Exactly, that's why I think we need to put logic and probability theory back into cutting edge ML. [1,2] are only early approaches that show potential direction

Exactly, that's why I think we need to put logic and probability theory back into cutting edge ML. [1,2] are only early approaches that show potential directions to achieve this. Deep learning is very useful, but only one piece of the whole AGI puzzle. Furthermore, many AI problems will benefit the generality of being formulated as a probabilistic program synthesis problem [3]. In this framework, lots of program semantics (~formal methods) concepts like abstract interpretation [4,5] might become very useful. They allow to explore huge program spaces very quickly. Lastly, Pearl's do calculus [6] is a good starting point closely related to [1]. [1] http://probmods.org/ http://probmods.org/ [2] http://pyro.ai/ http://pyro.ai/ [3] https://web.mit.edu/cocosci/Papers/Science-2015-Lake-1332-8.pdf https://web.mit.edu/cocosci/Papers/Science-2015-Lake-1332-8.... [4] http://www.concrete-semantics.org/concrete-semantics.pdf http://www.concrete-semantics.org/concrete-semantics.pdf [5] http://adam.chlipala.net/frap/frap_book.pdf http://adam.chlipala.net/frap/frap_book.pdf [6] http://bayes.cs.ucla.edu/BOOK-2K/causality2-epilogue.pdf http://bayes.cs.ucla.edu/BOOK-2K/causality2-epilogue.pdf