Anytime the topic of artificial intelligence comes up, two things dominate the discussion: It’s the technology of the future and China is ahead of the U.S. Whatever the reality, the White House earlier this year launched a National Intelligence Research Resource Task Force. It’s run out of the Office of Science and Technology Policy and the National Science Foundation. For a progress report on what the task force has been up to, Federal Drive with Tom Temin turned to the senior advisor for translation, innovation and partnerships at the National Science Foundation Dr. Erwin Gianchandani.
Insight by Confluent: Learn about how agencies are benefitting from that concept of data-in-motion to improve mission outcomes in this exclusive e-book.
Tom Temin: Dr. Gianchandani, good to have you on.
Erwin Gianchandani: Hey, thanks so much for having me, Tom.
Tom Temin: Tell us about this National Intelligence Research Resource Task Force – it’s a mouthful. And what’s the charter? And what have you been up to so far?
Erwin Gianchandani: Sure, absolutely. It is definitely a mouthful. So the AI Resource – I’m just going to say AI for short – the AI Research Resource Task Force is something that really was chartered by Congress as part of the National Defense Authorization Act of 2021. So the bill became law early this year. And as part of that bill, there’s a national AI initiative act. And as part of that act, there is a call for the federal government to stand up this task force to be able to really explore what a shared computing and data infrastructure that could provide AI researchers and students across all disciplines of science and engineering, with access to a holistic advanced computing ecosystem – what could that look like? And so the goal of the task force is really to do that sort of legwork of trying to be able to understand what can go into the design and the implementation of an AI research resource that could be available to the broader country that could really democratize access to this capability for all of AI research and development and education as well. And then to put that forward, that implementation plan, that governance plan, that sustainability plan, put that forward to the president and Congress to determine what the next steps might look like. So the task force launched back in June of this year, and we are sort of on a bit of a shot clock, if you will, to try to develop the implementation plan, get a first draft and interim report within a year of that date. So sometime next spring, early summer, and then have a final report in place by sometime next fall, so November or December of 2022, if you will.
Tom Temin: And of course, the immediate question comes up since this was part of an NDAA, but that’s not always fully defense-related items that end up in the NDAA. Does this idea encompass both military and intelligence uses, along with civilian uses? Or is this more civilian-oriented, that you envision?
Erwin Gianchandani: Well, so that’s a great question, Tom. And I think that that’s really something that will be up for the task force to help define scope. I will say that Congress was really I think, mostly focused on how do we democratize access to the cyber infrastructure? That is datasets, high quality representative data sets, that is an understanding of advanced computing capabilities. If you think about AI today, what fuels the AI revolution that we’re seeing today? it’s access to data and its access to advanced computing, whether it be high performance computing, or supercomputers or access to cloud computing or hybrid models of computing, where new and emerging forms of computing like neuromorphic computing and so forth, right? And so I think they were really interested in trying to see how could the broad constellation of academic researchers and students – so all different institution types: research primary institutions, community colleges, minority-serving institutions – how could they get access to these sort of core critical elements, if you will, to be able to fuel AI research and development? And in particular, how can we ensure that anyone anywhere regardless of what community you’re in, regardless of what institution you’re affiliated with, regardless of your own personal background, you could have access to this type of resource? So I think that your question about military versus civilian is a great one. I think we are initially anyways, in the task force discussions are very focused on the civilian access to this type of a resource, but we are certainly trying to keep an open mind in terms of our deliberations as we move forward.
Tom Temin: We are speaking with Dr. Erwin Gianchandani, he is the senior advisor for translation, innovation and partnerships at the National Science Foundation. And what about the data and access to data? Because a big challenge for many researchers is simply having what is the latest and greatest data set that is both useful to what it is they’re researching, and also has the lack of bias or the inclusion that is needed in so many AI applications.
Erwin Gianchandani: I’m so glad you asked about that, Tom. And I’m so glad you asked about sort of bias and personal privacy, actually is another factor here, thinking about civil liberties and civil rights too, when we think about the datasets. Or just more generally the research that’s enabled through this type of research resource. And so as we’re having these conversations on the task force, one of the things foremost in our mind is what are the plans that would allow us to be able to provide access to data through the resource? But at the same time, how do we ensure that any recommendations also take into account responsible privacy-enhancing techniques and capabilities and approaches? And how do we ensure that whatever we do, there are ground rules in place for user-based permissions, there are ground rules in place for data portability. And there are ground rules in place for de-identification of any personally identifiable or other privacy elements in the data sets themselves. And it may well be the case that sensitive data might not be provided through the resource given legal and policy restrictions. But anything that is provided has to be subject to these privacy enhancing techniques and these sorts of governance attributes that I just described.
