Robotics, Navigation, and Reinforcement Learning with Raia Hadsell

Google Cloud Platform Podcast

Episode | Podcast

Date: Wed, 18 Jul 2018 00:00:00 +0000

<p>On this episode of the podcast, <a href="https://twitter.com/Neurotic">Mark</a> and <a href="https://twitter.com/nyghtowl">Melanie</a> delve into the fascinating world of robotics and reinforcement learning. We discuss advances in the field, including how robots are learning to navigate new surroundings and how machine learning is helping us understand the human mind better.</p> <h5 id="raia-hadsell">Raia Hadsell</h5> <p><a href="https://twitter.com/raiahadsell">Raia Hadsell</a>, a senior research scientist at DeepMind, has worked on deep learning and robotics problems for the past 15 years. After completing a PhD at New York University, which featured a self-supervised deep learning vision system for a mobile robot, her research continued at Carnegie Mellon’s Robotics Institute and SRI International, and in early 2014 she joined DeepMind in London to develop artificial general intelligence. Her current research focuses on the challenge of interactive learning for AI agents and robots, including subjects such as neural memory for real world navigation and lifelong learning.</p> <h5 id="cool-things-of-the-week">Cool things of the week</h5> <ul> <li>AI Adventures How to Make a Data Science Project with Kaggle <a href="https://towardsdatascience.com/cooking-up-a-data-science-project-using-kaggle-datasets-and-kernels-fca2b678e268"> site</a></li> <li>Predict your future costs with Google Cloud Billing cost forecast <a href="https://cloudplatform.googleblog.com/2018/07/predict-your-future-costs-with-google-cloud-billing-cost-forecast.html"> blog</a> and <a href="https://cloud.google.com//billing/docs/how-to/reports#cost-forecast"> site</a></li> <li>Kaggle Competition Winning Solutions <a href="https://www.kaggle.com/sudalairajkumar/winning-solutions-of-kaggle-competitions/"> site</a> <ul> <li>Google Cloud Platform Podcast Episode 84: Kaggle with Wendy Kan <a href="https://www.gcppodcast.com/post/episode-84-kaggle-with-wendy-kan/"> podcast</a></li> </ul> </li> <li>Introducing Jib — build Java Docker images better <a href="https://cloudplatform.googleblog.com/2018/07/introducing-jib-build-java-docker-images-better.html"> blog</a> <ul> <li>Google Container Tools <a href="https://github.com/GoogleContainerTools/jib">site</a></li> </ul> </li> </ul> <h5 id="interview">Interview</h5> <ul> <li>Raia Hadsell <a href="http://raiahadsell.com/index.html">site</a></li> <li>Learning to Navigate Cities Without a Map <a href="https://arxiv.org/pdf/1804.00168.pdf">research paper</a> and <a href="https://deepmind.com/blog/learning-to-navigate-cities-without-a-map/"> blog</a></li> <li>Unsupervised Predictive Memory in a Goal-Directed Agent | MERLIN <a href="https://arxiv.org/pdf/1803.10760.pdf">research paper</a></li> <li>Nature: Vector-based navigation using grid-like representations in AI <a href="https://www.nature.com/articles/s41586-018-0102-6.epdf">research paper</a></li> <li>DeepMind has trained an AI to unlock the mysteries of your brain <a href="http://www.wired.co.uk/article/deepmind-newest-network-mimics-the-gps-cells-in-your-brain"> site</a></li> <li>Navigating with grid-like representations in artificial agents <a href="https://deepmind.com/blog/grid-cells/">blog</a></li> <li>DeepMind <a href="https://deepmind.com">site</a> and <a href="https://deepmind.com/blog/">blog</a></li> <li>Boston Dynamics <a href="https://www.bostondynamics.com">site</a></li> <li>Google Brain Robotics <a href="https://ai.google/research/teams/brain/robotics">site</a></li> <li>Transylvanian Machine Learning Summer School <a href="https://tmlss.ro">site</a></li> <li>IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures <a href="https://arxiv.org/pdf/1802.01561.pdf">research paper</a></li> <li>Edward Mozer - Grid Cells and the Brain’s Spatial Mapping System <a href="https://www.youtube.com/watch?v=Ekzec9e1Y4Y">video</a></li> <li>The Nobel Prize in Physiology or Medicine 2014 <a href="https://www.nobelprize.org/nobel_prizes/medicine/laureates/2014/">site</a></li> <li>TensorFlow <a href="https://www.tensorflow.org">site</a></li> </ul> <h5 id="question-of-the-week">Question of the week</h5> <p>How do you connect a Google Cloud Source repository to an existing Git repository? <a href="https://cloud.google.com/source-repositories/docs/connecting-hosted-repositories"> site</a> and <a href="http://blog.plataformatec.com.br/2013/05/how-to-properly-mirror-a-git-repository/"> blog</a></p> <h5 id="where-can-you-find-us-next">Where can you find us next?</h5> <p>We’ll both be at <a href="https://cloud.withgoogle.com/next18/sf/">Cloud NEXT!</a></p> <ul> <li>Mark will be talking about Agones <a href="https://cloudplatform.googleblog.com/2018/03/introducing-Agones-open-source-multiplayer-dedicated-game-server-hosting-built-on-Kubernetes.html"> blog</a></li> </ul> <p>Melanie will speak at <a href="http://pycon.ru/2018/en/">PyCon Russia</a> July 22nd</p>