They did a great job with what they've put together so far (and if I would've had reasonable expectations, I'd be happy) but I felt he hasn't addressed some of the toughest problems yet:

handling proprietary data

what if scenarios

training (ie new scientific discoveries require manual intervention)