Circuit Conspiracies
Talkat Near-Term Quantum Computers: Fault Tolerance + Benchmarking + Quantum Advantage + Quantum Algorithms · 2024-04-22..04-26·
summary
Poses the peakedness problem: pick random quantum circuits (1D nearest‑neighbor SU(4) layers), condition on having an output basis string with probability ≥δ, and ask if such circuits remain classically indistinguishable from random. Uses gradient‑descent to optimize a small number τp of final “peaking” layers after τr random layers, reports numerics (n up to ~16 qubits, τr≈n, τp ≪ τr) showing surprisingly large δ (≈0.5 at n=10, ≳0.1 at n=12) and exponential decay with n, and links to an analogous neural‑net postselection question (Paul Christiano) and arXiv/Zhang results. tau_r vs tau_p numerical optimization
AI-written summary · about
chapters (8)
- 0:00Quantum advantage motivation
- 2:00NISQ requirements
- 8:59Peakedness problem
- 18:38Peaked vs random circuits
- 25:41Peaking layers experiments
- 37:17AI safety and interpretability
- 47:26Summary and future
- 48:07Q&A
related talks
51mPeaked quantum advantage using error correction
45mQuantum Pseudo-Randomness From Domain Walls and Spectral Gaps
01h 08mPutting Quantum Supremacy on a Rigorous Footing
50mQuantum Benchmarking
01h 09mRecent Progress in Quantum Advantage
01h 09mAre Random Quantum Circuits PRUs?
45mNoise vs Imprecision in the Hardness of Random Quantum Circuits
59mUtility, Computational Cost, and Effective Quantum Volume of NISQ Experiments