Google’s New Chatbot, Aerial Surveillance, and Rooting Robots: The Week’s Most Interesting ArXiv Papers
OneZero rounds up this week’s noteworthy A.I. and robotics papers
Researchers publish hundreds of new A.I. research papers every week on arXiv, an online repository for scientific papers. Here are a few that we found interesting, and think you might, too.
“Towards a Human-like Open-Domain Chatbot”
Google built a chatbot called Meena that was trained on 40 billion words of conversation captured from social media. Meena supposedly outperforms chatbots like Cleverbot and Microsoft’s Xiaoice by focusing on a simple metric: a crowd worker’s decision of whether its responses “makes sense.”
“A Deep Learning Approach to Grasping the Invisible”
A new approach to robotic manipulation allows a robot arm to root around, find a hidden object, and pick it up — much like the famed truffle hogs. This has obvious applications in an Amazon warehouse, where items might be dropped or buried in bins… but let’s focus on the truffle opportunities here.
“On Learning Vehicle Detection in Satellite Video”
Today in aerial surveillance: a new way to recognize cars from satellite imagery.
“InsectUp: Crowdsourcing Insect Observations to Assess Demographic Shifts and Improve Classification”
Bees, hornets, and mosquitoes can give scientists insight into the overall health of an ecosystem, and an international cadre of A.I. researchers have turned that idea into an app. The free app automatically recognizes different species of insect for the user, and uploads information of where they were sighted to researchers. It was an unexpected hit: users uploaded more than 44,000 images.
“Examining the Benefits of Capsule Neural Networks”
Geoff Hinton, often credited for pioneering the math behind neural networks used in modern A.I., debuted a new algorithmic model called capsule networks in 2017. (Read how they work here.) A new paper suggests that capsule networks are better at understanding minute differences between objects than the algorithms used today.
