In this interview, FedHealthIT’s President, Susan Sharer, speaks with Gil Alterovitz, Director of Artificial Intelligence at Veterans Affairs about the new NAII and its vision, partnerships and next steps around AI.

The National Artificial Intelligence Institute (NAII)

NAII presents a unique opportunity to really help Veterans, to be the go-to place for Veterans around leveraging their own data and providing insight using Artificial Intelligence (AI). Historically, there are many cases of external companies working on AI and the challenges of getting the needed data to those companies.

We wanted to build expertise within the VA and the capacity to work with external organizations on developing and implementing AI models to help Veterans. It is important to keep data private and secure. By leveraging expertise within VA, we can identify smaller subsets of data that will allow us to collaborate and work with those external sources.

The NAII was set up as a joint initiative by the Office of Research and Development and the Office of the Secretary’s Center for Strategic Partnerships in VA with the intent to be able to advance those opportunities.

The Vision

NAII has a number of specific initiatives, flagship pilot projects around research and development and working with partners. These flagship projects, with Veteran input, will work in particular AI areas, such as deep learning, trustworthy AI, privacy-preserving AI, explainable AI, and multiscale AI analysis. We also want to be an affiliation hub for existing AI or related efforts, a resource to help those efforts and to highlight them in a way that they can grow, scale, and be successful.


The NAII leads AI Tech Sprints using the “Government Innovation Award”-winning framework for empowering an AI-able ecosystem through voluntary incentives linking the ecosystem of Federal, industry, academia, and non-profit organizations around AI R&D.

With these tech sprints, we can work around smaller data sets that can be made available to small groups to build a tool around a use case for their own use or perhaps something they want to take further.

We are also taking Veteran input into account in terms of priority use cases for AI and we are eager to engage with industry around those areas, to focus tech sprints around data for AI, so the tools they build can and would be useful for Veterans.

In the past, we had to create cooperative research agreements, and this was something that would take a long time to move through to enable that access to data. Few companies would try to enter into these because of the time and commitment required.

This approach, the tech sprints, creates a wider entry funnel, allowing many more organizations to get involved. The smaller data sets also make the processes shorter, so you can have a number of companies working on the data at the same time. It is much easier to create agreements around smaller data sets. This can be used for whatever the organization is driving, but using VA’s format and around AI questions related to Veteran-specific issues.

This process makes it easier for organizations to show proof-of-concepts and facilitates moving to a larger data sets.  Furthermore, it helps to continue testing and benefit refining problems via lessons learned.

Next Steps

We are looking at several areas where AI is at a tipping point. There are several technologies within AI that can help with specific use cases and we are thinking about that mapping between these now, with goal of improving health and well-being of Veterans.

There is a list of priority use cases we are investigating, some that have been identified within VA and the administration. There was recently an executive order and task force, PREVENTS initiative around suicide risk, but that is just one example.

We are thinking about things like deep learning and explainable AI where the technology can detail why it is making certain recommendations to help us understand the process, enable safe AI, and see what we can take learn it. There are also potential opportunities where we do not need to see the underlying data AI is using, just the conclusions it is making. These are slower to do right now, but longer term can open new avenues.

Our priority areas now are anywhere we either have to integrate a lot of information quickly so it is hard for one person to do, and where we have to integrate information from many different sources so where it would be hard for human to remember and process all of those sources. Where a computer can help analyze that large volume of information and provide reasons why it is thinking a certain way or summarize it in a way that is easier for a person to process, there is great opportunity.

About Gil Alterovitz

Gil Alterovitz is the Director of the National Artificial Intelligence Institute (NAII) at the U.S. Department of Veterans Affairs. Dr. Alterovitz has led national and international collaborative initiatives for developing novel informatics methods and approaches for integrating clinical, pharmaceutical, and genomic information, from research to point-of-care. He is also one of the core writers of the White House Office of Science Technology and Policy’s National Artificial Intelligence Research and Development Strategic Plan, which was recently updated. His work on integrative methods for “big data” in the biomedical informatics space has been published or presented in more than 50 peer-reviewed publications ranging from academic journals and international conferences to three books (including “Systems Bioinformatics: An Engineering Case-based Approach,” ranked #1 in new Amazon bioinformatics category).



Please enter your comment!
Please enter your name here