Microsoft Research (@MSFTResearch)

Microsoft Research @MSFTResearch

Founded in 1991, Microsoft Research is dedicated to conducting both basic and applied research in computer science and software engineering.

Tweets by Microsoft Research

The @MSFTQuantum team has created a new class of quantum error correction codes, called Floquet codes. Learn how the codes lower spacetime overhead by 10 times and substantially reduce requirements for a key process in the topological quantum stack: https://t.co/I2fSkyI1pL

Shared by Microsoft Research at 5/5/2022

People spend 250% more time in meetings than before the remote work transition. Discover how asynchronous collaboration can help workers get what they need to stay productive—even when they choose to skip a meeting—in the New Future of Work Report 2022: https://t.co/z1F6Ml101q

Shared by Microsoft Research at 5/5/2022

AI-powered robotic boat Clearbot scoops up plastic trash floating in Hong Kong’s harbor before it can get into the ocean. https://t.co/NWO0Zf8WaR

Shared by Microsoft Research at 5/2/2022

Designing new pharmaceuticals in the lab requires knowledge, creativity, experience, and intuition—and it can take years. Discover how Microsoft Research and Novartis are collaborating to help chemists streamline this complicated process: https://t.co/CeylMSesWM

Shared by Microsoft Research at 4/27/2022

PPE is the first reinforcement learning algorithm that can solve the problem of exogenous noise with a mathematical guarantee. Learn about the conceptualization of latent state discovery that works even in an environment with severe distractions: https://t.co/0JviYAqQ3q

Shared by Microsoft Research at 4/25/2022

Explore how Urban Innovation research teams at Microsoft are using sensing, analytics, and community engagement to empower communities to build capacity for climate resilience: https://t.co/9ObjDG1cwn @RheaplyInc @ELPCenter

Shared by Microsoft Research at 4/23/2022

Today in collaboration with @MIT and @Cornell we announce a new SOTA for unsupervised Semantic Segmentation. “STEGO” discovers the world’s objects and classifies every pixel of input images: https://t.co/kkDQCMQs6g

Shared by Microsoft Research at 4/21/2022

The research community at Microsoft is proud to partner with the @NSF RINGS—Resilient and Intelligent Next-Generation Systems—program. Microsoft’s efforts include advances in edge computing, AI, telecommunications, and cloud technologies: https://t.co/JjJcAdXJ7N

Shared by Microsoft Research at 4/21/2022

Edge computing requires minuscule compute footprints that are vulnerable to data drift—where live data diverges from training data. Learn how Ekya addresses this problem by enabling model retraining and inference to co-exist on the edge box: https://t.co/aCjm2cYn8G

Shared by Microsoft Research at 4/19/2022

Microsoft Chief Technology Officer @Kevin_Scott shared his thoughts on AI & Society in the Spring 2022 issue of Daedalus. The essay explores how today’s AI advances, like transfer learning and platform models, could spur a future where AI complements people’s creativity.

Shared by Microsoft Research at 4/18/2022

Microsoft is committed to delivering best-in-class productivity experiences. Explore how the Systems Innovation initiative is aiming to combine deep workload understanding with hardware & algorithmic research & novel AI/ML techniques to achieve this goal: https://t.co/vTf1PwN1eG

Shared by Microsoft Research at 4/5/2022

Microsoft Research has joined LinkedIn! Follow us to keep up with the latest advances, collaborations, and discussions centered around cutting-edge science and technology research: https://t.co/4CQXxRbiC7.

Shared by Microsoft Research at 4/5/2022

A collaboration between Princeton University and Microsoft Research initially sought to better identify digital fingerprints of political disinformation. Learn how researchers pivoted to use the same tools to address misinformation on COVID-19: https://t.co/6XljWjMbt7

Shared by Microsoft Research at 4/1/2022

AI can help write code, but often with errors that developers must fix. Learn how Jigsaw, a multi-modal tool that augments large language models with program analysis and synthesis methods, fixes bugs in machine-written software: https://t.co/Cjf3GpGEMI

Shared by Microsoft Research at 3/31/2022

We show how the proposed approach could improve the quality of X-ray image annotations in an efficient manner. We are releasing a new set of image labels and a benchmark to the research community to evaluate their own data-driven annotation methods.

