153 | John Preskill on Quantum Computers and What They're Good For
What this covers
John Preskill, a leading figure in quantum information science, joins Sean Carroll to sketch what quantum computers actually do and where they matter. The episode unfolds through a tension: quantum computers are genuinely powerful, but their power is narrowly distributed. Preskill argues that the clearest and most defensible application is simulating quantum systems—molecules, materials, catalysts—a direct echo of Richard Feynman's 1981 vision. Applications to cryptography, optimization, and general problem-solving, by contrast, rest on shakier ground. The conversation moves through how entanglement creates computational advantage, why measurement constrains what you can extract, and what the engineering challenges actually are. It then pivots to deeper terrain: how quantum error correction reveals something profound about the nature of space itself.
The substance divides into technology, physics, and honest skepticism. On the hardware side, Preskill surveys competing platforms—superconducting circuits and trapped ions leading, Rydberg atoms advancing—and explains why scaling each one hits distinct bottlenecks. He lays out the gap between today's noisy devices (50–100 qubits, error rates around 1% per operation) and the error rates needed for useful work (10^−5). On the foundations, he clarifies why the "parallel universes computing in parallel" picture misleads, why factoring lacks a rigorous hardness proof, and why NP-hard problems probably resist quantum speedup. Near the end, the discussion opens onto holographic duality and emergent spacetime: how quantum error correction maps onto the structure of gravity, how geometry itself appears to arise from entanglement, and why that robustness matters. Preskill is direct about timelines too, pushing back against hype by separating physics breakthroughs (five years) from practical applications for paying customers (not in the near term).
Preskill argues that quantum computers are genuinely transformative but narrowly so: their clearest payoff is simulating quantum systems like molecules and materials, while applications to cryptography, optimization, and general speed-ups are uncertain or modest, and the field must focus on better hardware and error correction rather than over-hyped near-term applications.
- Quantum advantage comes from entanglement being inexpressible classically, but extracting useful answers is constrained by measurement, so only specially structured problems get big speed-ups.
- Simulating quantum systems is the most defensible application; optimization and even factoring lack rigorous hardness proofs.
- Quantum error correction is both an engineering necessity and a deep scientific insight that now connects to emergent spacetime.
Quantum computing's novel processing model makes predicting its key applications impossible.
- Because quantum computing processes information so differently from classical computing, we have limited ability to forecast its most important applications; by analogy to how people once couldn't foresee uses for home computers or web browsers, useful applications of quantum computers will likely emerge that we cannot currently anticipate.
“if quantum computing is so different from the information processing we've been familiar with up to now, then we probably have a very limited ability to forecast what its most important applications are going to be”