Microsoft Research Blog

Research Blog

  4. An icon of people labeled with the pro “know the task” and the con “slow” followed by a plus sign and an icon representing AI labeled with the pro “fast” and the con “unaware of task” followed by an equal sign and an icon of shaking hands. An outer arrow labeled “The AI generates tests to highlight bugs” points up and around from the AI icon to the people icon. An arrow labeled “The person validates which bugs are real” points down and around from the people icon to the AI icon, representing an iterative feedback loop.

    Partnering people with large language models to find and fix bugs in NLP systems 

    May 23, 2022 | Scott Lundberg and Marco Tulio Ribeiro

    Advances in platform models—large-scale models that can serve as foundations across applications—have significantly improved the ability of computers to process natural language. But natural language processing (NLP) models are still far from perfect, sometimes failing in embarrassing ways, like translating “Eu não recomendo este prato”…

  6. This diagram shows a payload exchange between a server, inside Worker 0, and clients that live inside Workers 2 and 3. First, the server pushes the central ML model plus the clients’ data to Workers 2 and 3. Then, each client trains the model with their local data. Finally, the clients send the pseudo-gradients of this new model back to the server for aggregation and the creation of a new global model.

    FLUTE: A scalable federated learning simulation platform 

    May 16, 2022

    Federated learning has become a major area of machine learning (ML) research in recent years due to its versatility in training complex models over massive amounts of data without the need to share that data with a centralized entity. However, despite this flexibility and the…

  9. A view of a park with an agent walking along a trail. Sources of exogenous noise surround the agent, including ducks gliding on a pond, people in the background, and reflections on the water.

    PPE: A fast and provably efficient RL algorithm for exogenous noise 

    April 25, 2022 | Dipendra Misra and Yonathan Efroni

    Picture a person walking in a park by a pond. The surrounding environment contains a number of moving objects that change the quality of the environment: clouds moving to hide the sun, altering the quality of light; ducks gliding across the pond, causing its surface…

  10. Diagram showing Ekya’s architecture. Video data flows from a series of cameras into specialized, lightweight inference models and shared resource pools before reaching the edge.

    Don’t let data drift derail edge compute machine learning models 

    April 19, 2022

    Edge computing has come of age, with deployments enabling many applications that process data from IoT sensors and cameras. In 2017, we identified the symbiotic relationship between edge computing and video analytics in an article (opens in new tab), noting that live video analytics is…

  11. A flowchart showing inputs pre-processed before being fed into large language models including GPT-3, Codex, and others. The post-process output is returned to the end-user for verification. If they find the output incorrect, it is edited by them, and the learning is fed back into the pre-process and post-process mechanisms to improve them further.

    Jigsaw fixes bugs in machine-written software 

    March 31, 2022

    Large pre-trained language models such as GPT-3, Codex, and others can be tuned to generate code from natural language specifications of programmer intent. Such automated models have the potential to improve productivity for every programmer in the world. But since the models can struggle to…