At our AI-focused Transform 2020 event, taking place July 15-17 entirely online, VentureBeat will recognize and award emergent, compelling, and influential work through our second annual VB AI Innovation Awards. Drawn from our daily editorial coverage and the expertise 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.
Here are the nominees in each of the five categories — NLP/NLU Innovation, Business Application Innovation, Computer Vision Innovation, AI for Good, and Startup Spotlight.
Natural Language Processing/Understanding Innovation
A senior principal scientist at Amazon Research and faculty member at the University of California, Santa Cruz, Dr. Hakkani-Tur currently works on solving natural dialogue for Amazon’s Alexa AI. She has researched and worked on natural language processing, conversational AI, and more for over two decades, including stints at Google and Microsoft. She holds dozens of patents and has written or co-authored more than 200 papers in the area of natural language and speech processing. Recent work includes improving task-oriented dialogue systems, increasing the usefulness of open-domain dialogue responses, and repurposing existing data sets for dialogue state tracking for natural language generation (NLG).
BenevolentAI’s mission is to use AI and machine learning to improve drug discovery and development. The amount of available data is overwhelming, and despite a steady stream of new research, too many pharmaceutical experiments fail today. BenevolentAI helps by accelerating the indexing and retrieval of medical papers and clinical trial reports about new treatments for diseases that don’t have cures. Fact-based decision-making is essential everywhere, but for the pharmaceutical industry, the facts just need to be harvested in a synthetic, relevant, and efficient way.
Research continues to uncover bias in AI models. StereoSet is a data set designed to measure discriminatory behaviors like racism and sexism in language models. Researchers Moin Nadeem, Anna Bethke, and Siva Reddy built StereoSet and have made it available to anyone who makes language models. The teams maintains a leaderboard to show how models like BERT and GPT-2 measure up.
Hugging Face seeks to advance and democratize natural language processing (NLP). The company wants to contribute to the development of technology in this domain by growing the open source community, conducting research, and creating NLP libraries like Transformers and Tokenizers. Hugging Face offers free online tools anyone can use to leverage models such as BERT, XLNet, and GPT-2. The company says more than 1,000 companies use its tools in production, including Apple and Microsoft’s Bing group.
Business Application Innovation
Jumbotail’s technology updates traditional “mom-and-pop” stores in India, often known as “kirana” stores, by connecting them with recognized brands and other high-quality product producers to help transform them into modern convenience stores. Jumbotail does so without raising the cost to customers by collecting and mining millions of data points in real time every day. 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 financial tech platform for payments and credit. The insights and tech developed around this new business model empower producers and customers, and Jumbotail is poised to expand to other continents.
Rasa is an open source conversational AI company whose tools enable startups to build their own (close to) state-of-the-art natural language processing systems. These tools — some of which have been downloaded over 3 million times — bring AI assistants to life by providing the technical scaffolding necessary for robust conversations. Rasa invests in research to create conversational AI, furnishing developers at companies like Adobe, Deutsche Telekom, Lemonade, Airbus, Toyota, T-Mobile, BMW, and Orange with solutions to understand messages, determine intent, and capture key contextual information.
Dr. Richard Socher is probably best known for founding MetaMind, which Salesforce acquired in 2016, and for his contribution to the landmark ImageNet database. But in his most recent role as chief scientist and EVP at Salesforce (he just left to start a new company), Socher is responsible for bringing forth AI applications, from initial research to deployment.
To help domain experts without AI expertise deploy AI products and services, Platform.ai offers computer vision without coding. It’s an end-to-end rapid development solution that uses proprietary and patent-pending AI and HCI algorithms to visualize data sets and speed up labeling and training by 50-100 times. The goal is to empower companies to build “good” AI. Platform.ai can count big-name brands like GE, Claro, and Mattel as customers. The company’s founders include chief scientist Jeremy Howard, who is also the founding researcher of deep learning education organization Fast.ai and a professor at the University of San Francisco.
In their powerful work, “Large image datasets: A pyrrhic win for computer vision?,” researchers Abeba Birhane, Ph.D. 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 data sets. These data sets, including ImageNet and MIT’s 80 Million Tiny Images, have been cited hundreds of times in research. Birhane and Prabhu’s work is under peer review, but it has already resulted in MIT voluntarily and formally withdrawing the Tiny Images data set on the grounds that it contains derogatory terms as categories, as well as offensive images, and that the nature of images in the data set makes remedying it unfeasible.
