In this talk I will discuss the factors that impact the successful application of machine learning to real-world problems. I will describe problems the generation/collection of representative training data and the computational challenges of large datasets. These problems and solutions will be explored in the context of four applications. The first application is to create a global map of the land cover of the Earth's surface from remotely sensed data (satellite data). These maps are used by climatologists and environmentalists to track changes in the Earth's ecosystem such as global warming and rain forest deforestation. Given labeled training samples of various sites on the Earth's surface, the goal is to induce a classifier that can be used to predict the land cover for unlabeled sites. The second application is to train a classifier based on data collected from an "artificial nose" to discriminate vapors. The "nose" is a collection of sensors that have different reactions to different vapors. The third application is to find anomalous in light-curves in the data collected by astrophysicists. The goal is to find light-curves that suggest a transiting planet. The final application is spam-filtering, where the goal is to discriminate spam from good email.