Watch Google Quantum AI Reveal Willow Quantum Computing Chip – Video
Watch Google Quantum AI demonstrate Willow quantum computing chip
Introducing our newly developed quantum computing chip that can learn and evolve like the natural world around us. Willow from Google Quantum AI. Hello everyone, I am Julian Kelly, Director of Hardware at Google Quantum AI. Today, on behalf of our amazing team, I’m proud to announce Willow Willow is Google’s newest and most powerful superconducting quantum computing chip. and our next steps in building large-scale quantum computers and exploring applications. I’ve been fascinated by quantum computing since I first tried Cubis in 2008. Since joining Google in 2015, we’ve dreamed of making our mission a reality: building quantum computers to solve otherwise unsolvable problems. We launched our first chip, Foxtail, in 2017, followed by Bristol Cohen in 2018, and Sycamore in 2019, which fueled our milestone as the first computer to perform random circuit sampling on the computing task. Quantum computers that surpass the best classical supercomputers, and for many years we have used sycamore for random circuit sampling. Ability to squeeze significant performance from our hardware, including enabling scalable logic within our milestones. But we are ultimately limited by quantum coherence times the length of time the cubism remains in its intended state. With Willow, we’ve taken a big step forward. We increased the quantum coherence time fivefold, from 20 microseconds for Sycamore to 100 microseconds for Willow. We did all this without sacrificing any of the features that make our system so successful. This advancement is enabled by our new purpose-built superconducting quantum chip manufacturing facility in Santa Barbara, one of only a few in the world. We’re seeing exciting progress with Willow beyond Sycamore’s groundbreaking demo. Our logic qubits now operate below critical quantum error correction thresholds. The theory has been a highly sought-after target in quantum computing since its discovery in the 1990s. We have achieved this for the first time, and as the error rate is halved, errors in our logic qubits will be suppressed exponentially. Each time we add physical qubits at a distance of 3 to 5 to 7 surface coatings. Furthermore, the lifetimes of our logical qubits are now much longer than the lifetimes of all the physical qubits that compose them. This means that even if we make the quantum displacement larger and more complex by adding more toggle, we can actually improve its accuracy using quantum error correction. We benchmarked Willow against one of the world’s most powerful supercomputers on random circuit sampling. The results are quite surprising, and according to our best estimates, Willow’s calculation in 5 minutes would take the fastest supercomputers 10 to 25 years. This is a 1 followed by 25 zeros, or a time scale much longer than the age of the universe. The results highlight the exponentially growing gap between classical and quantum computing in certain applications. Let’s talk about the hardware approach. We’re leading the way at Google’s quantum artificial intelligence to make these things possible. Our returnable toggle and couplers enable ultra-fast gates and operations for low error rates, are reconfigurable to optimize in-situ hardware and run multiple applications and high connectivity for efficient expression of algorithms. We leverage this tuning capability to achieve repeatable high performance across the entire device. Let me explain one challenge with superconducting toggle joints is that not all toggle joints are created equal, some are outliers with unusually high ears. But this is where our trainable elbow joint really comes into play. We can fix these abnormal toggle joints by reconfiguring them to be consistent with the rest of the device. We can go a step further and have our researchers leverage tuning capabilities to continuously develop new calibration strategies to reduce errors across all toggle joints via software. Let’s quantify this and take a moment. In quantum computer specifications, our number of toggle connections is the average number of times each Cuba can interact with its neighbors. We quantify the error probability of running simultaneous operations, a single cubic gate, two cubic gates, measure the coherence time, which measures how long each qubit can retain its information, and measure the rate, which is the number of calculations we can run per second. The performance of an application is the performance of an entire system. benchmark. Willow has the best spot on the entire list. It has plenty of toggle, high connectivity, and can run a variety of applications. We measured low average error rates across all operations using multiple native two-cube gates. We have significantly doubled that, we have a very high measurement rate and Willow is below the error correction threshold and performs random circuit sampling. Willow looks to the future of possibilities well beyond classical computers. We continue to work toward building large-scale and useful error-corrected quantum computers that will push the boundaries of science and nature exploration and enable future commercial applications in areas such as pharmaceuticals, batteries, and fusion energy. We are excited to solve problems that cannot be solved tomorrow.
