Despite all the hype around quantum computers, they are still far too error-prone to be of real use. But recent experiments show that this may not always be the case, boosting the credibility of claims from companies like Google and IBM that we might get useful quantum computers as soon as 2029. These latest experiments represent key milestones and signal that we are entering a new age, say researchers.
鈥淪uddenly, really useful devices seem tantalisingly close, in a way that they never have done before,鈥� says at the University of Cambridge.
For much of the past decade, quantum computing companies were focused on building ever larger machines, steadily increasing the number of quantum bits, or qubits, in聽their systems. Qubits are units of quantum information, made using physical systems like the spin of an聽electron or the orientation of a聽photon. But these qubits were too聽prone to errors to reliably run algorithms of real-world use.
Advertisement
Companies now appear to be shifting their attention towards building error-free qubits, called logical qubits. These are collections of physical qubits that together can reduce errors to a good enough level to run such algorithms. 鈥淚t鈥檚 marking a reset in the whole conversation around quantum computing and new benchmarks,鈥� says Vicary. 鈥淭his聽is exciting because this is聽the聽time when quantum computers聽start to be useful.鈥�
In August, showing that as you build logical qubits by adding more physical qubits to a computer, the errors won鈥檛 snowball and become unmanageable 鈥� they will聽instead cross a threshold where, in principle, they will shrink as the system gets larger. This works by spreading information across a聽group of qubits, so if an error occurs聽in one, it won鈥檛 affect the overall computation.
鈥淚deally, you want to scale up while聽at the same time further reducing the error rates, improving the quality of your qubits,鈥� says at University College London. 鈥淭rying to do all of those things at once is very difficult, but what makes me optimistic is this Google experiment is an example of聽them doing exactly that.鈥�
However Google鈥檚 work didn鈥檛 involve performing computations on the qubits 鈥� instead, the researchers showed that these qubits can act as memory, says at Imperial College London.
, from researchers at Microsoft and quantum computing start-up Quantinuum, shows a combination of error-corrected qubits and computation. The team set up different combinations of qubits to make four logical qubits, then performed basic logical operations on the system, for example where the value of a qubit聽is聽flipped from positive to negative. 鈥淭hey have fewer rounds of聽error correction, so their quantum memory聽is stable for less time, but they can also do some computation with it,鈥� says Bondesan.
The quantum computer in Microsoft鈥檚 study uses a different hardware design to Google鈥檚, opting聽for a series of magnetically trapped charged particles instead of superconducting pieces of wire. This allows it to utilise an error-correction technique called a tesseract code, where qubits are arranged in a complex geometry known as a聽four-dimensional hypercube. 鈥淚n聽principle, they can host more logical qubits, with fewer physical qubits,鈥� says Bondesan. 鈥淚n this sense, it鈥檚 more efficient.鈥�
Other researchers have shown error correction working in more unusual quantum computers. 聽at Yale University and his colleagues tested a form of error correction called bosonic codes, which spreads out errors over vibrations in quantum computers. Instead of qubits, this system uses 鈥渜udits鈥�, which can take more values聽than just 1 and 0 and are聽theoretically more powerful. Meanwhile, researchers at Amazon鈥檚聽quantum computing team聽demonstrated , which, like Google鈥檚 work, could reduce errors as systems get larger.
鈥淭he Google and Microsoft approaches are really following more mainstream qubit-based quantum computing, whereas the introduction of bosonic codes in the Yale and the Amazon experiments are more novel and exploratory, but also very, very exciting,鈥� says Browne.
Seeing error correction work over聽so many different designs and聽experiments is a sign that previous theoretical results could work for real-world systems, says Browne. 鈥淭here were lots of nice theoretical ideas about fault-tolerant quantum computing and error correction, but聽none of them had been demonstrated, or [they were demonstrated] in a very limited way or special case,鈥� says Browne. 鈥淭hat鈥檚 really changed a lot in the last couple of years. There have been a lot of key milestones reached, demonstrating one by one different aspects of fault-tolerant quantum computing.鈥�
However, the rate of progress may yet falter and full fault-tolerant systems might still be a long way off, dashing the hopes raised by Google and IBM鈥檚 optimistic timetables. 鈥淚 would be surprised if we see things moving at a very consistent rate. I think there will be roadblocks,鈥� says Browne. 鈥淓ach of these platforms is going to expand as far as they can until they hit the next roadblock. It鈥檚 a little bit hard to predict where those roadblocks will be.鈥�
