As weve learned more and more about what we need and the different limits of all the components that make up a supercomputer, we were really able to say, If we could design our dream system, what would it look like? said OpenAI CEO Sam Altman. And then Microsoft was able to build it.
OpenAIs goal is not just to pursue research breakthroughs but also to engineer and develop powerful AI technologies that other people can use, Altman said. The supercomputer developed in partnership with Microsoft was designed to accelerate that cycle.
We are seeing that larger-scale systems are an important component in training more powerful models, Altman said.
For customers who want to push their AI ambitions but who dont require a dedicated supercomputer, Azure AI provides access to powerful compute with the same set of AI accelerators and networks that also power the supercomputer. Microsoft is also making available the tools to train large AI models on these clusters in a distributed and optimized way.
At its Build conference, Microsoft announced that it would soon begin open sourcing its Microsoft Turing models, as well as recipes for training them in Azure Machine Learning. This will give developers access to the same family of powerful language models that the company has used to improve language understanding across its products.
It also unveiled a new version of DeepSpeed, an open source deep learning library for PyTorch that reduces the amount of computing power needed for large distributed model training. The update is significantly more efficient than the version released just three months ago and now allows people to train models more than 15 times larger and 10 times faster than they could without DeepSpeed on the same infrastructure.
Along with the DeepSpeed announcement, Microsoft announced it has added support for distributed training to the ONNX Runtime. The ONNX Runtime is an open source library designed to enable models to be portable across hardware and operating systems. To date, the ONNX Runtime has focused on high-performance inferencing; todays update adds support for model training, as well as adding the optimizations from the DeepSpeed library, which enable performance improvements of up to 17 times over the current ONNX Runtime.
We want to be able to build these very advanced AI technologies that ultimately can be easily used by people to help them get their work done and accomplish their goals more quickly, said Microsoft principal program manager Phil Waymouth. These large models are going to be an enormous accelerant.
In self-supervised learning, AI models can learn from large amounts of unlabeled data. For example, models can learn deep nuances of language by absorbing large volumes of text and predicting missing words and sentences. Art by Craighton Berman.
Designing AI models that might one day understand the world more like people do starts with language, a critical component to understanding human intent, making sense of the vast amount of written knowledge in the world and communicating more effortlessly.
Neural network models that can process language, which are roughly inspired by our understanding of the human brain, arent new. But these deep learning models are now far more sophisticated than earlier versions and are rapidly escalating in size.
A year ago, the largest models had 1 billion parameters, each loosely equivalent to a synaptic connection in the brain. The Microsoft Turing model for natural language generation now stands as the worlds largest publicly available language AI model with 17 billion parameters.
This new class of models learns differently than supervised learning models that rely on meticulously labeled human-generated data to teach an AI system to recognize a cat or determine whether the answer to a question makes sense.
In whats known as self-supervised learning, these AI models can learn about language by examining billions of pages of publicly available documents on the internet Wikipedia entries, self-published books, instruction manuals, history lessons, human resources guidelines. In something like a giant game of Mad Libs, words or sentences are removed, and the model has to predict the missing pieces based on the words around it.
As the model does this billions of times, it gets very good at perceiving how words relate to each other. This results in a rich understanding of grammar, concepts, contextual relationships and other building blocks of language. It also allows the same model to transfer lessons learned across many different language tasks, from document understanding to answering questions to creating conversational bots.
This has enabled things that were seemingly impossible with smaller models, said Luis Vargas, a Microsoft partner technical advisor who is spearheading the companys AI at Scale initiative.
The improvements are somewhat like jumping from an elementary reading level to a more sophisticated and nuanced understanding of language. But its possible to improve accuracy even further by fine tuning these large AI models on a more specific language task or exposing them to material thats specific to a particular industry or company.
Because every organization is going to have its own vocabulary, people can now easily fine tune that model to give it a graduate degree in understanding business, healthcare or legal domains, he said.
