This document summarizes a talk given on global warming predictions. It discusses evidence from ice core data in Greenland and lake ice cover in Madison, Wisconsin. It then outlines several prediction methods that could be used, including simple extrapolation, moving averages, trends, linear regression, autoregression, the all poles method, the method of analogs, and artificial neural networks. The document demonstrates an artificial neural network approach using 6 neurons and lags of up to 9 years to predict lake ice cover. It concludes that averaging or taking the median of multiple predictions from different models may provide the most accurate longer-term prediction.