We recently published a paper on Neuroscience and Machine Learning (ML).
https://www.cell.com/neuron/pdf/S0896-6273(19)30740-8.pdf
In it we highlight some ways that ML utterly fails but brains succeed. This is a question of generalization, specifically how a network learns properties in data that generalizes well. In writing this it became helpful to distinguish between model bias and inductive bias. So I wrote a short blog post about it here:
http://xaqlab.com/2019/09/18/inductive-bias/
Thoughts are welcome.