GeNIe for “what-if” analyses: simple powerful graphical Belief/ Bayesian Networks. And its freeware!
Peter Norvig explains probability modelling. Peter gave a similar talk at Parc, covering how Google Translate works by overlapping segments of results and probability to improve the translation model. During beta the translated results can be voted on / and improved by real people. Combining big data sets and human tuning seems to simplify and improve algorithms results faster than waiting for an artificial intelligence.
