Probabilistic generative models can be used for compression, denoising, inpainting, texture synthesis, semi-supervised learning, unsupervised feature learning, and other tasks. Given this wide range of applications, it is not surprising that a lot of heterogeneity exists in the way these models are formulated, trained, and evaluated. As a consequence, direct comparison between models is often difficult. This article reviews mostly known but often underappreciated properties relating to the evaluation and interpretation of generative models with a focus on image models. In particular, we show that three of the currently most commonly used criteria---average log-likelihood, Parzen window estimates, and visual fidelity of samples---are largely independent of each other when the data is high-dimensional. Good performance with respect to one criterion therefore need not imply good performance with respect to the other criteria. Our results show that extrapolation from one criterion to another is not warranted and generative models need to be evaluated directly with respect to the application(s) they were intended for. In addition, we provide examples demonstrating that Parzen window estimates should generally be avoided.

One key buzzword these days is "algorithm," which technically means any computational formula but which has come to mean a formula that predicts our behavior. Amazon and Netflix have algorithms that predict what books a user is likely to want to read or what movies and TV shows he or she is likely to want to watch. Facebook has an algorithm that predicts the news a user is likely to want. Dating sites like Match.com and OkCupid use algorithms to predict with whom we would fall in love. Google, with the most famous algorithm of all, predicts what we want when we type a search term.

As the average age of farmers globally creeps higher and retirement looms, Japan has a solution: robots and driver-less tractors. The Group-of-Seven agriculture ministers meet in Japan's northern prefecture of Niigata this weekend for the first time in seven years to discuss how to meet increasing food demand as aging farmers retire without successors. With the average age of Japanese farmers now 67, Agriculture Minister Hiroshi Moriyama will outline his idea of replacing retiring growers with Japanese-developed autonomous tractors and backpack-carried robots. U.S. Agriculture Secretary Tom Vilsack has warned that left unchecked, aging farmers could threaten the ability to produce the food the world needs. The average age of growers in developed countries is now about 60, according to the United Nations.

A trip to Iceland wouldn't be complete without a dip in the Blue Lagoon, a man-made geothermal pool on Reykjanes peninsula. A trip to Iceland wouldn't be complete without a dip in the Blue Lagoon, a man-made geothermal pool on Reykjanes peninsula. The premise is simple: Correspondents, editors and producers from our newsroom share the pieces that have kept them reading, using the #NPRreads hashtag. Each weekend, we highlight some of the best stories. You have storms, you have darkness, but the pool is a place to find yourself again.

Peter Singer is a professor of bioethics at Princeton University and Laureate Professor at the University of Melbourne His books include Animal Liberation, The Life You Can Save, The Most Good You Can Do, and, most recently, Famine, Affluence and Morality. Last month, AlphaGo, a computer program specially designed to play the game Go, caused shock waves among aficionados when it defeated Lee Sedol, one of the world's top-ranked professional players, winning a five-game tournament by a score of 4-1. Why, you may ask, is that news? Twenty years have passed since the IBM computer Deep Blue defeated world chess champion Garry Kasparov and we all know computers have improved since then. But Deep Blue won through sheer computing power, using its ability to calculate the outcomes of more moves to a deeper level than even a world champion can.

Sure, robots and intelligent machines are likely to replace jobs in the not so distant future. The situation, though, isn't as dire as some would have you believe, according to Tom Davenport, co-author of Only Humans Need Apply: Winners and Losers in the Age of Smart Machines. The book is due out in May. Instead of stealing humans' jobs, artificial intelligent systems and robotics will help many people do their jobs better. "We have a new generation of technologies and we need to work with them if we're going to be productive and effective," Davenport said in an interview.

Research funded by the National Science Foundation may have found an ingenious solution to poaching: applying game theory and computer science to real-life situations. One of the biggest factors in why there are so many endangered animals today is poaching – a centuries-old problem. The dwindling tiger population is one of the most glaring examples of this. Whether for sport, medicine, pelts or other body parts, poaching remains a huge threat to wildlife. Patrols have long been the most direct form of human intervention in wildlife protection.

Computer chip giant Nvidia has developed a "miracle" chip that is expected to significantly accelerate breakthroughs in artificial intelligence research. Nvidia's Tesla P100 chip crams in 15 billion transistors within its 610-square-millimeter frame, around three-times more than most processors or graphics chips on the market. According to the company's CEO, this makes the Tesla P100 the largest computer chip ever made. "Three years ago, we dedicated ourselves on the single greatest endeavour in the history of our company," Jen-Hsun Huang, CEO of Nvidia, said at the GPU Technology Conference earlier this month. "We decided to be all in on AI. For the first time, we would design [a chip] that is dedicated to this field of work. Dedicated to accelerating AI; dedicated to accelerating deep learning. "I think we are going to realize looking back that one of the biggest things that ever happened is AI." The Tesla P100 is the product of around 2.5 billion worth of research and development at the hands of thousands of computer engineers. The Tesla P100 chip contains more than 15 billion transistors and is described by Nvidia's CEO as "a beast of a machine." "The odds of this working at all is approximately zero," said Huang. "We are changing so many things in one project.

In the race to develop self-driving cars, the United States and Europe lead in technology, but China is coming up fast in the outside lane with a regulatory structure that could put it ahead in the popular adoption of autonomous cars on its highways and city streets.

Venture capitalists pride themselves on their ability to pick winning ideas and winning people. But could artificial intelligence do a better job? Founders Factory, a U.K. startup accelerator, has developed an AI platform that identifies high-potential entrepreneurs. The hope is to avoid the unconscious bias that normally privileges some demographic groups and backgrounds, and prevents others from getting ahead. "I was interested in getting around the bias of selection, that if you've gone to a good school or university, you probably have a good network and a good chance of doing fairly well," says Tom Bowles, who created the software.