Second part of the very good article about having more data vs. better algorithms

Very interesting thoughts about the fact that sometimes it pays more to add more data rather than fine-tuning the weights on your fancy machine-learning algorithm.