a16z Podcast: The End (and Beginning) of Programming. Has Peter Levine in conversation with GitHub co-founder and former CEO Chris Wanstrath. podcast 
Declarative rather than programmatic development systems has been my career focus. Because SMEs (small to midsize enterprises) need tools to do business w/o spending time on programming. Very few firms can product productivity software for volume markets that people intuitively find useful. Microsoft Excel is standard. We can now go beyond Excel but are not not because we are aiming productivity tools at if then else developers not data developers.
Fei-Fei Li.
To grok what I'm looking to convey I had to turn to Fei-Fei Li Professor of Computer Science, Stanford University; Chief Scientist of AI/ML, Google Cloud. She gave a great explanation of how deep learning actually improves, by studying new babies and how they learn so quickly using vision. In nutshell Google's AI applies the methods of closed loop feedback by conducting many experiments, correct tests are reward and incorrect tests demoted. Thus an autonomous feedback explores data and develops a model that can be deployed at scale w/o programming.
YOUTUBE UvauKbNiX7k Fei-Fei Li.