Computational models of an ongoing objective process, as in weather forecasting, must continually assimilate new observational data as they run. Both "truth" and "model" are chaotic systems that thus synchronize through a limited exchange of information in one direction - a phenomenon that can be characterized as machine perception. A recent suggestion has been to envision the fusion of different models analogously, as 3-way synchronization of the different models with reality. This phenomenon may be useful for improving climate projection by combining a few different models that differ in regard to the magnitude of global warming and regional predictions. Several machine learning approaches have been proposed to train the connections linking corresponding variables in the different models. Stochastic approaches can avoid non-global local optima, but it seems likely that an intelligent "expert system" approach would improve the supermodel.