Predictions 2020: Quantum Sweet Dreams
Through 2024, you will not purchase a quantum computer or rent it in the cloud, unless for training or basic research. But the adventure is worth watching.
Net of the hype, 2019’s most remarkable events were the appearance of IBM Q System One and Google’s proof that a quantum computer could solve problems intractable on classical Turing machines (“quantum supremacy”).
The former was aimed to affirm IBM’s lead and push the industry by initiating a virtuous circle of investments: a bold and beautiful endeavour, although a machine of limited applicability. The latter was an experimental confirmation of what quantum information theory predicted in the early 1990s: in Scott Aaronson‘s words, “an outcome that’s at once expected and mind-boggling, conservative and radical”.
This fascinating technology is farther in the future than you could deduce from the media. Building general-purpose quantum computers or even quantum-based specialized accelerators will not be easy:
- Quantum computers are plagued by electronic noise, far more than ordinary digital computers.
The processing overhead leading to noise-free QC is (exponentially) enormous. Today, hundreds or thousands of physical qubits are needed to obtain one fault-tolerant, “logical” qubit (the abstract entity on which Boolean logic operations can be reliably executed);
- While we wait for solutions, we ought to limit ourselves to “noisy” quantum computers, called NISQs, and it is unsure if they could offer advantages over classical ones for practically meaningful applications. For sure, stuff like threatening public-key cryptography is a long way away;
- It is yet undecided what would be the best hardware foundation for even just a NISQ (like, e.g., trapped-ions or superconducting);
- Analog quantum computers (circuits that emulate the laws of physics) are being considered as alternatives to digital (based on Boolean logic), and this is not small dilemma;
- Hardware most often requires extremely challenging environmental conditions, like exceptional robustness toward vibrations or near-zero K operating temperature (that’s why IBM Q System One is a beautiful engineering feat);
- The software stack is underdeveloped, from operating system to programming languages. Even if a full-scale noise-free QC appeared magically tomorrow, we would have 1/100 of the programmers necessary to do anything with it.
Quantum computing is a wonderful discipline and a challenging subject for some of the world’s most brilliant minds and best high-tech enterprises. But it is still developing in the labs and likely to continue like that for a decade at least.
I am personally confident that actual progress will be fast. But beware: it will still be much slower than what is promised by the media hype.