Today, Julian Kelly, a research scientist at Google Quantum AI Labs, wrote an essay in Google Research that introduced peer reviewed Google's latest 72-bit Quantum Bit general purpose computer.
'We just started testing,' said John Martinis, a physicist at Google. 'We are very optimistic about what we know now,' Martinis said if everything goes well, quantum hegemony is likely to come true in a matter of months .
Kelly said the goal of Google Quantum AI Lab is to build quantum computers that can be used to solve real world problems.Google's strategy is to explore recent applications using systems that are compatible with general purpose error correction quantum computers To enable quantum processors to run algorithms beyond the scope of classical simulations, it not only requires large-scale quantum bits, but also the processor's low error rate guarantees for readout and logic operations, such as single-bit gates and Two-door
At the annual meeting of the American Physical Society in Los Angeles, Google showed off a new quantum processor called Bristlecone, a gate-based superconducting system designed to study the systematic error rate and scalability of quantum bit technology, as well as its advantages in quantum simulation, optimization And machine learning applications.
A record quantum 72 quantum computer, a 1% error rate, can achieve quantum hegemony
According to Julian Kelly, the latest device follows the physics principle of the linear array technology of the nine qubit quantum computers proposed by Google, and the best results shown by the technology are as follows: Low reading error rate (1%), Single qubit gates (0.1%) and most important double qubit gates (0.6%). The device uses the same pattern as 9 qubits for coupling, control, and readout, but expands it to include 72 A square array of qubits.
In the experiment, the researchers chose this size device to demonstrate the quantum hegemony of the future, studied the first-order and second-order error correction using surface coding, and promoted the development of quantum algorithms on the actual hardware.
Before researching a specific application, the ability to quantify the quantum processor is important. Google's theory team has developed a benchmarking tool to accomplish this task by applying a random quantum circuit to the system. The tasks were randomly assigned and the distribution of the sampled output was checked by a classical simulation method. For a quantum processor with a small enough operational error, it can have a supercomputer beyond the classic in a clear computer science-related problem. Performance, that is, 'quantum hegemony' These random circuits must be large in terms of both the quantum bits and the calculated length and depth.
Although no one can achieve this goal yet, Google researchers calculated that the goal of quantum hegemony can be perfected by using 49 qubits, a circuit depth of more than 40, and a 2 bit error of less than 0.5%. Prove they believe that the quantum processor superior to the supercomputer experiment proved to be the watershed in this area, but also one of the main goals of the future.
Google had hoped to achieve similar performance on nine qubits but had 72 qubits now and pointed out that Bristlecone would be a convincing principle in the future when it comes to building larger quantum computers Proofing that Google also said running a device such as the Bristlecone at a low system error requires a combination of technologies such as software, control electronics and the processor itself, there is a need for careful system engineering in a few rounds of iterations Observation.
At Google Quantum AI Lab, quantum physics can be achieved using Bristlecone, and learning how to build and operate the device at this level can be an exciting challenge as they look forward to sharing the results, which can also help more researchers Conduct experiments in this area.
Google: From 49 to 72, we are not only a little better than IBM
As we said before, Google has said it is going to build a 49-qubit quantum computer to realize quantum hegemony, claiming that they are the result of their calculation, and why did you jump to the '72 'qubit this time?
It has long been assumed that 50 qubits of quantum computers are 'starting prices' for quantum hegemony. "Shortly after Google dubbed Quantum Hegemony a 49-qubit quantum computer, IBM said that their research Shows that for some specific quantum applications, it may take 56 or more qubits to achieve quantum hegemony.
Not only that, on November 12, 2017, IBM announced the successful development of a 20-qubit quantum computer and successfully built and tested the world's first 50-qubit quantum computing prototype.
Last week, at the IBM Inaugural Index Developer Conference in San Francisco, IBM showed off its 50 qubits of prototypes and the internal structure was also exposed.
However, in order to realize the quantum hegemony, we have to talk about the quantum simulation just mentioned.At present, the most powerful supercomputer can only simulate 46 quantum bits.It is very challenging to simulate a general-purpose quantum computer on a traditional electronic computer In November 2017, a joint research team consisting of Prof. Yuan Shengjun, Faculty of Physics and Technology, Wuhan University, Prof. Jinfeng Ping, Prof. Kristel Michielsen from the Julich Supercomputing Center in Germany, and Prof. Hans De Raedt from University of Groningen, Netherlands, A series of simulations of a general purpose quantum computer were implemented on a 45-qubit simulation at Light of the Supercomputer Shenwei · Taihu Lake at the National Supercomputing Center in Wuxi, China Then in December 2017 the team made another breakthrough, achieving 46 qubits Simulation, setting a current world record.
Extensibility is one of the challenges in implementing a universal quantum computer, and from 45 to 46, it looks like adding only one bit, but in computer simulations, doubling the memory of a computer for every additional one bit of quantum. For example, simulating a quantum computer with 45 qubits requires at least 0.5 PB (about 0.5x10 ^ 15 bytes) of memory.
So, to simulate a 72-qubit computer, it would take millions of megabytes of RAM (2 ^ 26 (72-46) bytes). We probably will not be able to use that much memory in supercomputers, so if Bristlecone The ability to run common algorithms faster than the most powerful supercomputers so far will be just around the corner.
Seize the 2018 quantum hegemony contest point, small commercial quantum computer appears within 5 years
In addition to quantum bits, quantum hegemony also requires a low error rate, a quickest but erroneous quantum computer that is not as good as a classic supercomputer.
According to Google, the minimum error rate for a quantum computer must be within 1% and has a scale close to 100. For the moment, at the error rate, Google has achieved this goal on a 72-bit quantum computer, with a single quantum The bit gate is 0.1% and the double quantum bit gate is 0.6%.
When we can achieve error rates of 0.1-1% for tens or even millions of qubits, the quantum computer will start to really solve the real problem efficiently, which can take ten years or more.
But at least Google believes that before we can create large-scale quantum-bit quantum computers, we might first implement small, even commercial quantum computers, or quantum computing commercial applications.
In 2017, the Google Quantum team said in the Nature issue that they are convinced that even in the absence of a theory that can be fully error-corrected, there will still be small-scale devices in five years that are associated with quantum computing, which will also bring investors Short-term returns. "Early quantum computing devices will have commercial applications in quantum simulation, quantum-assisted optimization, and quantum sampling, and faster computational speed has clear commercial advantages from artificial intelligence to financial and medical fields. '
Google hopes their 72-qubit Bristlecone quantum computer not only enables quantum hegemony but also serves as a test fixture for studying quantum bit scalability and error rates, developing applications such as quantum simulation, optimization and machine learning.
By the end of January 2018, the Financial Times released news that in the coming weeks, Google and Microsoft will each announce two major milestones in quantum technology.
Now Google gives their 72 qubits quantum computer, next, look at Microsoft.