Everything is heritable:

• "Assortative mating and differential fertility by phenotype and genotype across the 20th century", Conley et al 2016 (Dysgenics found in the USA, 1920-1955: education/intelligence polygenic scores decrease over time, inversely correlate with offspring - education/intelligence polygenic score is dropping 0.0029 SDs per year, or 0.087SD over the 30 years covered, implying that the USA genetic potential has dropped 0.28SD since 1920 or <4 IQ points.)
• "The Genetics of Success: How Single-Nucleotide Polymorphisms Associated With Educational Attainment Relate to Life-Course Development", Belsky et al 2016 (excerpts)
• "The Genome Project-Write", Boeke et al 2016 (media; commentary)
• "Why We're Different: A Conversation with Robert Plomin" (Plomin takes a victory lap: "It might be hard for people to believe this, but forty years ago it was dangerous to talk about genetic influence in psychology.")
• "Demonstrating the validity of twin research in criminology", Barnes et al 2014
• "Pleiotropy across academic subjects at the end of compulsory education", Rimfeld et al 2015
• "Genetics affects choice of academic subjects as well as achievement", Rimfeld et al 2016
• "The Genetic Basis of Mendelian Phenotypes: Discoveries, Challenges, and Opportunities", Chong et al 2015 (But enough about common variants & complex traits - how have we been doing in finding the causes of genetic diseases like severe Mendelian diseases? Very well.)
• "A Prospective Study of Sudden Cardiac Death among Children and Young Adults", Bagnall et al 2016 (Rare genetic mutations implicated in >13% of unexplained cardiac deaths. Makes one wonder how much of the long tail of deaths are due to rare variants, and how much of a benefit there would be to erasing the ~80k mutations everyone carries... (Population genetics rule: rare variants are more harmful than common variants; and 1 harmful mutation, 1 reproductive death to purge it from the population.) Very cool application of genetics.)

• genetics of breast size (commentary):

  • Wade et al 2010 "Body mass index and breast size in women: same or different genes?"
  • "Genetic variants associated with breast size also influence breast cancer risk", Eriksson et al 2012
  • "Large-scale genotyping identifies a new locus at 22q13.2 associated with female breast size", Li et al 2013

Politics/religion:

• "Wealth, Health, and Child Development: Evidence from Administrative Data on Swedish Lottery Players", Cesarini et al 2016
• "Okhrana: The Paris Operations of the Russian Imperial Police" (The back and forth secret war of the Okhrana with the myriads of Russian revolutionaries across Europe, documented by the complete archives of the Paris Okhrana office smuggled to America after the Russian Revolution. When you see how easily and thoroughly the Okhrana had infiltrated the Russian revolutionaries, you start to see why the Communist leadership would be extraordinarily paranoid about spies - but also that the revolutionaries were, well before the Revolution, generally highly nasty folks; many of the mentioned revolutionaries would be summarily executed by their comrades.)
• "The Unbelievable Tale of Jesus's Wife", King's response (The Gospel of Jesus's Wife is probably a modern forgery. But it gets weirder. And kinkier. Forgeries like this always raise troubling issues about religious scriptures: if this forgery had been kept in private collections for another century before becoming known, in all likelihood, most of the damning evidence would either have disappeared or become inaccessible, and all that would be left is a few worries over the appearance. Most scriptures have even more vexed provenances than does the Gospel of Jesus's Wife, with blackouts of centuries not uncommon, and known destruction of variants (eg the well-known destruction of all variants of the Koran). Of course, you might think, who would dare counterfeit the Word of God Himself? Yet, humans are strange and inscrutable and can talk themselves into anything - why did Fritz do it? Probably even he doesn't really know. Who knows how many Fritzes there have been throughout history...)
• "My Four Months as a Prison Guard"

AI:

• "Unifying Count-Based Exploration and Intrinsic Motivation", Bellemare et al 2016 (Video; big exploration improvement for DQN-like agents: where DQN can only get to two rooms in Montezuma's Revenge, because it takes actions mostly at random and it is unlikely that it will randomly do the 15 or 20 exact moves which will get it a reward, this version, with a way of measuring novelty and exploring novel states until a reward is found, can make it to 15 rooms!)
• "Improved Techniques for Training GANs", Salimans et al 2016 (Even better image generation at 128px now. Synthesizing art can't be far off.)
• "InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets", Chen et al 2016 (are GANs and VAEs giving us a breakthrough in unsupervised learning?)
• "Early Visual Concept Learning with Unsupervised Deep Learning", Higgins et al 2016:
• "Progressive Neural Networks", Rusu et al 2016 (NN factored to build different sub-nets for different tasks while allowing later learned tasks to draw on earlier-trained subnets. See also "Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning", Parisotto et al 2015.)
• "Towards an integration of deep learning and neuroscience", Marblestone et al 2016
• "DeepMath - Deep Sequence Models for Premise Selection", Alemi et al 2016
• "Convolutional Sketch Inversion", G. et al 2016 (facial sketch->illustration)
• "Learning to learn by gradient descent by gradient descent", Andrychowicz et al 2016 (Replacing simple SGD, Adadelta, RMSprop etc by a very small RNN. Probably never practical but still amusing.)

Statistics/meta-science:

• "Generalized Network Psychometrics: Combining Network and Latent Variable Models", Epskamp et al 2016
• "Statistically Controlling for Confounding Constructs Is Harder than You Think", Westfall & Yarkoni 2016 (This is part of why results in sociology/epidemiology/psychology are so unreliable: not only do they usually not control for genetics at all, they don't even control for the things they think they control for. You have not controlled for SES by throwing in a discretized income variable measured in one year plus a discretized college degree variable. Variables which correlate with or predict some outcome such as poverty, may be doing no more than correcting some measurement error (frequently, due to the heavy genetic loading of most outcomes, correcting the omission of genetic information). This is why within-family designs are desirable even without worries about genetics: they hold constant shared-environment factors so you don't need to measure or model them.)
• "Linear Programming: How Does It Work?"