This paper provides the most complete evidence to date on the importance of monetary aggregates as a policy tool in an inflation forecasting experiment. Every possible definition of 'money' in the USA is being considered for the full data period (1960 -2006), in addition to two different approaches to constructing the benchmark asset, using the most sophisticated non-linear artificial intelligence techniques available, namely, recurrent neural networks, evolutionary strategies and kernel methods. Three top computer scientists in three top UK universities (Dr Peter Tino at the University of Birmingham, Dr Graham Kendall at the University of Nottingham and Dr Jonathan Tepper at Nottingham Trent University) are competing to find the best fitting US inflation forecasting models using their own specialist artificial intelligence techniques. Results will be evaluated using standard forecasting evaluation criteria and compared to forecasts from traditional econometric models produced by Dr Binner. This paper therefore addresses not only the most controversial questions in monetary economics -exactly how to construct monetary aggregates and to what level of aggregation, but also addresses the ever increasing role of artificial intelligence techniques in economics and how these methods can improve upon traditional econometric modelling techniques. Lessons learned from the experiment will have direct relevance for monetary policymakers around the world and econometricians/forecasters alike. Given the multidisciplinary nature of this work, the results will also add value to the existing knowledge of computer scientists in particular and more generally speaking, any scientist using artificial intelligence techniques.