Write it well, write it efficiently, write it as though the underlying hardware isn’t going to get any faster. Because it isn’t
A discussion with Jon Oberlander of the University of Edinburgh about machine learning and enterprise NLP