To kick off three days of AI panels, discussions, fireside chats, and networking at Transform 2020, VentureBeat presented the second annual AI Innovation Awards. Drawn both from our daily editorial coverage and the expertise, knowledge, and experience of our nominating committee members, these awards give us a chance to shine a light on the people and companies making an impact in AI.
Amid four nominees in each in our five categories — Natural Language Processing/Understanding Innovation, Business Application Innovation, Computer Vision Innovation, AI for Good, and Startup Spotlight — the winners have emerged.
Research continues to uncover bias in AI models; StereoSet is a dataset designed to measure discriminatory behaviors like racism and sexism in language models while ensuring that the models otherwise offer strong performance. Researchers Moin Nadeem, Anna Bethke, and Siva Reddy built StereoSet and have made it available to anyone who makes language models. They maintain a leaderboard to show how models like BERT and GPT-2 measure up.
Business Application Innovation: Jumbotail
Jumbotail’s technology revolutionizes traditional “mom & pop” stores in India, often known as “kirana stores,” by connecting them with brands and other high-quality product producers to help them emerge as modern convenience stores. Jumbotail does so without raising the cost to customers, by collecting and mining in real time millions of data points everyday. Thanks to its AI backend, Jumbotail became India’s leading online wholesale food and grocery marketplace, with a full stack that includes integrated supply chain and logistics, as well as an in-house fintech platform for payments and credit. The insights generated and tech developed around this new business model empowers producers and customers, and is poised to extend to other continents.
In their powerful work, “Large image datasets: A pyrrhic win for computer vision?,” researchers Abeba Birhane, PhD candidate at University College Dublin, and Dr. Vinay Prabhu, principal machine learning scientist at UnifyID, examined the problematic opacity, data collection ethics, labeling and classification, and consequences of large image datasets. These datasets, including ImageNet and MIT’s 80 Million Tiny Images, have been cited hundreds of times in research. This paper is under peer review, but already it’s resulted in MIT voluntarily and formally withdrawing the Tiny Images dataset on the grounds that it contains derogatory terms as categories as well as offensive images, and that the nature of the images in the dataset makes it infeasible to remedy the problems.
AI for Good: Dr. Timnit Gebru
Dr. Timnit Gebru continues to be one of the strongest voices in the AI community fighting racism, misogyny, and other biases — not just in the actual technology, but within the wider community of AI researchers and practitioners. She’s the co-lead of Ethical AI at Google and cofounded Black in AI, a group dedicated to “sharing ideas, fostering collaborations, and discussing initiatives to increase the presence of Black individuals in the field of AI.” Her work includes Gender Shades, the landmark research exposing the racial bias in facial recognition systems, and Datasheets for Datasets, which aims to create a standardized process for adding documentation to datasets to increase transparency and accountability.
Startup Spotlight: Dr. Daniela Braga, DefinedCrowd Corp
DefinedCrowd Corp creates high-quality training data for enterprises’ AI and machine learning projects, including with voice recognition, natural language processing, and computer vision workflows. The company crowdsources data labeling and more from hundreds of thousands of paid contributors and passes the massive curation on to its enterprise customers. Customers include several Fortune 500 companies. The startup’s cofounder and CEO, Dr. Daniela Braga, has credentials in speech technology and crowdsourcing dating back nearly two decades, including nearly seven years at Microsoft that included work on Cortana. She’s led DefinedCrowd Corp through several rounds of funding — most recently, a large $50.5 million round in May 2020.