Hidden Patterns #7

Hello! Today's newsletter focuses on Reinforced Learning and some articles on how computer simulations can be used to learn tasks and transfer to real robots. Today's robots can be programmed to perform various functions but require a human to input a sequence to perform that task. The robots can perform more generic tasks and step up in the autonomy game by transferring knowledge from simulations into the real world. Metaverse can contribute to building the knowledge database for robots to perform tasks.

News:

• OpenAI Announces Question-Answering AI WebGPT [Link]
• Researchers Build AI That Builds AI [Link]
• Meta has built an AI supercomputer it says will be world’s fastest by end of 2022 [Link]
• Tactical Artificial Intelligence Wargaming that uses AI [Link]

Articles:

• Reinforcement learning for the real world [Link]
• Introducing deep reinforcement learning to NLU ranking tasks [Link]
• Three Ways AI Is Helping Manufacturers [Link]
• The industrial Metaverse: Where simulation and reality meet [Link]
• Reinforcement learning improves game testing, EA’s AI team finds [Link]

Learn:

• Alpha Zero Cheat Sheet [Link]
• Key Concepts in RL [Link]

Apps:

• Self-Driving Cars in the Browser [Link]
• PuckWorld: Deep Q Learning [Link]

Video:

• Learning to Walk in Minutes Using Massively Parallel Deep RL