At our AI-focused Transform 2020 occasion, going down July 15-17 solely on-line, VentureBeat will acknowledge and award emergent, compelling, and influential work by our second annual VB AI Innovation Awards. Drawn from our day by day editorial protection and the experience of our nominating committee members, these awards give us an opportunity to shine a light-weight on the individuals and firms making an affect in AI.
Here are the nominees in every of the 5 classes — 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 school member on the University of California, Santa Cruz, Dr. Hakkani-Tur at present works on fixing pure dialogue for Amazon’s Alexa AI. She has researched and labored on pure language processing, conversational AI, and extra for over twenty years, together with stints at Google and Microsoft. She holds dozens of patents and has written or co-authored greater than 200 papers within the space of pure language and speech processing. Recent work consists of enhancing task-oriented dialogue programs, growing the usefulness of open-domain dialogue responses, and repurposing current information units for dialogue state monitoring for pure language era (NLG).
BenevolentAI’s mission is to make use of AI and machine studying to enhance drug discovery and improvement. The quantity of obtainable information is overwhelming, and regardless of a gentle stream of latest analysis, too many pharmaceutical experiments fail as we speak. BenevolentAI helps by accelerating the indexing and retrieval of medical papers and scientific trial reviews about new remedies for illnesses that don’t have cures. Fact-based decision-making is important in all places, however for the pharmaceutical business, the details simply must be harvested in an artificial, related, and environment friendly manner.
Research continues to uncover bias in AI fashions. StereoSet is an information set designed to measure discriminatory behaviors like racism and sexism in language fashions. Researchers Moin Nadeem, Anna Bethke, and Siva Reddy constructed StereoSet and have made it out there to anybody who makes language fashions. The groups maintains a leaderboard to point out how fashions like BERT and GPT-2 measure up.
Hugging Face seeks to advance and democratize pure language processing (NLP). The firm needs to contribute to the event of know-how on this area by rising the open supply neighborhood, conducting analysis, and creating NLP libraries like Transformers and Tokenizers. Hugging Face affords free on-line instruments anybody can use to leverage fashions equivalent to BERT, XLNet, and GPT-2. The firm says greater than 1,000 corporations use its instruments in manufacturing, together with Apple and Microsoft’s Bing group.
Business Application Innovation
Jumbotail’s know-how updates conventional “mom-and-pop” shops in India, typically often called “kirana” shops, by connecting them with acknowledged manufacturers and different high-quality product producers to assist remodel them into fashionable comfort shops. Jumbotail does so with out elevating the price to prospects by amassing and mining tens of millions of knowledge factors in actual time daily. Thanks to its AI backend, Jumbotail grew to become India’s main on-line wholesale meals and grocery market, with a full stack that features built-in provide chain and logistics, in addition to an in-house monetary tech platform for funds and credit score. The insights and tech developed round this new enterprise mannequin empower producers and prospects, and Jumbotail is poised to increase to different continents.
Rasa is an open supply conversational AI firm whose instruments allow startups to construct their very own (near) state-of-the-art pure language processing programs. These instruments — a few of which have been downloaded over three million occasions — convey AI assistants to life by offering the technical scaffolding crucial for strong conversations. Rasa invests in analysis to create conversational AI, furnishing builders at corporations like Adobe, Deutsche Telekom, Lemonade, Airbus, Toyota, T-Mobile, BMW, and Orange with options to know messages, decide intent, and seize key contextual info.
Dr. Richard Socher might be greatest identified for founding MetaMind, which Salesforce acquired in 2016, and for his contribution to the landmark ImageNet database. But in his most up-to-date position as chief scientist and EVP at Salesforce (he simply left to begin a brand new firm), Socher is chargeable for bringing forth AI purposes, from preliminary analysis to deployment.
To assist area specialists with out AI experience deploy AI services, Platform.ai affords pc imaginative and prescient with out coding. It’s an end-to-end speedy improvement answer that makes use of proprietary and patent-pending AI and HCI algorithms to visualise information units and pace up labeling and coaching by 50-100 occasions. The aim is to empower corporations to construct “good” AI. Platform.ai can depend big-name manufacturers like GE, Claro, and Mattel as prospects. The firm’s founders embody chief scientist Jeremy Howard, who can be the founding researcher of deep studying schooling group Fast.ai and a professor on the University of San Francisco.
In their highly effective 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 studying scientist at UnifyID, examined the problematic opacity, information assortment ethics, labeling and classification, and penalties of enormous picture information units. These information units, together with ImageNet and MIT’s 80 Million Tiny Images, have been cited lots of of occasions in analysis. Birhane and Prabhu’s work is underneath peer overview, however it has already resulted in MIT voluntarily and formally withdrawing the Tiny Images data set on the grounds that it accommodates derogatory phrases as classes, in addition to offensive photos, and that the character of photos within the information set makes remedying it unfeasible.
