Questions that the panel will focus on include:
 

- What does it mean to "understand" the brain, and especially the ventral stream, in the era of deep networks?  Can we come up with better, richer ways of comparing representations in models and brains, which are more revealing of how brains actually compute?

- How should we understand the top-down and recurrent processing that occurs in the vision, and its interaction with feedforward bottom up processing?  What are the best ways to compare integrated models of bottom-up, top-down and recurrent processing to neural data, in humans and nonhuman primates?

- To what extent can we understand vision as inverting a generative model?  Where is the generative model in the brain, and how is it used, in online perception, learning, or imagination?

- How can we scale the successes we've had in understanding object recognition in the brain, to other aspects of vision, other perceptual modalities, and beyond perception?