The Road to General AI
summary
Unity serves as a high‑speed sandbox for concrete ML results: reinforcement learning agents learn a chicken crossing task from tabula rasa to robust policy in minutes, curriculum learning produces stepwise skills, multi‑agent play yields emergent goalie/striker roles, and attention reduces sensor load. Industry-scale examples show self‑play beating champions, OpenAI trains hand manipulation with 300 billion frames and Rubik flips with 10 trillion frames, AlphaFold2 reaches ~90% accuracy, and simulated robots transfer to real hardware.
AI-written summary · about
chapters (15)
- 0:00Intro
- 1:55Board & trivia games in AI
- 4:36The 4 dimensions to AI
- 6:12Nature's learning method: Reinforcement
- 7:07Chicken crossing the road: Tabula rasa
- 7:49Demo
- 9:14Control learning
- 11:12Curriculum learning
- 12:31Memory in learning
- 13:14Hierarchical learning
- 13:37Attention
- 15:34Multi-agent learning
- 18:31Traits in learning
- 19:43On the road to a greatly improved AI
- 31:30Outro
related talks
36mOn the Road to Artificial General Intelligence
56mCan I do AI?
48mMachine Takeover - Improving Your Game With Machine Learning & AI
01h 01mIntroduction to Practical AI
48mRobots Among Us: Advances in AI for Everyday Androids
10mMachine Learning & Unity ML-Agents update
31mFrom Single AI to Multi AI Systems
37mA practical perspective on training reinforcement learning bots.