Crowdsourcing Artificial Intelligence for Science: YOU can do a lot!
summary
A crowdsourced reinforcement learning challenge uses the crowdAI platform and a donated NVIDIA DGX‑1 to teach a simulated skeleton named Bob with 18 muscles to walk, run and avoid reward‑gaming; about 500 participants and ~2000 submissions compete with starter kits and baseline code, iterate on simulator speed and parallelism, and push solutions that reach up to 4.5 m/s, reveal surprising engineering tweaks, and generalize to tasks like satellite building detection, adversarial robustness, Minecraft and Doom agent competitions.
AI-written summary · about
chapters (6)
- 0:00Start
- 0:24What Can You Do with Artificial Intelligence and What Can You Do for Artificial Intelligence
- 0:34What Can You Do with Artificial Intelligence
- 1:35What Can You Do for Artificial Intelligence
- 10:14Reinforcement Learning
- 18:24The Ai Generated Music Challenge
