Collaborating Authors

Microsoft researchers say NLP bias studies must consider role of social hierarchies like racism

AITopics Custom Links

As the recently released GPT-3 and several recent studies demonstrate, racial bias, as well as bias based on gender, occupation, and religion, can be found in popular NLP language models. But a team of AI researchers wants the NLP bias research community to more closely examine and explore relationships between language, power, and social hierarchies like racism in their work. Published last week, the work, which includes analysis of 146 NLP bias research papers, also concludes that the research field generally lacks clear descriptions of bias and fails to explain how, why, and to whom that bias is harmful. "Although these papers have laid vital groundwork by illustrating some of the ways that NLP systems can be harmful, the majority of them fail to engage critically with what constitutes'bias' in the first place," the paper reads. "We argue that such work should examine the relationships between language and social hierarchies; we call on researchers and practitioners conducting such work to articulate their conceptualizations of'bias' in order to enable conversations about what kinds of system behaviors are harmful, in what ways, to whom, and why; and we recommend deeper engagements between technologists and communities affected by NLP systems."

How to drive change with data


Optimization and The NFL's Toughest Scheduling Problem - June 23 At first glance, the NFL's scheduling problem seems simple: 5 people have 12 weeks to schedule 256 games over the course of a 17-week season. The scenarios are potentially well into the quadrillions. In this latest Data Science Central webinar, you will learn how the NFL began using Gurobi's mathematical optimization solver to tackle this complex scheduling problem.

Machine Learning in Coupon Recommendation


How multi-armed bandits can help Starbucks send personalized offers to its customers. Coupon systems have been widely used to enhance customers' engagement in digital-based platforms. Coupon systems have been widely used to enhance customers' engagement in digital-based platforms. By offering users a challenge and a corresponding reward, companies' services become not only more attractive, but most importantly it can lead users to become frequent customers, thus enhancing a brand's impact on its customers. However, knowing which coupon to provide can be a rather complex task since each customer profile responds differently to each offer, and frequently offering them bad deals might drag them away from your business.

Predicting the Stock price Using TensorFlow


A very simple graph that adds two numbers together. In the figure above, two numbers are supposed to be added. Those numbers are stored in two variables, a and b. The two values are flowing through the graph and arrive at the square node, where they are being added. The result of the addition is stored into another variable, c.

Microsoft and Sony to create smart camera solutions for AI-enabled image sensor ZDNet


Sony and Microsoft have joined together to create artificial intelligence-powered (AI) smart camera solutions to make it easier for enterprise customers to perform video analytics, the companies announced. Announced last week, the IMX500 is the world's first image sensor to contain a pixel chip and logic chip. The logic chip, called Sony's digital signal processor, is dedicated to AI signal processing, along with memory for the AI model. "Video analytics and smart cameras can drive better business insights and outcomes across a wide range of scenarios for businesses," said Takeshi Numoto, corporate vice president and commercial chief marketing officer at Microsoft. "Through this partnership, we're combining Microsoft's expertise in providing trusted, enterprise-grade AI and analytics solutions with Sony's established leadership in the imaging sensors market to help uncover new opportunities for our mutual customers and partners." Sony and Microsoft also announced that they will create a smart camera managed app powered by Azure Internet of Things (IoT) and cognitive services that it hopes to use alongside the IMX500 sensor to provide new video analytics use cases for enterprise customers.

Using AI to map marine environments


Sonar is commonly used to map the ocean floor, and seabed composition (e.g. Salinity, depth and water temperature also affect how sound waves are propagated through water. This means that sonar measurements at different depths and distances can give accurate soundings of the ocean's properties, for example how underwater currents propagate, how the deeper ocean changes with the climate or where best to listen to whales. Working with Systems Engineering & Assessment Ltd (SEA), scientists at the University's Institute for Mathematical Innovation (IMI) have developed an Artificial Intelligence (AI) algorithm which could improve underwater mapping by making sense of incomplete data and working out how many measurements are needed to give an accurate survey. The research was part of a project contracted by The Defence and Security Accelerator (DASA), a part of the Ministry of Defence, to improve monitoring of the UK's vast marine territories using high tech sonar.

Towards a Practical Neural-Symbolic Framework


Deep learning is still bearing fruits. However, the standard types of networks are exhausting their possibilities, and researchers seek out such extensions to the basic neural network models, which will weaken their inherent limitations. Some extensions such as self-attention layers have enjoyed great practical success. Remarkably, many shortcomings of neural networks mirror the advantages of symbolic systems (and vice versa). Indeed, one can note that both self-attention layers and capsule networks are attempts to work around the notorious variable binding problem described in the Fodor and Pylyshyn's paper, which is easily solved in symbolic systems but is very inconvenient for neural networks.

Deliveroo Chooses EduMe's Workforce Success Platform


Deliveroo and EduMe today announced an exclusive new global partnership that will drive the success of the food delivery giant with effective onboarding, training and continuous learning by using EduMe's platform. The initiative is being rolled out to Deliveroo's entire global network of riders. It will take advantage of EduMe's experience as the training provider of choice by other leading technology companies. This will help facilitate effective onboarding at scale for new riders. Furthermore, an integration with hiring platform Fountain will be leveraged to present a seamless engagement and onboarding experience for new riders.

Coronavirus Tests the Value of Artificial Intelligence in Medicine


Dr. Albert Hsiao and his colleagues at the University of California–San Diego health system had been working for 18 months on an artificial intelligence program designed to help doctors identify pneumonia on a chest X-ray. When the coronavirus hit the United States, they decided to see what it could do. The researchers quickly deployed the application, which dots X-ray images with spots of color where there may be lung damage or other signs of pneumonia. It has now been applied to more than 6,000 chest X-rays, and it's providing some value in diagnosis, said Hsiao, the director of UCSD's augmented imaging and artificial intelligence data analytics laboratory. His team is one of several around the country that has pushed AI programs developed in a calmer time into the COVID-19 crisis to perform tasks like deciding which patients face the greatest risk of complications and which can be safely channeled into lower-intensity care.