robotics Cassie Robot Learns to Walk Autonomously Using AI Researchers use reinforcement learning to teach a bipedal robot named Cassie to walk without hand-tuned programming. This approach overcomes the sim-to-real gap and brings robots closer to navigating unpredictable human environments.
reinforcement learning Reinforcement Learning Framework Enables Robust Walking in Bipedal Robots Researchers develop a model-free reinforcement learning method that trains robust locomotion policies in simulation and successfully transfers them to the real Cassie bipedal robot.
neurips 2020 NeurIPS 2020 Highlights Breakthrough Reinforcement Learning Research Researchers at NeurIPS 2020 unveil innovative reinforcement learning techniques focusing on sample efficiency, offline learning, and novel exploration methods. Leading institutions introduce new approaches to solve complex tasks with sparse rewards.
openai OpenAI Builds 16-Game Arcade to Test Whether AI Agents Cheat or Truly Learn OpenAI creates a suite of 16 simplified arcade games designed to test if AI agents are actually learning generalizable skills or just finding clever ways to cheat the system. By changing the rules and environments, researchers can see if the AI truly understands the game mechanics.
openai OpenAI Agents Develop Clever Tool Use During Hide-and-Seek Simulation OpenAI researchers observe machine learning agents teaching themselves sophisticated tool use and rule-breaking strategies in a virtual hide-and-seek game. This simulated environment serves as a stepping stone for teaching AI to navigate real-world physics.
artificial intelligence AI Breakthroughs in Drug Discovery and Quantum Chemistry Signal Industry Shifts Two recent AI advancements from InSilico Medicine and DeepMind show the technology moving beyond minor efficiencies to disrupt pharmaceutical and materials science industries.
openai OpenAI Agents Evolve Tool Use and Teamwork Through Hide-and-Seek OpenAI researchers discover that AI algorithms develop surprisingly complex behaviors like tool use and collaboration simply by playing hundreds of millions of virtual hide-and-seek games.
microsoft Microsoft Adds Personalization and Handwriting APIs to Cognitive Services Microsoft expands its Cognitive Services platform ahead of the Build developer conference with new machine learning tools, including a reinforcement learning-based Personalizer. The new APIs also bring handwriting recognition and conversation transcription capabilities to third-party developers.
deepmind DeepMind AI Achieves Historic Victory Against Pro Gamers in StarCraft II DeepMind's AlphaStar algorithm defeats professional StarCraft II players by mastering complex strategy and imperfect information. This breakthrough demonstrates new levels of artificial intelligence that could eventually apply to real-world strategic planning.
microsoft Microsoft Acquires AI Startup Bonsai to Boost Azure Autonomous Systems Microsoft acquires Berkeley-based Bonsai to advance its reinforcement learning capabilities and integrate them into the Azure cloud platform. The deal aims to make artificial intelligence development more accessible to developers of all skill levels.
microsoft Microsoft Acquires AI Startup Bonsai to Boost Cloud Offerings Microsoft purchases Bonsai, a startup specializing in reinforcement learning, to enhance its Azure cloud platform. The acquisition supports Microsoft's ongoing strategy to commercialize artificial intelligence for enterprise customers.
neurips NeurIPS 2018 Breaks Attendance and Submission Records in Montreal The 32nd NeurIPS conference draws a record 8,300 attendees and accepts 1,010 papers out of nearly 5,000 submissions. Reinforcement learning and generative models dominate the event, while speech processing earns notable spotlight recognition.