Shared by Microsoft Research at 3/24/2022

We hope that our benchmark and noisy/clean label set can empower researchers and developers to develop their own methods to speed up image annotation even more, saving time for experts in the field working on ML projects. Explore the repo: https://t.co/xTOIQnljAj.

Shared by Microsoft Research at 3/24/2022

In this image, we see cat images sampled from an internet dataset (CIFAR10) that are easy to label versus ones that are hard to label. The plot shows the distribution of how likely a lower quality image of a cat can be mislabeled as a dog by human annotators.

Shared by Microsoft Research at 3/24/2022

In a @naturecomms paper, Microsoft researchers & collaborators show how to speed up image label correction times by 400% to remove label noise and make better use of experts’ time—such as clinicians labeling images for machine learning research projects: https://t.co/0lkoqVcBXg.

Shared by Microsoft Research at 3/24/2022

The 2022 VizWiz Grand Challenge Workshop launched 3 machine learning challenges with Microsoft-sponsored awards. Tasks include visual question answering & few-shot object recognition on images & videos taken by people who are blind. Submissions due May 6: https://t.co/mQXOHGtWdu

Shared by Microsoft Research at 3/24/2022

Tomorrow at 12pm ET on Physics ∩ ML, Greg Yang explains how to tune GPT-3 on a single GPU based on the recent paper "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer." Sign up here and bring any questions: https://t.co/89XjiwVfeR

Shared by Microsoft Research at 3/22/2022

Mixture of Experts help different parts of AI models learn different tasks and translate between multiple languages. Learn how Z-code MoE models are making @MStranslator better & faster in the first MoE deployment for a machine translation provider: https://t.co/Yl6kMEdyjo

Shared by Microsoft Research at 3/22/2022

Topological qubits are controlled by a fundamental property called the topological gap. Learn how Microsoft Quantum researchers have observed a 30 μeV topological gap, showing that the underlying physics behind their approach to quantum computing is sound: https://t.co/6g2NC6jfnl

Shared by Microsoft Research at 3/14/2022

Apply for Reinforcement Learning Open Source Fest 2022: you’ll get to collaborate with our Real World RL Team and will receive a $10K stipend. Learn more and apply: https://t.co/yXLJdO0fSB

Shared by Microsoft Research at 3/9/2022

Microsoft researchers are releasing SynapseML v0.9.5 on GitHub. Explore the release, with scalable multivariate anomaly detection, text to speech, updated analytics features, and new responsible AI tools such as Individual Conditional Expectation analysis: https://t.co/xNoToHF4VR

Shared by Microsoft Research at 3/8/2022

When a neural network is too large to pretrain more than once, tuning its hyperparameters is practically impossible. Today, we announce μTransfer—a new technique that can tune the 6.7 billion parameter GPT-3 model using only 7% of the pretraining compute: https://t.co/RnS5HZboq0

Shared by Microsoft Research at 3/8/2022

COMPASS builds a general-purpose multimodal pretrained model for autonomous systems. Learn how the representation can address a variety of application scenarios, such as drone navigation, visual odometry and auto racing: https://t.co/wFX5IzrGtF

Shared by Microsoft Research at 2/23/2022

We're looking forward to being a part of the "Out-of-distribution generalization and adaptation in natural and artificial intelligence" #NeurIPS2021 workshop on December 14. Sign up to hear from Yoshua Bengio, @AnimaAnandkumar, @wellingmax, and more: https://t.co/Bd92tDpt0Y

Shared by Microsoft Research at 12/9/2021

We're excited to be sponsoring the 4th Annual @ApacheTVM Open Source ML Acceleration Conference! Join the TVM community this Dec 15-17 for a free event covering #DeepLearning optimization, with tutorials, research talks, and case studies. Register now at https://t.co/BApVE4IR6C!