An assistant professor within the School of Interactive Computing at Georgia Tech and a analysis scientist at Facebook AI Research, Dr. Dhruv Batra focuses totally on machine studying and pc imaginative and prescient. His long-term analysis aim is to create AI brokers that may understand their environments, carry natural-sounding dialogue, navigate and work together with their setting, and contemplate the long-term penalties of their actions. He’s additionally cofounder of Caliper, a platform designed to assist corporations higher consider the info science expertise of potential machine studying, AI, and information science hires. And he helped create Eval.ai, an open supply platform “for evaluating and comparing machine learning (ML) and artificial intelligence (AI) algorithms at scale.”
Ripcord affords a portfolio of bodily robots that may digitize paper data, even eradicating staples. Employing pc imaginative and prescient, lifting and positioning arms, and high-quality RGB cameras that seize particulars at 600 dots per inch, the corporate’s robots are in a position to scan at 10 occasions the pace of conventional processes and deal with nearly any format. Courtesy of partnerships with logistics corporations, Ripcord transports information from prospects equivalent to Coca-Cola, BP, and Chevron to its amenities, the place it scans them and both shops them to satisfy compliance necessities or shreds and recycles them. The firm’s Canopy platform uploads paperwork to the cloud practically immediately and makes them out there as searchable PDFs.
AI for Good
Authors Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dandres constructed an online calculator so anybody can perceive the carbon emissions their analysis generates. Machine studying analysis calls for excessive compute sources, and whilst the sector achieves key technological breakthroughs, the authors of the calculator consider transparency about the environmental impact of these achievements must be generalized and included in any paper, weblog publish, or publication a couple of given work. They additionally present a easy template for standardized, straightforward reporting.
Niramai developed noninvasive, radiation-free early-stage breast most cancers detection for girls of all age teams utilizing thermal imaging applied sciences and AI-based analytics software program. The firm works with varied authorities and nonprofit entities to allow low-cost well being check-ups in rural areas in India. Prevention and early detection are key to enhancing the outcomes of cancers, however well being facilities should not at all times outfitted with costly screening machines. Because thermal imaging is protected, cost-effective, and straightforward to deploy, it will possibly enhance early screening in low-tech amenities world wide.
Dr. Pascale Fung is director of the Centre for AI Research (CAiRE) on the Hong Kong University of Science and Technology (HKUST). Among different accolades and honors, Fung represents the college at Partnership on AI and is an IEEE fellow due to 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 permits researchers and medical professionals to shortly entry info from the COVID-19 Open Research Dataset (CORD-19).
Dr. Timnit Gebru continues to be one of many strongest voices battling racism, misogyny, and different biases in AI — not simply within the precise know-how, however inside the wider neighborhood of AI researchers and practitioners. She’s the co-lead of Ethical AI at Google and cofounded Black in AI, a bunch devoted to “sharing ideas, fostering collaborations, and discussing initiatives to increase the presence of Black individuals in the field of AI.” Her work consists of Gender Shades, the landmark analysis exposing the racial bias in facial recognition programs, and Datasheets for Datasets, which goals to create a standardized course of for including documentation to information units to extend transparency and accountability.
Relimetrics develops full-stack pc imaginative and prescient and machine studying software program for QA and course of management in Industry 4.zero purposes. Unlike many different rivals within the discipline of visible inspection, Relimetrics proposes an end-to-end stream that may be adopted by giant teams, in addition to smaller producers. Industry 4.zero is related to a plethora of technological stacks, however few are in a position to scale to giant and small producers throughout a number of industries but stay easy sufficient for area specialists to deploy them, which is the place Relimetrics is available in.
Dr. Daniela Braga, DefinedCrowd
DefinedCrowd creates high-quality coaching information for enterprises’ AI and machine studying initiatives, together with voice recognition, pure language processing, and pc imaginative and prescient workflows. The firm crowdsources information labeling and extra from lots of of 1000’s of paid contributors and passes the large curation on to its enterprise prospects, which embody a number of Fortune 500 corporations. The startup’s cofounder and CEO, Dr. Daniela Braga, has credentials in speech know-how and crowdsourcing courting again practically twenty years, together with practically seven years at Microsoft that included work on Cortana. She has led DefinedCrowd by a number of rounds of funding — most not too long ago, a big $50.5 million spherical in May 2020.
Flatfile needs to exchange guide information janitoring for enterprises with its AI-powered information onboarding know-how. Flatfile is content material agnostic, so an organization in primarily any business can reap the benefits of its Portal and Concierge platforms, that are in a position to run on-premises or within the cloud. Flatfile has accomplished two funding rounds, certainly one of which wrapped up in June 2020. As of September 2019, the corporate had attracted 30 prospects with primarily no paid promoting. Less than a 12 months later, it had 400 corporations on its waitlist, starting from startups as much as publicly traded corporations.
DoNotPay, based by British-born entrepreneur Josh Browder, affords over 100 bots to assist shoppers cancel memberships and subscriptions, combat firms, file for advantages, sue robocallers, and extra. While a lot of the corporate’s automation engine is rules-based, it leverages third-party machine studying companies to parse phrases of service (ToS) agreements for problematic clauses, equivalent to pressured arbitration. To handle challenges stemming from the pandemic, DoNotPay not too long ago launched a bot that helps U.S.-based customers file for unemployment. In the long run, the startup plans to convey to market a Chrome extension that may work proactively for customers within the background.