What’s New

• 2017 August 18, Andrew Brock, Theodore Lim, J. M. Ritchie, and Nick Weston. SMASH: One-Shot Model Architecture Search through HyperNetworks. arXiv:1708.05344. code. (Architecture Search)
• 2017 August 9. DeepMind and Blizzard open StarCraft II as an AI research environment. (DeepMind; Starcraft II)

Main Contents

1. Overview
2. AI and Society
   1. Impacts:
      1. Applications and Demos
      2. Progress and Forecasts
   2. Timing: Prosaic AI, Roadmaps
   3. Research Groups: supported by Industry and Policies
      1. DeepMind: Publications
      2. OpenAI: Publications
      3. Google Brain: Publications
      4. FAIR: Publications
   4. Safety
      1. Blog: Towards Safe and Beneficial Intelligence
      2. Effective Altruism
      3. Organizations
      4. Oracle AI
3. Understanding AI
   1. Introduction
   2. Philosophy of Intelligence
4. Course
   1. System: Google Cloud Platform
   2. Programming in Python 3 with Jupyter Notebook
   3. Deep Learning in TensorFlow
5. Frontiers (excluding Deep Reinforcement Learning)
   1. Learning Goals
      1. Concept Learning
      2. Continual Learning
      3. Few-Shot Learning
      4. Multi-Task Learning
      5. Predictive Learning
      6. Transfer Learning
      7. Unsupervised Learning
   2. Learning Techniques
      1. Attention
      2. Curriculum Learning
      3. Hierarchical Learning:
      4. Imitation Learning
      5. Meta-Learning
      6. Memory
   3. Reinforcement Learning
      1. Safe Reinforcement Learning
      2. Agents
      3. Model-Free Algorithms
      4. Intrinsic Motivation
         • Exploration
         • Empowerment
      5. Auxiliary Tasks
      6. Multi-Agent Learning
      7. General Reinforcement Learning
      8. Reinforcement Learning Implemtations
   4. Deep Learning
      1. Network Architecture
         • Architecture Search
         • Memory Augmented Neural Networks
         • Neural Interface
      2. Learning Rules
      3. Evolutionary Algorithms
      4. Adversarial Learning
      5. Generative Models
      6. Platforms
      7. Deep Learning Theory
      8. Deep Learning Implementations
   5. Experiments
      1. Vision
      2. Language: Program Induction
      3. Robotics: Autonomous Driving
      4. Automated Theorem Proving
   6. Neuroscience
   7. Alternative Theories
6. Deep Reinforcement Learning
   1. Outstanding Challenges
      1. Symbol Grounding
      2. World Model
      3. Planning
      4. Reasoning
   2. Environments
7. Research