Tom Temin: And do you envision people coming to this resource when it is established at some point in the future and using it for their own commercial purposes? If that’s the purpose of it, since a lot of it will have originated as a taxpayer resource? Or is it primarily aimed, do you think, at federal needs for artificial intelligence?
Erwin Gianchandani: Yeah, so Tom, that’s a great question. And I think it’s one that we want to see sort of where the task force goes over the course of its deliberations over the next few months. I should note that all of the meetings in the task force are open and public. This is a federal advisory packed committee. And so they’re noticed we put out notice notifications in the Federal Register, we really want folks to engage with us in the meetings, we’ve had two now that we’re about a half day to three-quarters of a day long. And we had a number of folks from beyond the task force who attended and asked questions, and so forth. And so I think getting that diversity of perspectives is really critical. The other thing that I’ll also note that’s very relevant as we just issued a request for information – actually not just – we issued it at the end of July, and we recently extended that RFI period through the beginning of October. And so we’re really looking again, for inputs from all different stakeholders, all different perspectives, regardless of sector, regardless of background, to help inform our thinking. And to really come back to your question about who should have access to this particular resource. I will say that Congress chartered NSF and OSTP and NSF is in the business of supporting, we’re a federal funding agency that largely supports research at academic institutions. But we will see where the inputs that we get and where the deliberations of the task force take us.
Tom Temin: And are you thinking about the sustainability of this effort? Because over the years, I’ve seen a lot of these types of things launch – a number of years ago, there was data.gov, for example, and performance.gov. And this, that, and the other.gov. And they start strong with a lot of fanfare. But then the effort, frankly, fades and data.gov is still there, and it’s been developed, but it’s not what I think anyone envisioned when they first launched it. So what about the long-term sustainability of this type of effort?
Erwin Gianchandani: Couldn’t agree more time about the importance of sustainability for this type of an effort. For this to really have, I think, the long term impact that we want it to have, and for it to allow the us to continue to be a leader, when it comes to innovations in AI research, as well as education, we really need to think hard and long and fast about what is that sustainability plan. And that sustainability plan, Tom, is not just about the set of folks who might come together, and the set of resources that might come together. Remember, NSF, the Department of Energy, other agencies in the federal government, we’re already funding some of these types of resources, advanced computing capabilities, access to data sets, and so forth. This is in some sense about trying to stitch together some of those existing investments and new ones into the future. And so as we do that, thinking about that sustainability piece, how do you ensure constant refreshing of the compute capacity and capabilities, for instance, right, that is all a factor that we’re going to be focusing on as we look into this fall and into the early part of next year with the task force discussions.
Tom Temin: And so eventually, you envision federal entities that have large scale computing capabilities, being able to maybe contribute time?
Erwin Gianchandani: Potentially very much. So that’s very much a possibility. Contribute their resources to this type of national AI research resource, be able to allow for allocations on those types of resources and so forth. Absolutely, that’s certainly within the realm of possibility.
Tom Temin: And one final question on the realm of possibilities: Could it also include a kind of commons, where people that wanted to do work that could be in the open source area, could contribute their learnings to a kind of commons or platform that makes those new discoveries available to anyone as an open source basis?
Erwin Gianchandani: Yeah. So I think, Tom, that’s an excellent question. And it’s actually one that briefly came up in one of our recent task force meetings, and indeed, it’s something that we will be looking at closely, too is this notion of as you’re conducting the research, how do you then in turn make the results of that research available for broader consumption and in particular, the data sets that might emerge. I mean, that is actually a key aspect of reproducibility and replicability of research these days. And so it’s certainly something that’s at the top of the minds of the task force members.
Tom Temin: And so ironically, this could greatly expand AI work but at the same time, reduce duplication of effort?
Erwin Gianchandani: Could be, absolutely.
Tom Temin: Dr. Erwin Gianchandani is a senior advisor for translation, innovation and partnerships at the U.S. National Science Foundation. Thanks so much for joining me.
Erwin Gianchandani: Thanks for having me, Tom.
Tom Temin: And we’ll check back with you in another year to see how it’s going.