Using the ideas exposed in this blog and something more I am developing a program , an inductive deterministic inference engine
Inductive because I don’t use proof checking but only inductive inference.
Deterministic not in a strong sense , the result is always probably correct but the program is done over strong deterministic hypothesis of the world.
The main categorization of it is a sequence predictor so given a training set as bit string the objective is to find the next value of the bit string.
The objective is to predict correct results for very difficult bit strings normally identified as “random” by the other existing inference technique .
The interesting characteristics of the software are the possibility to define what is possible to call “distribution data” or “domain context” or “priori knowledge” defined with a huge set of existing/experimental/empirical real data and the usage of cellular automata evolution to make the inference.
The state of the art is a working release optimized for gpu system.