Read the original post:
- As it closes in on Arm, Nvidia announces UK supercomputer dedicated to medical research - TechCrunch - October 8th, 2020
- With Crossroads Supercomputer, HPE Notches Another DOE Win - The Next Platform - October 8th, 2020
- What happens when two planets crash together? This supercomputer has the answer - Digital Trends - October 8th, 2020
- Supermicro Details Its Hardware for MN-3, the Most Efficient Supercomputer in the World - HPCwire - September 2nd, 2020
- I confess, I'm scared of the next generation of supercomputers - TechRadar - September 2nd, 2020
- Bradykinin Hypothesis of COVID-19 Offers Hope for Already-Approved Drugs - BioSpace - September 2nd, 2020
- Stranger than fiction? Why we need supercomputers - TechHQ - September 2nd, 2020
- Google Says It Just Ran The First-Ever Quantum Simulation of a Chemical Reaction - ScienceAlert - September 2nd, 2020
- This Equation Calculates the Chances We Live in a Computer Simulation - Discover Magazine - September 2nd, 2020
- 17 of the best computers and supercomputers to grace the planet - Pocket-lint - August 31st, 2020
- Supercomputer finds best way to air out classroom to ward off virus : The Asahi Shimbun - Asahi Shimbun - August 31st, 2020
- The Supercomputer Breaking Online Gaming Records and Modeling COVID-19 - BioSpace - August 31st, 2020
- When it comes to hurricane models, which one is best? - KHOU.com - August 31st, 2020
- Natural Radiation Including Cosmic Rays From Outer Space Can Wreak Havoc With Quantum Computers - SciTechDaily - August 31st, 2020
- The Tech Field Failed a 25-Year Challenge to Achieve Gender Equality by 2020 Culture Change Is Key to Getting on Track - Nextgov - August 31st, 2020
- Cerebras Systems Expands Global Footprint with Toronto Office Opening - HPCwire - August 31st, 2020
- CSC's Supercomputer Mahti is Now Available to Researchers and Students - HPCwire - August 28th, 2020
- Here's the smallest AI/ML supercomputer ever - TechRadar - August 28th, 2020
- When it comes to hurricane models, which one is best? - 12newsnow.com KBMT-KJAC - August 28th, 2020
- SberCloud's Cloud Platform Sweeps Three International Accolades At IT World Awards - Exchange News Direct - August 28th, 2020
- Supercomputer Market Growth, Future Prospects And Competitive Analysis (2020-2026) - Bulletin Line - August 28th, 2020
- A continent works to grow its stake in quantum computing - University World News - August 28th, 2020
- Supercomputer predicts where Spurs will finish in the 2020/21 Premier League table - The Spurs Web - August 28th, 2020
- Has the world's most powerful computer arrived? - The National - August 28th, 2020
- Galaxy Simulations Could Help Reveal Origins of Milky Way - Newswise - August 28th, 2020
- ALCC Program Awards Computing Time on ALCF's Theta Supercomputer to 24 projects - HPCwire - August 10th, 2020
- A Quintillion Calculations a Second: DOE Calculating the Benefits of Exascale and Quantum Computers - SciTechDaily - August 10th, 2020
- GE plans to give offshore wind energy a supercomputing boost - The Verge - August 10th, 2020
- From WarGames to Terms of Service: How the Supreme Courts Review of Computer Fraud Abuse Act Will Impact Your Trade Secrets - JD Supra - August 10th, 2020
- New Audis To Use Supercomputer That Controls Almost Everything - Motor1 - August 10th, 2020
- Japanese supercomputer ranked as worlds most powerful system - August 10th, 2020
- Top 10 Supercomputers | HowStuffWorks - August 10th, 2020
- What are supercomputers currently used for? | HowStuffWorks - August 10th, 2020
- GE taps into US supercomputer to advance offshore wind - reNEWS - August 10th, 2020
- Summit supercomputer to advance research on wind power for renewable energy - ZDNet - August 10th, 2020
- BSC Researchers Create Spin-Off Platform to Accelerate the Development of New Chemicals - HPCwire - August 10th, 2020
- Supercomputer study of mobility in Spain at the peak of COVID-19 using Facebook and Google data - Science Business - August 9th, 2020
- Julia and PyCaret Latest Versions, arXiv on Kaggle, UK's AI Supercomputer And More In This Week's Top AI News - Analytics India Magazine - August 9th, 2020
- Audi To Over-Complicate Cars With Supercomputers And Repair Costs Could Skyrocket - Top Speed - August 9th, 2020
- Supercomputer COVID-19 insights, ionic spiderwebs, the whiteness of AI TechCrunch - Best gaming pro - August 8th, 2020
