http://www.technologyreview.com/s/602094/ais-language-proble... by Will Knight on 2016-08-09 addresses the same problem of NLU (Natural Language Understanding). I have long been working on True AI programs that think with automated reasoning (inference) in English, German and Russian. My webpage at
http://ai.neocities.org/SOTA.html invites AI Projects to expand upon my
http://github.com/PriorArt/AGI/wiki/MindGrid software by assigning their own staff members to work on specific AI mind-modules. I feel that I have been able to solve the AI language problem because I first spent thirteen years creating at
http://mind.sourceforge.net/theory5.html a linguistic Theory of Mind for concept-based artificial intelligence. My
http://www.linkedin.com/in/mentifex profile indicates that creating the first True AI is not a team project or the work of a committee, because some one ambitious intellect must grapple and wrestle with the ordeal of initial creation, after which teams and IT departments and commercial AGI Projects can parcel out the individual enhancement assignments for the further development of an Artificial General Intelligence. At
http://groups.google.com/d/msg/de.sci.informatik.ki/4ikvL8rO... I discuss in German the difference between a simple, primitive, concept-based AI which operates on the assumption that all verbal inputs to it are true statements and not lies, so that the AI may think and reason without suspicion, and on the other hand a conscious AI which bears the onerous burden of judging whether statements made to it are truth or lies.