More on Dota 2 
Transitioning entirely to neural machine translation - Gathering human feedback
Topic Suggestions for Millions of Repositories - Better exploration with parameter noise
- Using AI for new visual storytelling techniques in VR
- Proximal Policy Optimization
- Robust adversarial inputs
Piecewise regression: When one line simply isn’t enough - Hindsight Experience Replay
- Teacher–student curriculum learning
- Deal or no deal? Training AI bots to negotiate
- Learning from human preferences
- Accelerating machine learning for computer vision
- Learning to cooperate, compete, and communicate
- UCB exploration via Q-ensembles
- Under the hood: Photo Search
- Robots that learn
- A novel approach to neural machine translation
- Equivalence between policy gradients and soft Q-learning