Top 10 reasons #deeplearning isn’t getting us to artificial general intelligence. A critique of deep learning, 5 years into its resurgence, by @garymarcus https://arxiv.org/abs/1801.00631
cool paper, but aren't recursive neural nets (mentioned as "clever tricks" in 3.3) exactly the sort of thing you are asking for in 3.3? the Lake and Baroni critique was only about recurrent nets, so doesn't (obviously) extend to recursive nets