Last year at CoreOS Fest I spoke to a gentleman from Google who said they think limiting Tensorflow nodes processing power and/or trimming out some nodes entire

Last year at CoreOS Fest I spoke to a gentleman from Google who said they think limiting Tensorflow nodes processing power and/or trimming out some nodes entirely appears to cause the entire model to become more robust and perform better by spreading certain attributes (like edge detection) across multiple nodes. A type of holistic training, if you will. I see correlation with a limitation on visualizations. Regarding this particular article, members of this research group at New South Wales and related studies have been making a variety of claims regarding visualization over the past few years, including what appear to be loaded comments about Aphantasia. These comments include: “It’s a lack of the mind’s eye. My hunch is that there is something a bit different about their brains and that there may be a genetic component.” — Adam Zeman, University of Exeter “If there’s too much activity and noise, it’s like sketching on a dirty piece of paper: it’s hard to see the drawing.” — Joel Pearson, University of New South Wales “It could also alter moral judgement.” — Alice Klein, reporter for New Scientist while relating comments by Joel Pearson. I would present a counter argument that our visual imaginations are not something to be considered "severely limited". Instead, our imaginations are something that may be limited, either by intent (see Zen) or by circumstances (non-visualizer). If someone happens to be a great visualizer, then good for them, but don't tell me I'm limited. ;) Judging it to be something we lack and thus need more of is questionable. Perhaps the desired outcome is to have less of it, rather than more.