Tearing down AlphaStar
summary
A breakdown of DeepMind’s AlphaStar AI for StarCraft II, explaining its internal architecture and the individual components (models, modules, and training elements) that make up the system. For each component the talk explains what it does and how it contributes to AlphaStar’s overall gameplay strength.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 5:56Why Starcraft for AI
- 14:21AlphaStar architecture
- 21:48Neural networks in AlphaStar
- 31:26Unite Embedding and LSTM
- 42:41Training AlphaStar
- 47:18Results and demo
- 50:16Conclusion
related talks
19mKeynote: AlphaStar: How Google Infrastructure Made it Possible
59mGPT-4 vs Starcraft II – Strategic Decision Making using Large Language Models
24mApplied AI research with StarCraft, Torch and the BEAM
01h 03mIEEE CoG 2019 - Industry Day - David Silver (DeepMind)
48mMachine Takeover - Improving Your Game With Machine Learning & AI
44mGPT-4 vs Starcraft II – Strategic Decision Making using Large Language Models
14mAI for Science
43mDebugging AI