Just another attempt at the old "black box" everybody envisioned in the '80s. I have not seen the new product, but I have seen virtually every other R2V product since the early '80s. Probably nothing new here either, polylines with hundred of vertices that are not needed, partial recognition of base elements, attempts at putting elements on a particular layer based on weight and all of the rest of the failed attempts of the past.
Maybe we should wait until AI is a reality for use on something like this. Besides, look at all of the Indian Engineers this would put out of work if this really worked.