerful earthquake occurred under Tangshan, China and flattened the city, taking with it over one quarter of a million lives. A few years ago in the mountains of Peru, cameras used to study animal behavior captured a mass exodus of the animals away from the mountain tops. Not long after that, a 7.0 magnitude earthquake occurred, after which the animals returned.
Not long ago, California scientists expected an approximately 6.0 magnitude earthquake to occur in Parkfield within a window of a few years. That was based on there being a fairly regular pattern or periodicity of similar sized earthquakes occurring in that area during many decades in the past. The prediction fell like a failed souffle.
Although some researchers in the field believe that fairly accurate predictions may some day be possible on a regular basis (once enough information is known to be able to do so), probably most earth scientists believe that the best that can be done is forecasting earthquakes rather than predicting them.
With prediction, one attempts to estimate when an earthquake will occur in a given area within a short period of time, usually within a few days time or possibly as much as a few years. With forecasting, the estimates are given with percentage probabilities of occurring within a few decades, usually 30 or 50 years into the future.
In order to forecast such quakes, the history of activity along the fault to be assessed must be studied to see a pattern of behavior. Possible future behavior is mapped out based on usually many centuries of past behavior; the further back in time one goes, the more accurate the future forecast is likely to be.