For months, we have been exposed to an overdose of enthusiasm. To listen to the major technology companies, the marriage of artificial intelligence and quantum computing will soon produce a superintelligence that makes today’s data centres obsolete.
Science moves at another pace. A recent exchange between Andrew Yao and Gilles Brassard, both Turing Award laureates, helps put matters in proportion: the quantum shift matters, but not for the reasons currently being sold to us.
A conventional computer handles bits: 0 or 1. A quantum computer uses qubits, which can occupy a superposition of possible states before measurement and become entangled, meaning their properties are correlated in ways with no classical equivalent.
In principle, this may make some forms of calculation (including parts of machine learning) more efficient. A recent academic review finds substantial potential, but also unresolved questions about noise, error correction and meaningful advantage over classical approaches.
In practice, we are in what physicists call the NISQ era: noisy, intermediate-scale quantum computing. Machines from IBM, Google and others already carry out highly specialised calculations beyond the reach of conventional methods, a milestone often described as quantum utility. Yet their qubits remain so delicate that even slight environmental interference can destroy the quantum state, a phenomenon known as decoherence.
Brassard is clear on this point: putting an exact date on the arrival of a universal quantum computer (the kind that could transform commercial AI) would be premature. The research is bearing fruit, but fault-tolerant systems are not around the corner.
The real risk lies elsewhere
While the market dreams of quantum brains, cybersecurity specialists face a problem that has already begun: Harvest Now, Decrypt Later.
Hostile actors can collect encrypted files today (industrial secrets, medical records, financial data) and hold onto them. They cannot read them now. They are waiting for a sufficiently capable quantum processor to break the public-key cryptography on which much of today’s digital security depends.
That is the uncomfortable arithmetic: if information protected today will still matter a decade from now, it may already be exposed to the risk of being collected and decrypted later. The threat is explained here.
Change the locks now
The operational response is already under way; it does not depend on waiting for a fully fledged quantum computer. The US National Institute of Standards and Technology released its first principal post-quantum cryptography standards in 2024 and advises organisations to begin migrating now. These are mathematical algorithms designed to resist attacks from both conventional and future quantum computers, and they can be deployed on existing infrastructure.
That is the lesson in the Yao–Brassard conversation. Serious innovation begins in interdisciplinary basic research, often long before it acquires a commercial label. If we are serious about preparing for the quantum era, the task is not to fantasise about tomorrow’s AI, but to strengthen today the cryptography that underpins our economies. Ipsi dixerunt.
This article is republished from Futuribles. Here's the original article in Spanish and English.