Where Is Wall-E? How Robotics and AI Can Change the Face of Recycling
summary
An experiment shows ChatGPT mislabels a realistic robot doll as a baby, highlighting limits of common sense but usefulness of language interfaces; recycling prototypes combine foundation models, incremental learning and natural interaction to recognize billions of objects, learn new damaged items in real time, and let humans guide robots by demonstration. Roboton benchmarks shrink the robot‑to‑human gap from ~120s extra to ~20s, but model training still takes weeks; a vehicle‑parts demo disassembles components successfully yet fails because no business model prevents shredding.
AI-written summary · about
chapters (11)
- 0:00Intro
- 1:17Recycling
- 3:18Circular Economy
- 3:52Automation and Robotics
- 4:48Recycling is hard
- 5:39Where is WallE
- 6:47Whats the problem
- 7:50Smart and fast robot
- 9:23Training AI models
- 10:33We are ready now
- 11:49Natural interaction
related talks
16mOn the ethical challenges raised by robots powered by AI
17mBeyond Automation: AI and the Evolution of Robotic Intelligence
12mAI Disassembly for a Sustainable Circular Economy
16mReal-World Robotics as the Next Frontier for AI?
45mHyper Innovation in Exponential Times
03mMaking good robots
18mOn the History and Future of AI
52mRead, Write, and Tinker with AI