have 25 minutes help

interpreting visual dis

George has a dataset of eight data points with a single predictor variable. When he attempts linear regression MSE for his training error is 224.2; When George attempts cubic regression MSE for his training error is 67.5. On the other hand his ten-fold crossvalidation error for cubic regression is 323.6 and his cross validation error for linear regression is 293.3; What explains these results; which is the best model to use for this dataset