How to Pass Turing Test?

Recently a chatbot fooled human judges in thinking it is a 13 year old boy for whom English is the second language. Here is the basic details of the structure of this system: it uses a hierarchy of patterns specified in XML structure for state estimation, and a set of rules that activate based on state, with knowledge that is stored in the database. Watch the original conference Chatbots 3.0 conference held in 2010:

Curse of Dimentionality

Classifiers that tend to model non-linear decision boundaries very accurately (e.g. neural networks, KNN classifiers, decision trees) do not generalize well and are prone to overfitting. Therefore, the dimensionality should be kept relatively low when these classifiers are used. If a classifier is used that generalizes easily (e.g. naive Bayesian, linear classifier), then the number of used features can be higher since the classifier itself is less expressive.

which features should be used?