Shared by Microsoft Research at 12/8/2021

Microsoft researcher @SebastienBubeck & his student @geoishard received an Outstanding Paper Award at #NeurIPS2021 for their work to prove a universal law of robustness & the necessity of overparametrization in deep neural networks. Read the paper: https://t.co/NcnB0upQNP

Shared by Microsoft Research at 12/1/2021

Microsoft researchers developed a method for training generative models—Manifold Matching via Metric Learning—that unlocks the ability to use latest breakthroughs in metric learning. Learn how the method can accelerate generative model research: https://t.co/vXh49BSDJr #ICCV2021

Shared by Microsoft Research at 11/29/2021

Microsoft Research welcomes Visiting Researcher, Prof Frank Noé (@FrankNoeBerlin), a leading expert in machine learning for the natural sciences. Learn more about opportunities across our global labs. https://t.co/Gjxnl4pQkj

Shared by Microsoft Research at 11/4/2021

Turing Bletchley can perform image-language tasks in 94 languages using a common image-language vector space, without any metadata or surrounding text. This Universal Image Language Representation model represents a breakthrough in deep learning. https://t.co/Jy1EXPcEdI

Shared by Microsoft Research at 11/1/2021

Build a strong foundation for your career by working on real-world projects using our cutting-edge technologies. Make valuable connections and have lots of fun along the way! Discover Internship opportunities at Microsoft Research Cambridge UK: https://t.co/J1rwAnxagK

Shared by Microsoft Research at 11/1/2021

At Microsoft Research, we aim to empower the next generation of exceptional computing research talent. Today, we're thrilled to announce and congratulate this year's Microsoft Research PhD Fellowship recipients from around the world: https://t.co/pg6IpkiDEf

Shared by Microsoft Research at 10/20/2021

Join the researchers to learn more about this work and how we can improve digital health delivery at the Microsoft Research Summit, Oct 19–21. Open to all: https://t.co/YocrCPJfMY

Shared by Microsoft Research at 10/19/2021

What does it take to build a better digital health system? Microsoft researchers partnered with @GlobalFund & @StopTB on a report highlighting lessons learned across 13 countries. Discover what the report says about improving digital health systems: https://t.co/DT2458rzZ9

Shared by Microsoft Research at 10/19/2021

Amit Sharma, Senior Researcher, Microsoft Research

Shared by Microsoft Research at 10/19/2021

The need for collaborative research that addresses global challenges has never been greater. The first ever Microsoft Research Summit convenes more than 300 world-leading researchers and engineers, October 19-21. Register today and join in. Open to all: https://t.co/pokwff2C36

Shared by Microsoft Research at 10/18/2021

In the Research for Industry track at Microsoft Research Summit, join @RanveerChandra—along with members of the National Academies, senior executives in industry & a Nobel laureate—to learn how tech can help industry be more efficient & create opportunity: https://t.co/PsTnBN5RnW

Shared by Microsoft Research at 10/17/2021

The inaugural 2021 Microsoft Research Summit starts next week! Join us on Oct 19–21 for three days of scientific discovery and technology innovation with experts from around the globe. Open to all! Register at: https://t.co/xCHKkSWFt4

Shared by Microsoft Research at 10/15/2021

Research at Microsoft is one key to finding sustainable pathways for businesses and organizations. Join @LucasJoppa & other experts to explore how technology can support this journey toward becoming carbon negative at Microsoft Research Summit, Oct 19–21: https://t.co/mwoGLI99ZK

Shared by Microsoft Research at 10/12/2021

The promises and challenges of causal machine learning

Shared by Microsoft Research at 10/8/2021

Join Peter Lee, CVP of Microsoft Research & Incubations, for some reflections on the motivation and inspiration behind the work we do in research, and what we can expect to see in Microsoft Research’s fourth decade. Register now: https://t.co/CsCyXmhZNO

Shared by Microsoft Research at 10/7/2021

How does technology combat inequities in healthcare? How does it increase access to care & promote health resilience at all levels of society? Oct 19–21, join the first Microsoft Research Summit & discover how we partner to achieve better health for all: https://t.co/GGljzRloyM

Shared by Microsoft Research at 10/6/2021

Reinforcement learning has promise for human-oriented AI, industrial applications like logistics, & real-world impact in everything from gaming to robotics. Learn how research is bridging the gap between theory & applications at Microsoft Research Summit: https://t.co/YYX7i8W635

Shared by Microsoft Research at 9/30/2021

How do AI and machine learning make the cloud better? Join researchers, academia, and industry at Microsoft Research Summit, Oct 19–21, to learn how AI and machine learning can improve cloud performance, security, and sustainability. Register now: https://t.co/Uu8PK7Iqao

