Boudica works on a small pool of controlled data that it is supplied with and is therefore far less likely to contain biases and inaccuracies than LLMs which learn from huge pools of varied data.
Boudica’s use of multiple separate thesaurus models and its probability matrices also ensure that only the optimum response is given to a user.