量子コンピューターは「何でも速い超すごいPC」ではなかった。本当の得意技とは？

Many people think a quantum computer is just a super-fast PC. That is not true. Quantum computers will not make games or videos faster. They are good only at some special problems, like studying molecules or breaking some codes.

The biggest problem is errors. Quantum bits are very weak, and heat or small vibrations can break them. So scientists often keep them very, very cold.

Quantum computers are often described as super-fast versions of ordinary computers, but this is misleading. Ordinary computers use bits that are either 0 or 1. Quantum computers use qubits, which can be in a mix of 0 and 1 called superposition.

However, the real key is interference. Calculations are designed so that paths leading to the right answer grow stronger while wrong paths cancel out. This only works for certain problems, such as simulating molecules for new medicines or factoring huge numbers, which could threaten some internet encryption.

Everyday tasks like watching videos will not get faster. The biggest challenge is errors, because qubits are easily disturbed by heat and vibration. In 2024, Google showed that larger error-correction groups could actually reduce errors.

Quantum computers are often portrayed as machines that will make every task dramatically faster, but this picture is largely inaccurate. Whereas classical bits are either 0 or 1, qubits can exist in superposition. The common claim that they "try every answer at once" is misleading, since measurement yields only a single outcome.

The real advantage comes from interference: algorithms are designed so that paths toward correct answers reinforce one another while incorrect ones cancel out. This works only for particular problems, notably simulating molecules and materials, and factoring large numbers, as Peter Shor showed in 1994, which threatens some current encryption. In 2019, Google reported a 53-qubit processor completing a specialized task in about 200 seconds, claiming a supercomputer would need 10,000 years, but rivals disputed this and classical methods have since improved.

The greatest obstacle is error, as qubits are disrupted by heat, vibration and stray electromagnetic waves. In 2024, Google's 105-qubit Willow processor showed that enlarging error-correction groups cut the error rate by more than half at each step. Breaking current encryption, though, would need far more qubits; one 2021 estimate put the figure at about 20 million. Even so, quantum-resistant encryption standards appeared in 2024.