Wednesday, October 10, 2018

PROPOSED Iowa Initiative for Artificial Intelligence – Brief Overview

The incipient Initiative for Artificial Intelligence at the University of Iowa is motivated by the fact that Artificial Intelligence (AI), including machine learning, synthetic vision, decision making, smart systems, etc., is reshaping every aspect of society and industry.  Progress in the science and technologies associated with big data, computational power, sensors, and interconnectivity have enabled unprecedented advances which will influence all aspects of our research and educational endeavors.  It is critical for the colleges, in close collaboration with ITS, to build and foster the AI capabilities for the University of Iowa campus.

The proposed Iowa Initiative for Artificial Intelligence (IIAI) will focus on (goals, problems, benefits to UI):

  1. Increasing availability of campus-wide AI-specific computational environment with effective administrative structure – building on the successful model of High Performance Computing on campus and leveraging existing funding.
    1. Current AI/GPU/machine learning computational resources require continuous growth and efficient management, UI is behind on their availability for supporting our research and educational missions.

   Resulting from the IIAI initiative, centralized AI-enabled computer cluster capacity will be continuously enhanced and maintained to serve research needs on campus, benefitting Colleges, Centers, and interdisciplinary researchers.

  1. Development of AI/GPU methods for the computer cluster requires availability of a smaller-scale interdisciplinary development lab(s) to design, debug, and test applications efficiently. Such an environment does not currently exist.

   Dedicated AI-development labs will be established. First two will be built in FY19 (Seamans Center, Pappajohn Biomedical Discovery Building), with more to be built as needed to support growing the community.

  1. Building AI-enabling project-related expertise brings together an interdisciplinary community of experts, educators, students, and research users.
    1. AI and deep learning (DL) will augment and perhaps partially gradually replace conventional analysis approaches. To stay competitive, interdisciplinary collaborative access to AI/DL knowhow is critical.

   IIAI will provide access to AI/DL interdisciplinary knowledge via hiring and properly training interdisciplinary AI/DL researchers (PhD level) who will act as knowledge consultants available to these research groups, who will contribute to and take part in new research grant proposals thus increasing grant submission success rates.

   Development labs will naturally contribute to cross-pollination of AI/DL knowledge as they bring expert and novice users together in a collaborative environment.

   IIAI will raise the overall level of AI expertise on campus and increase the number of competitive AI-related research proposals.

  1. AI/DL training and education of (potential) users at UI is currently not sufficient/efficient and needs to be flexible to quickly respond to the changing landscape of AI/DL.

   IIAI will join forces with UI3 to jointly offer relevant and accessible AI/DL education and training at all levels of expertise to maximize the impact on the UI research community.

  1. Facilitating research access to relevant data necessary for extramurally funded AI-research projects.
    1. AI/DL projects depend to a large extent on availability of training data.  Many barriers exist to providing access to large volumes of data across disciplines, thus making rapid research progress difficult at UI.

   IIAI will work closely with ICTS, CPH, UI3, TSB, as well as other UI and non-UI entities to develop policies and agreements to facilitate data access.  These policies will help resolve data ownership issues as they emerge.

   IIAI will maintain repositories of commonly used relevant research datasets that are too large for single labs to maintain, but can be shared across many labs on campus.  IIAI will collaborate with the campus HPC community to provide data transfer mechanisms both within the university and with external collaborators.  IIAI will provide access to datasets that would not otherwise be easily available.