Shared by Microsoft Research at 9/30/2021

The need to build a more resilient society has never been more pressing. Learn how researchers work with community partners to develop technology that drives more inclusive innovations at the inaugural Microsoft Research Summit. Open to all: https://t.co/hkflgIG05L

Shared by Microsoft Research at 9/28/2021

Microsoft Turing Universal Language Representation model, T-ULRv5, has topped the XTREME leaderboard for evaluating cross-lingual generalization. Learn how recent advances enable T-ULRv5 to not only excel beyond previous models, but also train 100x faster: https://t.co/gX4ooVYGFX

Shared by Microsoft Research at 9/28/2021

Discover the research being done to reach carbon negative & what’s still needed. At the Microsoft Research Summit, learn the key problems we need to solve to limit Earth's warming & how international experts are contributing to the solutions. Register: https://t.co/YzS4HYjfKZ

Shared by Microsoft Research at 9/28/2021

Explore the latest technological advances in healthcare, biology, biotechnology & precision medicine. At the Microsoft Research Summit, learn how Microsoft researchers & global experts are making strides in achieving a healthy global society. Register: https://t.co/BjAzDDHdF1

Shared by Microsoft Research at 9/27/2021

Open to everyone! The first-ever 2021 Microsoft Research Summit, Oct 19 - 21, with over 150 sessions across 16 tracks, provides the global research community with an opportunity learn from experts pushing the frontiers of technology. Register now: https://t.co/v93MdVj6QJ

Shared by Microsoft Research at 9/23/2021

Now the reveal—*drumroll* please. The QUESTION you’ve been waiting for: What is the Microsoft Research Summit? Registration for this inaugural event is now open to all. Join us Oct 19–21. Learn about the research covered in 150+ sessions and register now: https://t.co/sbkg3ffnzp https://t.co/fEIQPLvKD4

Shared by Microsoft Research at 9/23/2021

One of the greatest levers we have to achieve societal impact at scale is policy. Learn how researchers at Microsoft and their collaborators are building technology for evidence-based policies to make societies more resilient: https://t.co/sp67vqEs8m

Shared by Microsoft Research at 9/23/2021

Farming is a low-margin business. DeepMC improves micro-climate forecasting to make farming more efficient. Learn how this deep learning framework sorts through mountains of climate data to inform key decisions: https://t.co/B93d3GJ3J1 #KDD21 @RanveerChandra

Shared by Microsoft Research at 9/15/2021

DeepSpeed enables training of 8x bigger MoE models, uses fewer resources, and attains excellent throughput & near-linear scalability. Discover how we combine multidimensional parallelism and heterogenous memory technologies to train MoE models like Z-code: https://t.co/4zGHWqJSCk

Shared by Microsoft Research at 8/18/2021

Understanding molecules can help address pressing societal challenges. As researcher @wellingmax prepares to join Microsoft Research, he shares his vision for predicting molecular properties & searching the molecular space w/ @ChrisBishopMSFT: https://t.co/XzPytzHvrZ #MSRPodcast

Shared by Microsoft Research at 7/26/2021

The infinite-width limit is key to overparametrized neural networks, yet existing theories don’t allow feature learning in practical networks. Learn how @TheGregYang & @EdwardJHu use the Maximal Update Parameterization to unlock feature learning: https://t.co/e3WgThtw6G #ICML2021

Shared by Microsoft Research at 7/22/2021

Machine learning has the potential to disrupt the field of molecular simulation, and the societal gains can be huge. In this #MSRPodcast, @ChrisBishopMSFT explores the science and opportunities with new hire and leading ML researcher @wellingmax: https://t.co/XzPytzHvrZ

Shared by Microsoft Research at 7/20/2021

More broadly, this work demonstrates the potential of integrating domain knowledge with deep learning models to improve our prediction of nucleic acid interactions. Full paper: https://t.co/hKu1sibZ7J

Shared by Microsoft Research at 7/19/2021

Migrating telecom networks to the cloud was just an intriguing idea--until Microsoft researchers showed how it could be done over @Azure. Today, software and services are replacing physical equipment and enabling world-class service: https://t.co/fG8QK3EQPc

Shared by Microsoft Research at 7/19/2021