Great find on the article. Perhaps one day the judges for the Loebren Prize will be replaced by these "self-learning" robots. There is a discussion about this on Metaculus-
Last accident, Falcon Heavy had a big delay. Apparently SpaceX pulled engineering resources off of FH to work on F9. Given that the 2018 launch window is narrow, it's quite likely that Red Dragon won't launch until the 2020 launch window.
p.s. are you involved with Metaculus in some way? All of your recent postings appear to mention metaculus.com.
Another aspect of the project's challenge is funding. While technically plausible in principle to develop such system, the total funding needed is on the order of $5-10 Billion. There is a relevant discussion about this on [Metaculus](http://www.metaculus.com/questions/208/to-the-stars-1-will-t...), for those who are interested.