• July 2020
    M T W T F S S

Scientists Succeed in Measuring Electron Spin Qubit Without Demolishing It RIKEN

March 3, 2020

Scientists at the RIKEN Center for Emergent Matter Science in Japan have successfully measured electron spin in a silicon quantum dot (QD) without altering its spin. Such measurements have been difficult, as spin typically is impacted by the process of reading out single electron spins in silicon, which is to convert the spins into charges for fast detection. The researchers used the Ising model to transfer spin data of an electron in a QD to an electron in a neighboring QD; they then were able to measure the neighbor’s spin while leaving the original electron untouched. Using this method, said RIKEN’s Seigo Tarucha, yielded a non-demolition fidelity rate of 99%, a readout accuracy of 95%, and a theoretical accuracy upgrade to 99.6%.


How the Cloud Has Opened Doors for Hackers

The Washington Post
Craig S. Smith
March 2, 2020

Corporate transfers of operations to the cloud have elevated the threat of hacking, as the cloud can be accessed remotely with ease. Manav Mital, co-founder of cloud security startup Cryal, said cloud companies manage the upkeep and security of physical servers, but client requirements for ease of access have spawned new apps and databases, and increasingly complex services that are difficult to manage and monitor. Although companies still shield private data behind firewalls and other security measures, more people and programs require access to data in the cloud, making it easier for bad actors to find potential vulnerabilities. The Ponemon Institute estimated that cloud breaches cost each individual company $3.92 million on average.


Cal Receives Largest Donation Ever—$252M—for Datacenter

San Francisco Chronicle
Michael Cabanatuan
March 2, 2020

The University of California, Berkeley (UC Berkeley) received an anonymous $252-million donation, which will be used to break ground on the Data Hub, a facility for students and faculty studying computing and data science. This gift is the largest donation ever given to the school, while the datacenter’s completion will require another $300 million. The Data Hub will contain the university’s Division of Computing, Data Science, and Society. UC Berkeley’s Jennifer Chayes said the facility “will be a magnet, bringing together scholars from disciplines across campus to forge new collaborations and take on some of the most critical questions facing society today, from biomedicine, to climate change and sustainability, to making data-informed public policy on issues of societal significance.”


ZeRO & DeepSpeed: New system optimizations enable training models with over 100 billion parameters

February 13, 2020 |

By DeepSpeed Team ;

Rangan Majumder; Junhua Wang

The latest trend in AI is that larger natural language models provide better accuracy; however, larger models are difficult to train because of cost, time, and ease of code integration. Microsoft is releasing an open-source library called DeepSpeed, which vastly advances large model training by improving scale, speed, cost, and usability, unlocking the ability to train 100-billion-parameter models. DeepSpeed is compatible with PyTorch. One piece of that library, called ZeRO, is a new parallelized optimizer that greatly reduces the resources needed for model and data parallelism while massively increasing the number of parameters that can be trained. Researchers have used these breakthroughs to create Turing Natural Language Generation (Turing-NLG), the largest publicly known language model at 17 billion parameters, which you can learn more about in this accompanying blog post.



Researchers Simulate 61-Bit Quantum Computer with Data Compression

January 23, 2020

Researchers at the University of Chicago and the U.S. Department of Energy’s Argonne National Laboratory used data compression to fit a 61-quantum bit (qubit) simulation of Grover’s quantum search algorithm on a supercomputer with 0.4% error. The researchers also modeled other quantum algorithms with significantly more qubits and quantum gates than previously possible. Although classical quantum-circuit simulation is essential for better understanding the operations and behaviors of quantum computation, the current full-state simulation limit is 48 qubits. This research offers a new tool for scaling quantum-circuit simulation by applying lossless and lossy data compression to the state vectors.


Supercomputer Simulations Reveal Details of Galaxy Clusters

UC San Diego News Center
Jorge Salazar; Jan Zverina
January 24, 2020

A multi-institutional study led by University of California, San Diego (UCSD) researchers developed high-resolution computer simulations of galactic clusters. The RomulusC models apply black-hole physics to examine the molecular gas within and enclosing the intracluster medium, using ultraviolet light from quasars shining through the gas to explore the clusters in greater detail. The researchers used the Trident software tool to render synthetic absorption line spectra into a realistic spectrum for direct comparison to existing observations. The scientists also tapped four supercomputers, including UCSD’s Comet system, to model the clusters. Said the University of Washington’s Iryna Butsky, “Simulations are extremely important in being able to make progress in theoretical work.”


The Case of the Elusive Majorana: So-Called ‘Angel Particle’ Still a Mystery Penn State News

Sam Sholtis
January 3, 2020

A study by researchers at Pennsylvania (Penn) State University and Germany’s University of Wurzburg casts doubt on the reported discovery of the chiral Majorana fermion in 2017. Analog versions of Majorana fermions are considered a potential pathway for building a topological quantum computer, with quantum bits shielded from environmental decoherence. The researchers analyzed dozens of devices similar to the one used to generate the so-called angel particle in the 2017 report. They found that the feature claimed to be the manifestation of the Majorana fermion was unlikely to be triggered by its existence. Said Penn State’s Cui-Zu Chang, “An important first step toward this distant dream of creating a topological quantum computer is to demonstrate definitive experimental evidence for the existence of Majorana fermions in condensed matter. Over the past seven or so years, several experiments have claimed to show such evidence, but the interpretation of these experiments is still debated.”


Intel’s Jim Clarke on its New Cryo-controller and why Intel isn’t Late to the Quantum Party

By John Russell

December 9, 2019

Intel today introduced the ‘first-of-its-kind’ cryo-controller chip for quantum computing and previewed a cryo-prober tool for characterizing quantum processor chips. The new controller is a mixed-signal SoC named Horse Ridge after one of the coldest regions in Oregon and is designed to operate at approximately 4 Kelvin.


Python Programming Language, AWS Skills Demand Has Exploded

Liam Tung
November 20, 2019

Analysis of Indeed.com job search engine listings over the last five years found explosive growth in demand for skills in Python, showing the coding language is the most popular one, or on its way to becoming most popular. Job listings mentioning Python climbed from 8% in September 2014 to 18% in September 2019, with the upsurge often credited to growth in data science and interest in machine learning and artificial intelligence, helped by abundant third-party Python packages and developer tools. Indeed also revealed skyrocketing demand for developers with Amazon Web Services (AWS) skills, with about 14% of current listings calling for AWS knowledge. Indeed Hiring Lab economist Andrew Flowers said, “A big reason behind the exceptional growth of Python and AWS is that the underlying tech job mix is changing in ways that favor these programming languages.”


World’s Fastest Supercomputers Hit Higher Speeds with Linux

Steven J. Vaughan-Nichols
November 18, 2019

The latest Top500 supercomputer ratings found Linux-powered supercomputers currently average 1.14 petaflops in speed, with the Oak Ridge National Laboratory’s Summit system topping the list with a High-Performance Linpack benchmark of 148.6 petaflops. Ranked second is Lawrence Livermore National Laboratory’s Sierra system, with a speed of 94.6 petaflops, despite using the same Power9 central-processing units and Nvidia Tesla V100 graphic-processing units as Summit. Although China owns nearly half the world’s fastest supercomputers, U.S. systems have a 37.8% aggregate performance share to China’ 31.9%. The Green500 list of the most energy-efficient supercomputers ranked the A64FX prototype the top system, with 16.9 gigaflops/watt. Meanwhile, HDR InfiniBand-based machines account for 40% of the Top500’s aggregate performance, and Ethernet-based systems account for 29%.