Read your article. Not sure about most of it as I’m not a PhD in AI or data science. I’m intrigued by the ability to build AI from “descriptive, contextual models” instead of big data. If the resulting systems are accurate, then it is indeed a big leap forward.

On NLP however (which I’ve spent the first half of 2018 quite immersed in because I have one toe in the translation industry), my understanding from the experts is full human level NLP AI will be not possible for at least another 25–30 years. Speech recognition and synthesis may be there, but understanding humor, slangs, puns and autonomous multi-lingual translation with high accuracy is still a long time off.