After midnight in a sweltering room in Pasadena in July 1976, Viking Mars team members sat hunched around a bulky monotone computer monitor, tensely awaiting the first data from the world's first successful Mars probe lander, the only Mars lander ever specifically designed to detect life. Over the next weeks each of Viking's first life-detection experiments came back with a striking signature. As the data trickled back into the Space Operations Facility, it became clear that carbon dioxide was released when organic compounds were added to Martian soil, though not when the mixture was superheated. This was a life signature, and exactly what had happened with the experiment on Earth. When water was added to the soil, oxygen was released, just as on Earth. The remote probe, panning for life, had found its signature in its first two experiments.

We present MILABOT: a deep reinforcement learning chatbot developed by the Montreal Institute for Learning Algorithms (MILA) for the Amazon Alexa Prize competition. MILABOT is capable of conversing with humans on popular small talk topics through both speech and text. The system consists of an ensemble of natural language generation and retrieval models, including neural network and template-based models. By applying reinforcement learning to crowdsourced data and real-world user interactions, the system has been trained to select an appropriate response from the models in its ensemble. The system has been evaluated through A/B testing with real-world users, where it performed significantly better than other systems. The results highlight the potential of coupling ensemble systems with deep reinforcement learning as a fruitful path for developing real-world, open-domain conversational agents.

A set of decades-old algorithms has finally met with adequate data and computing power. Organizations around the world are using this artificial intelligence to make better decisions; government agencies are not far behind either. AI has shown with tremendous potential and unbelievable promise. It is but natural that AI be applied to automate workflows based on something each citizen uses everyday -- language. Majors companies like IBM, Amazon and Microsoft, as well as upstarts like ours, Coseer, are investing in AI for language. The obvious course is to start with algorithms that have been so successful elsewhere.

In the summer of 1957, the Earth stood witness as a meteorite cratered in rural Pennsylvania, bringing with it a people-eating plague never seen: an alien amoeba with the taste for human flesh. While we had Steve McQueen around for the first invasion, humanity is now defended against microbial marauders from outer space by NASA and its international counterparts. Biological contamination goes both ways, mind you. Just as important as keeping extraterrestrial organisms from reaching the surface (aka "backward contamination") is ensuring that our planetary probes carry as few microbial hitchhikers from Earth as possible ("forward contamination"). To that end, in 1958, the U.S. National Academy of Sciences (NAS) issued a decree urging "that scientists plan lunar and planetary studies with great care and deep concern so that initial operations do not compromise and make impossible forever after critical scientific experiments."

Humanity has been on the road for a very long time--from the beginning, when each individual had to collect sufficient food to survive every single day--to the point where we invented agriculture. At that point, we moved from 99% survival and 1% reproduction to a brand new model. Growing food marked the introduction of leisure. Since then, every step in our evolution has proceeded along the lines of doing more and more with less and less. You might recall the 1899 story of Charles H. Duell, Commissioner of the U.S. Patent Office, lobbying President McKinley for its closure, claiming that "everything had already been invented."

"It has always been our ambition to get inside that white space, and now we are there the space can no longer be blank," wrote the polar explorer Captain Scott, on crossing the 80th parallel of the Antarctic continent for the first time in 1902. Fast-forward more than a century--and the deep ocean floor around Antarctica still offers a "white space", beyond the reach of scuba divers, only partially mapped in detail by sonar from ships and seldom surveyed by robotic vehicles. So I jumped at the chance to join a team from the BBC on an expedition to the Antarctic Peninsula for Blue Planet II, to help them as a scientific guide. Thanks to the crew of the research ship Alucia, we dived in minisubmarines to 1km deep in the Antarctic for the first time. And while we didn't face anything like the physical hardships endured by early polar explorers on land, those dives did give us the opportunity for some unique science.

We present MILABOT: a deep reinforcement learning chatbot developed by the Montreal Institute for Learning Algorithms (MILA) for the Amazon Alexa Prize competition. MILABOT is capable of conversing with humans on popular small talk topics through both speech and text. The system consists of an ensemble of natural language generation and retrieval models, including template-based models, bag-of-words models, sequence-to-sequence neural network and latent variable neural network models. By applying reinforcement learning to crowdsourced data and real-world user interactions, the system has been trained to select an appropriate response from the models in its ensemble. The system has been evaluated through A/B testing with real-world users, where it performed significantly better than many competing systems. Due to its machine learning architecture, the system is likely to improve with additional data.

UC San Diego is creating an outdoor site where it can test fly unmanned aerial vehicles, which are rapidly coming into common use by everyone from police investigating crime scenes to scientists looking for archaeological remains. The aerodrome will be a net cage that will be 30 feet high and roughly 50 feet long and wide, making it similar to a facility that's being built at the University of Michigan, a leader in drone research. San Diego chipmaker Qualcomm gave UC San Diego $200,000 to create the flight center, which is meant to help promote the school's quickly expanding research in robotic systems. The campus recently announced that it will begin testing driverless vehicles on university roads next year, using golf carts to deliver packages. The research will begin about the time that engineers start to extensively use the aerodrome.

Is there a global theory for the shapes of fish? But for most of the history of biology, it's not the kind of thing anyone would ever have asked. Stephen Wolfram is the creator of Mathematica, Wolfram Alpha and the Wolfram Language; the author of A New Kind of Science; and the founder and CEO of Wolfram Research. Sign up to get Backchannel's weekly newsletter, and follow us on Facebook, Twitter, and Instagram. And it's now 100 years since D'Arcy Thompson published the first edition of his magnum opus On Growth and Form--and tried to use ideas from mathematics and physics to discuss global questions of biological growth and form. Stretch one kind of fish, and it looks like another. Yes, without constraints on how you stretch. It's not quite clear what this is telling one, and I don't think it's much. But just to ask the question is interesting, and On Growth and Form is full of interesting questions--together with all manner of curious and interesting answers. D'Arcy Thompson was in many ways a quintessential British Victorian academic, steeped in the classics, and writing books with titles like A Glossary of Greek Fishes (i.e. But he was also a diligent natural scientist, and he became a serious enthusiast of mathematics and physics. And where Aristotle (whom Thompson had translated) used plain language, with perhaps a dash of logic, to try to describe the natural world, Thompson tried to use the language of mathematics and physics.

With the recent renaissance of deep convolution neural networks, encouraging breakthroughs have been achieved on the supervised recognition tasks, where each class has sufficient training data and fully annotated training data. However, to scale the recognition to a large number of classes with few or now training samples for each class remains an unsolved problem. One approach to scaling up the recognition is to develop models capable of recognizing unseen categories without any training instances, or zero-shot recognition/ learning. This article provides a comprehensive review of existing zero-shot recognition techniques covering various aspects ranging from representations of models, and from datasets and evaluation settings. We also overview related recognition tasks including one-shot and open set recognition which can be used as natural extensions of zero-shot recognition when limited number of class samples become available or when zero-shot recognition is implemented in a real-world setting. Importantly, we highlight the limitations of existing approaches and point out future research directions in this existing new research area.