Reinforcement Learning and the Next Gen of AI
summary
Sony Research builds reinforcement‑learning agents that become superhuman in Gran Turismo (Nature cover), trained end‑to‑end in massive PlayStation-backed infrastructure and compressed to run on consoles across ~400 cars; a commercial December power pack sells for $39. The same RL pipeline trains a table‑tennis robot that beats top-ranked pros, using high‑speed perception that tracks logo spin at up to ~100 km/h and ~9,000 rpm and learning from simulation with minimal handcrafting to master tactics and control.
AI-written summary · about
chapters (5)
- 0:00Introduction to Sony AI
- 1:32AI in Gran Turismo racing
- 7:33Bringing AI to casual gamers
- 9:36Reinforcement learning in robotics
- 13:47Future of AI and reinforcement learning
related talks
37mA practical perspective on training reinforcement learning bots.
13mReinforcement learning and it's growing role in AI
48mMachine Takeover - Improving Your Game With Machine Learning & AI
22mDeveloping AI in Racing Games Using Reinforcement Learning
31mDesigning game AI with reinforcement learning
30mReinforcement Learning - ChatGPT, Playing Games & More
29mThe Present & Future of AI and Games
32mThe Road to General AI