Physics. For quantum information, two wrongs can make a right.

PERSPECTIVES C an you reliably send information down a telegraph wire that doesn't always transmit signals correctly? Claude Shannon put classical information theory on a firm footing when he showed that you can correct for transmission errors as long as there is some tiny correlation between what gets sent and what is received. What's more, Shannon quantified how much information could be reliably communicated. From its onset, classical information theory was intimately entwined with communication. The birth of quantum information theory began from an apparently different direc-tion—cryptography—when it was realized that if you can reliably send someone quantum states, then you can use those states to exchange private messages that cannot be cracked by even the most powerful computer (1). This cannot be done classically without exchanging a physical key beforehand that is as long as the message you want to send. However, we are still wrestling with the corresponding question that was so central to classical information theory: How much quantum information can we reliably send down a noisy channel? On page 1812 of this issue, Smith and Yard (2) have discovered that we may be further from answering this question than we think, but that intriguing clues might come from the very place that initially sparked our interest in quantum information: cryptography. Classically, a telegraph wire that is so noisy that no information can be reliably sent through it is useless. These are called zero-capacity channels. But what about the quantum case, such as trying to send information (which might be conveyed by the polarization of a single photon) through a fiber-optic cable that is so noisy it cannot be used to send any quantum state reliably? Because our intuition tends to be classical, it was generally believed that a channel that cannot convey quantum information would also be useless. Yet a few years ago, the Horodecki brothers and I found that although these channels cannot be used to send quantum states, they can be used to send classical private messages. Indeed, one can classify all states that, if shared over some channel, are private (3). What's more, this privacy is verifiable , which means that practical cryptography can be performed over these zero-capacity fibers (4). The belief that quantum cryptography required being able to reliably send quantum states turned out to be wrong. Now, Smith and Yard, using results from (5), have shown a remarkable property of …