A radical new technique lets AI learn with practically no data is the next article in technology review. It's talking about this new paper from the University of Waterloo in Ontario that suggests essentially a quite unsupervised approach to representing data at a more global scale. And certainly we are trending toward massive, massive, massive datasets in NLP and in computer vision. So it seems like if we can compress those datasets in some way, while still being able to train to the same level, this could be pretty impactful as it might take.