The Vector Institute, an independent nonprofit dedicated to advancing AI, today established a team to commercialize its industrial and health care research. This group will create tools, frameworks, and model templates while helping organizations operationalize models within their markets. Moreover, Vector says it will be responsible for procuring computing infrastructure to scale the institute’s research.
Vector launched in March 2017 with $135 million in total funding over five years from the government of Canada, the government of Ontario, and nearly 40 Canada-based companies across a range of segments. The institute comprises faculty members and graduate researchers working in AI who specialize in areas like computer vision, reinforcement learning, and natural language processing. While Vector offers its researchers industry opportunities, it doesn’t provide fee-for-service consulting or compete with private sector companies. Rather, it’s an independent entity whose mission is to work with other institutions, industry, startups, incubators, and accelerators to further AI research and drive its application and deployment.
The formation of this new team, which Vector says figures into its 2020-2023 strategic plan, is a reflection of the cost of AI research. A Synced report estimated that the University of Washington’s Grover fake news detection model cost $25,000 to train in about two weeks. OpenAI reportedly racked up millions of dollars training its GPT-3 language model, and Google spent an estimated $6,912 training BERT, a bidirectional transformer model that redefined the state of the art for 11 natural language processing tasks.
Vector’s new engineering team will be charged with supporting the institute’s industry sponsor companies through knowledge transfer programs. (Any Canada-based company undertaking AI R&D efforts or already leveraging AI in products can apply to become a sponsor.) The team will also collaborate with the Schwartz Reisman Institute for Technology and Society to encourage responsible AI development, assisting partners in addressing challenges relating to AI governance, fairness, explainability, interpretability, privacy, and security.
Beyond this, the engineers will seek to accelerate Vector’s ability to conduct and reproduce experiments, pushing research forward by enhancing the institute’s infrastructure and applying innovations through open source publication and maintenance. And they will operate the thousands of GPUs, petaflops of computation, large-scale databases, training frameworks, and workflow tools that make up what Vector calls “one of the world’s most significant nonprofit machine learning computational infrastructures.” (Vector recently purchased 240 Nvidia Quadro RTX 6000 GPUs it has temporarily repurposed for Ontario’s Pandemic Threat Response project.)
The team will contribute to the Vector Institute’s completed NLP Project, in which partners worked with Vector researchers to create and reproduce natural language processing models “conducive to application.” The next three years will see at least 10 more similar projects.
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here