An assistant professor in the School of Interactive Computing at Georgia Tech and a research scientist at Facebook AI Research, Dr. Dhruv Batra focuses primarily on machine learning and computer vision. His long-term research goal is to create AI agents that can perceive their environments, carry natural-sounding dialogue, navigate and interact with their environment, and consider the long-term consequences of their actions. He’s also cofounder of Caliper, a platform designed to help companies better evaluate the data science skills of potential machine learning, AI, and data science hires. And he helped create Eval.ai, an open source platform “for evaluating and comparing machine learning (ML) and artificial intelligence (AI) algorithms at scale.”
Ripcord offers a portfolio of physical robots that can digitize paper records, even removing staples. Employing computer vision, lifting and positioning arms, and high-quality RGB cameras that capture details at 600 dots per inch, the company’s robots are able to scan at 10 times the speed of traditional processes and handle virtually any format. Courtesy of partnerships with logistics firms, Ripcord transports files from customers such as Coca-Cola, BP, and Chevron to its facilities, where it scans them and either stores them to meet compliance requirements or shreds and recycles them. The company’s Canopy platform uploads documents to the cloud nearly instantly and makes them available as searchable PDFs.
AI for Good
Authors Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dandres built an online calculator so anyone can understand the carbon emissions their research generates. Machine learning research demands high compute resources, and even as the field achieves key technological breakthroughs, the authors of the calculator believe transparency about the environmental impact of those achievements should be generalized and included in any paper, blog post, or publication about a given work. They also provide a simple template for standardized, easy reporting.
Niramai developed noninvasive, radiation-free early-stage breast cancer detection for women of all age groups using thermal imaging technologies and AI-based analytics software. The company works with various government and nonprofit entities to enable low-cost health check-ups in rural areas in India. Prevention and early detection are key to improving the outcomes of cancers, but health centers are not always equipped with expensive screening machines. Because thermal imaging is safe, cost-effective, and easy to deploy, it can improve early screening in low-tech facilities around the world.
Dr. Pascale Fung is director of the Centre for AI Research (CAiRE) at the Hong Kong University of Science and Technology (HKUST). Among other accolades and honors, Fung represents the university at Partnership on AI and is an IEEE fellow because of her contributions to human-machine interactions. Through her work with CAiRE, she has helped create an end-to-end empathetic chatbot and a natural language processing Q&A system that enables researchers and medical professionals to quickly access information from the COVID-19 Open Research Dataset (CORD-19).
Dr. Timnit Gebru continues to be one of the strongest voices battling racism, misogyny, and other biases in AI — 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 data sets to increase transparency and accountability.
Relimetrics develops full-stack computer vision and machine learning software for QA and process control in Industry 4.0 applications. Unlike many other competitors in the field of visual inspection, Relimetrics proposes an end-to-end flow that can be adopted by large groups, as well as smaller manufacturers. Industry 4.0 is associated with a plethora of technological stacks, but few are able to scale to large and small manufacturers across multiple industries yet remain simple enough for domain experts to deploy them, which is where Relimetrics comes in.
Dr. Daniela Braga, DefinedCrowd
DefinedCrowd creates high-quality training data for enterprises’ AI and machine learning projects, including 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, which 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 has led DefinedCrowd through several rounds of funding — most recently, a large $50.5 million round in May 2020.
Flatfile wants to replace manual data janitoring for enterprises with its AI-powered data onboarding technology. Flatfile is content agnostic, so a company in essentially any industry can take advantage of its Portal and Concierge platforms, which are able to run on-premises or in the cloud. Flatfile has completed two funding rounds, one of which wrapped up in June 2020. As of September 2019, the company had attracted 30 customers with essentially no paid advertising. Less than a year later, it had 400 companies on its waitlist, ranging from startups up to publicly traded companies.
DoNotPay, founded by British-born entrepreneur Josh Browder, offers over 100 bots to help consumers cancel memberships and subscriptions, fight corporations, file for benefits, sue robocallers, and more. While much of the company’s automation engine is rules-based, it leverages third-party machine learning services to parse terms of service (ToS) agreements for problematic clauses, such as forced arbitration. To address challenges stemming from the pandemic, DoNotPay recently launched a bot that helps U.S.-based users file for unemployment. In the future, the startup plans to bring to market a Chrome extension that will work proactively for users in the background.