- Every Superman Movie Climax, Ranked From Worst To Best - Screen Rant - August 8th, 2020
- Atos Partners with University of Oxford on Largest AI Supercomputer in the UK - HPCwire - August 7th, 2020
- Five Movies Worth Watching About the Threat of Nuclear War - Council on Foreign Relations - August 7th, 2020
- Break it down: A new way to address common computing problem - Washington University in St. Louis Newsroom - August 7th, 2020
- GE Research uses summit supercomputer for study on wind power - Windtech International - August 7th, 2020
- Research: A Survey of Numerical Methods Utilizing Mixed Precision Arithmetic - HPCwire - August 7th, 2020
- Atos signs 5m supercomputing deal to support Oxford University-led AI research push - ComputerWeekly.com - August 6th, 2020
- Atos partners with University of Oxford on largest AI supercomputer in the UK - Yahoo Finance - August 6th, 2020
- How coronavirus antibody testing works - Livemint - August 6th, 2020
- Researchers Use Supercomputers To Discover New Pathway For Covid-19 Inflammation - Forbes - August 6th, 2020
- Supercomputer-Powered Research Uncovers Signs of 'Bradykinin Storm' That May Explain COVID-19 Symptoms - HPCwire - July 31st, 2020
- Celtic and Rangers title race outcome predicted by betting supercomputer - HeraldScotland - July 31st, 2020
- Nvidia reportedly in advanced talks to buy Arm - ZDNet - July 31st, 2020
- NVIDIA Claims To Have Won MLPerf Benchmarking, But Google Says Otherwise - Analytics India Magazine - July 31st, 2020
- Continental is supercharging the development of driver-assistance tech - CNET - July 31st, 2020
- COVID-19 Pandemic Can Help More of Us Learn About Climate Change - UT News | The University of Texas at Austin - July 31st, 2020
- From rocks to icebergs, the natural world tends to break into cubes - Science Magazine - July 31st, 2020
- New Data on Genetic Expression In Severe COVID-19, Pre-Existing Immune Response - Bio-IT World - July 31st, 2020
- Superman's Glasses Are Secretly Used For Mind Control - Screen Rant - July 31st, 2020
- The Israeli company that has come as close as possible to the sun - Haaretz.com - July 31st, 2020
- Celtic and Rangers title race outcome predicted by betting supercomputer - Glasgow Times - July 31st, 2020
- PEARC20 Plenary Introduces Five Upcoming NSF-Funded HPC Systems - HPCwire - July 31st, 2020
- NIH Awards $6M to UConn Health Biological Computer Modeling Teams - HPCwire - July 31st, 2020
- Continental Debuts the Fastest Supercomputer in the Automotive Industry and It's Built for AI - EnterpriseAI - July 30th, 2020
- WATCH: Supercomputer generates 3D videos that show how Earth may have lost half of its atmosphere to create th - Business Insider India - July 26th, 2020
- Repeated intelligence failures: Time to worry - The Sunday Guardian - July 26th, 2020
- Supercomputer Market to witness an impressive growth during the forecast period 2020 - 2026 - CueReport - July 26th, 2020
- What is supercomputer? - Definition from WhatIs.com - July 26th, 2020
- What is supercomputer? - Definition - July 26th, 2020
- Impact of Covid-19 on Supercomputer Market Comprehensive Growth 2020-2027 with Top key vendor IBM Corporation, Cray Inc., Lenovo Inc., Sugon, Inspur -... - July 25th, 2020
- Super-computer Henry Cavill breaks the Internet again with more geek content - KSRO - July 24th, 2020
- How Equity Is Lost When Companies Hire Only Workers With Disabilities - The New York Times - July 24th, 2020
- Earth may have lost half of its atmosphere to create Moon, reveals study using 3D videos - Republic World - Republic World - July 24th, 2020
- "Super"-computer: Henry Cavill breaks the Internet again with more geek content - wcsjnews.com - July 24th, 2020
- Solar Opposites EPs Tease Whats To Come On Season 2 Of Hulu Animated Comedy Comic-Con@Home - Deadline - July 24th, 2020
- NVIDIA and University of Florida Release New AI Curriculum Spanning All Educational Disciplines - Motley Fool - July 21st, 2020
- Supercomputing Pipeline Aids DESI's Quest to Create 3D Map of the Universe - HPCwire - July 21st, 2020
- "Super"-computer Henry Cavill breaks the Internet again with more geek content - 1310kfka.com - July 20th, 2020
- It's time to decide what we want Downriver's future to look like - Southgate News Herald - July 20th, 2020
- The worlds supercomputers joined forces against COVID-19 why such collaborations are critical for tackling future emergencies - The European Sting - July 20th, 2020