Member-only story
Quantum Sundays |78⟩ Quantum Computing Breakthroughs and Their Use in Modern Artificial Intelligence
A practical tour of what quantum computing can, cannot, and might someday do for AI, and what AI is already doing for quantum computing
TL;DR: Despite billions in investment and relentless vendor hype, quantum computers will do essentially nothing for mainstream AI through 2030, because three walls stand in the way: the QRAM data-loading bottleneck (getting a billion classical numbers into a quantum state costs as much as just processing them), dequantization (Ewin Tang and others proved the headline “exponential” quantum ML speedups can be matched classically), and the barren-plateau dilemma (quantum neural networks expressive enough to matter become untrainable, while trainable ones tend to be classically simulable). The genuine breakthroughs of 2024–2026, including Google Willow’s below-threshold error correction and Quantinuum’s 48 error-corrected logical qubits, are historic hardware milestones that remain orders of magnitude short of the thousands of logical qubits ML workloads would demand. The strongest results at this intersection actually run in reverse: DeepMind’s AlphaQubit neural decoder and reinforcement-learning calibration are fixing quantum computing’s hardest…