IBM has announced the release of a quantum computing chip and plans to build a supercomputer within 10 years.

Wallstreetcn
2023.12.05 07:46
portai
I'm PortAI, I can summarize articles.

IBM released the quantum computing chip "Heron" and the quantum computer IBM Quantum System II, and made plans to produce a supercomputer in 2033. The Heron processor has three to five times better performance and a 2/3 lower error rate than previous quantum processors. IBM has also demonstrated a new modular system that promises to produce compelling quantum machines by 2033.

The global supercomputer development competition is in full swing, and IBM unexpectedly released its ace card on Monday, unveiling the company's most powerful quantum computing chip to date, as well as the IBM Quantum System Two, and setting a grand blueprint to produce supercomputers by 2033.

Quantum Computing Chip Sets New Record for Error Rate

On December 4th, local time, IBM unveiled the quantum computing chip "IBM Quantum Heron" (Heron) for the first time at the company's Quantum Summit, making it IBM's first practical-level quantum processor.

The "Heron" processor has 133 fixed-frequency qubits, surpassing the "Eagle" processor with 127 qubits.

IBM claims that compared to the "Eagle", the "Heron" processor's device performance has improved by 3 to 5 times, and its error rate has reached a historic low, two-thirds lower than previous quantum processors.

Next year, more "Heron" processors will join IBM's industry-leading utility-scale system cluster.

New Modular System Unveiled, Supercomputers Are Not Far from Reality

In addition, IBM also unveiled its first quantum computer, the IBM Quantum System Two, which has over 1,000 qubits, equivalent to qubits in a regular computer. It is reported that this quantum computer will be powered by three "Heron" processors.

IBM showcased a new modular system to the industry, connecting processors inside the machine and then connecting the machines together to form a modular system. When combined with new error-correcting codes, it is expected to produce remarkable quantum machines by 2033, including a supercomputer with 1,000 logical qubits, fully unleashing the power of quantum computing.

Dario Gil, Senior Vice President and Director of Research at IBM, said, "We are in an era where quantum computers are being used as tools to explore new scientific fields."

"As we continue to advance quantum systems, expanding and delivering value through modular architecture, we will further enhance the quality of utility-scale quantum technology stacks and put them in the hands of our users and partners, who will push the boundaries of quantum technology to solve more complex problems."

Key Barrier of Quantum Computing - High Error Probability

Compared to traditional computers, quantum computing utilizes quantum entanglement and superposition to achieve more powerful parallel computing capabilities and much faster computing speeds. However, these quantum states are notorious for being unpredictable and prone to errors. To address this issue, physicists have been attempting to encode an information quantum bit, known as a "logical qubit," by inducing multiple physical qubits (such as each physical qubit or individual ion encoded in a superconducting circuit) to work together.

Researchers generally agree that the most advanced error correction techniques require over 1,000 physical qubits per logical qubit, and a machine capable of useful computation would need millions of physical qubits.

However, in recent months, physicists have become increasingly interested in an alternative error correction scheme called quantum low-density parity-check (qLDPC).

According to a preprint by IBM researchers, this number could be reduced by a factor of 10 or more. The company stated that it will now focus on building chips designed to accommodate qubits that have undergone qLDPC correction in around 400 physical qubits, and then connect these chips together.

Mikhail Lukin, a physicist at Harvard University in Cambridge, Massachusetts, described IBM's preprint as "an excellent theoretical work."

"That being said, implementing this method with superconducting qubits seems challenging and may even take several years to attempt concept validation experiments on this platform," Lukin said.

The problem is that qLDPC technology requires each qubit to be directly connected to at least six other qubits. In traditional superconducting chips, each qubit is only connected to 2-3 neighboring qubits.

However, Oliver Dial, a condensed matter physicist and IBM Quantum's Chief Technology Officer at the Thomas J. Watson Research Center in Yorktown Heights, New York, said that the company has a plan: it will add an additional layer of quantum chips to its quantum computer design to allow for the extra connections required by the qLDPC scheme.

Jay Gambetta, IBM Quantum's Vice President, stated that the company has been taking a dual-track approach to prepare the hardware, including developing the capability to continuously manufacture a large number of high-quality qubits.

He said that Condor, with over 1,121 superconducting qubits, demonstrates the company's progress in this area, and IBM unveiled this processor on Monday.

"It's about 50% smaller in qubits," Gambetta told the media. "The yield is there—we're getting close to 100% yield."

The second problem that IBM has been focusing on is the errors that occur when manipulating individual or pairs of qubits.

Changing the state of a qubit produces subtle signals that can leak into neighboring qubits, a phenomenon known as crosstalk. "Egret," the new processor, is one of the smaller ones and represents the efforts of IBM's research team over the past four years to improve gate performance.

"It's a beautiful device," Gambetta said. "It's about five times better than previous devices, with significantly fewer errors, and the crosstalk cannot be measured accurately."

When will quantum computing be commercialized?

Despite the milestone significance of quantum computing research, commercialization has not yet been achieved.

"This has always been a dream, and it has always been a distant dream," said Dial. "In fact, getting it close enough to see where we are today is huge for me."

IBM has extended its quantum development roadmap by 10 years to 2033 to build more powerful systems with better computing and error-correction capabilities.

In addition, by the end of 2024, IBM plans to establish eight quantum computing centers in the United States, Canada, Japan, and Germany to ensure widespread use of the Quantum System Two by researchers.

Gambetta also stated, "We need some time to transition from scientific value to commercial value. But in my view, the distinction between research and commercialization is becoming increasingly blurred."

IBM researchers have stated that recent progress has enhanced their confidence in the long-term potential of quantum computing, although they have not predicted when quantum computing will enter the mainstream commercial market.