The Quanta Podcast Audio Edition: AI Comes Up With Bizarre Physics Experiments. But They Work.
21 snips
Apr 2, 2026 Researchers let AI redesign precision interferometers and uncovered alien-looking optical layouts that nevertheless cut quantum noise. Graph-based algorithms reinvented long-known quantum experiments in unfamiliar ways. Machine-learning models also surfaced better dark-matter formulas and hidden symmetries from collider data, while humans still supervise and interpret the surprising outputs.
AI Snips
Chapters
Transcript
Episode notes
AI Suggested A 3 Kilometer Ring For LIGO
- Rana Adhikari's team fed an unconstrained AI all LIGO components and it produced alien-looking interferometer designs.
- After months they decoded a counterintuitive trick: adding a 3 km ring to circulate light, an idea tied to old Russian theory that reduces quantum noise.
AI Surfaces Overlooked Physics Ideas
- AI can highlight overlooked ideas even in fields with thousands of experts, surfacing solutions humans miss.
- Examples include LIGO sensitivity gains and rediscovering theoretical tricks nobody had pursued experimentally.
Graph Optimization Led To New Entanglement Swap
- Mario Krenn's PyTheus represented optical setups as graphs and optimized them to produce desired quantum outputs.
- The algorithm found a radically different entanglement-swapping design, later built and validated by Xiaosong Ma's team in China.