A Generative adversarial network, or GAN, is one of the most powerful machine learning models proposed by Goodfellow et al. for learning to generate samples from complicated real-world distributions. GANs have sparked millions of applications, ranging from generating realistic images or cartoon characters to text-to-image translations. Turing award laureate Yann LeCun called GANs “the most interesting idea in the last 10 years in ML.” In the context of generating images, GANs consist of two parts. 1) A parameterized (deconvolutional) generator […]

Episode 123 | June 10, 2021 In the world of economics, researchers at Microsoft are examining a range of complex systems—from those that impact the technologies we use to those that inform the laws and policies we create—through the lens of a social science that goes beyond the numbers to better understand people and society. Interviewed by Senior Principal Researcher Hunt Allcott, Economist David Rothschild discusses how the news media has evolved alongside social media  

AI agents capable of understanding natural language, communicating, and accomplishing tasks hold the promise of revolutionizing the way we interact with computers in our everyday lives. Text-based games, such as the Zork series, act as testbeds for development of novel learning agents capable of understanding and interacting exclusively through language. Beyond requiring the use of imagination and myriad concepts of everyday life to solve, these fictional world–based narratives are also a safe sandbox for AI testing that avoids the expense […]

Episode 122 | June 2, 2021  In the world of economics, researchers at Microsoft are examining a range of complex systems—from those that impact the technologies we use to those that inform the laws and policies we create—through the lens of a social science that goes beyond the numbers to better understand people and society.  In this episode, Senior Principal Researcher Dr. Hunt Allcott speaks with Microsoft Research New England office mate and  

“If you think about the world that we’re in now, software plays a key role in our lives – in everything we do. And at the core of solving big problems and fundamental challenges is the ability to use software to attack the problem and make it better. I work in the field of programming languages, and we go to work thinking ‘what do we do to make software better?’ Because if we make software better, then we can make […]

Last month, the DeepSpeed Team announced ZeRO-Infinity, a step forward in training models with tens of trillions of parameters. In addition to creating optimizations for scale, our team strives to introduce features  

Accessibility and inclusion represent a growing space in the technology landscape, and how research and development are being used to empower people across abilities is expanding in exciting ways. Instead of treating disabilities as conditions in need of solutions—as has been the case over the years with the medical-based approach to accessibility—research is moving toward a social approach, examining and addressing the societal expectations imposed on people with disabilities and society’s lack of inclusion. The world comprises people of various […]

The genetic sequence for COVID-19 was first published in January 2020. Before the end of the year, new vaccines—which typically take five to ten years to develop—were approved for emergency use in multiple nations. This astounding achievement is emblematic of a “reset” in the relationship between science and society. Faced with a frightening health crisis, the world called on scientists and researchers  

Episode 121 | May 19, 2021 In the world of economics, researchers at Microsoft are examining a range of complex systems—from those that impact the technologies we use to those that inform the laws and policies we create—through the lens of a social science that goes beyond the numbers to better understand people and society. In this episode, Dr. Hunt Allcott, Senior Principal Researcher at Microsoft Research New England, talks with 

Understanding video is one of the most challenging problems in AI, and an important underlying requirement is learning multimodal representations that capture information about objects, actions, sounds, and their long-range statistical dependencies from audio-visual signals. Recently, transformers have been successful in vision-and-language tasks such as image captioning and visual question answering due to their ability to learn multimodal contextual representations. However, training multimodal transformers end-to-end is difficult because of the excessive memory requirement. In fact, most existing vision and language transformers […]