Recommendation 2
We urge NIH to take a central role in conceiving, developing and refining these tools, setting standards for their efficiency and efficacy, and making them broadly available and enabling in each of its domains. In doing so, NIH should be mindful of two important considerations: First, in most cases, creation of best-in-class tools will require highly sophisticated computer science expertise and capacity. Thus, NIH should contribute their biomedical research and health expertise, and relevant data sets, to collaborations with computer scientists at DOE National Labs, National Institute of Standards and Technology (NIST), or DOD; joint projects might be developed with NSF Computational and Data-Enabled Science and Engineering in Mathematical and Statistical Sciences Program. Such collaborations would deliver great strategic and intellectual value to NIH; moreover, the budget implications would be relatively modest, as most necessary resources will be underwritten, with many already in place, in the computation-rich collaborating agencies.
Second, NIH must acknowledge and address evident risks and potential negative consequences inherent in these technologies (as considered in the White House Blueprint for an AI Bill of Rights, and in the Bipartisan House Task Force Report on Artificial Intelligence). Primary among these problems for NIH is that many or most data compilations involving human subjects or materials are not representative of national or regional population demographics. It will be essential to develop policies, technologies, and practices that eliminate such problems in future data sets and algorithms,
Made with FlippingBook - Online magazine maker