The Man Who Ran America's $9B Science Agency: "AI Isn't a Private-Sector Miracle"
Sethuraman Panchanathan, known as Panch, ran the US National Science Foundation for almost five years and is now at Arizona State University. He discusses the decades of public funding behind modern AI, the government's role in research and safeguards, and how access to AI can be widened through partnerships between government and industry. Hosts Autria Godfrey, Stephen Horn and Laila Rizvi ask whether public institutions can keep pace with the private sector.
Transcript
AI is often viewed as a private sector revolution. Silicon Valley builds the models, Big Tech owns the compute, venture capital funds the race, and governments are left trying to catch up. But that's not the whole story. Modern AI was actually built on decades of public research powered by federal funding and public institution academic work. The breakthroughs now being commercialized by private companies did not appear out of nowhere. They came from a much wider system: universities, government agencies, federally funded institutions, and public investment.
Dr. Sethuraman Panchanathan, also known as Panch, should know. Panch used to run the National Science Foundation here in the United States, one of the most important and powerful public science agencies in the world. And for decades under his direction, the NSF was investing in AI long before names like Anthropic or OpenAI even existed.
So today, we're asking him the question: if public money helped build the foundations of AI, but now private companies control much of its future, who should be reaping the benefits? Is there a role for the government? Who should be in charge of shaping what comes next, or is the private sector already too far ahead in claiming ownership?
Straight ahead, we are talking AI, publicly funded science, and the future of innovation with former National Science Foundation Director, Dr. Sethuraman Panchanathan.
All right, Stephen, here on "Agents of Tech," we tend to discuss the AI race as between private companies, whether it's here in the United States or globally. Do you agree that there is a place where taxpayers should be able to say, "Hey, wait a minute, I helped fund some of this. Where's my refund?"
So Autria, if you've talked to any mathematician or physicist in the United States of America, they will tell you very clearly, we only have an iPhone today because of their research. You know, America for many years, its competitive advantage has come from the foundational research that it does. Well, things have changed, right, as you say. But you still get a large impact from universities now, particularly with young people who are coming into the labor market who are powering a lot of these innovative companies. But you're right, what is the role for government, Laila? Because you know, it's not particularly fast moving, is it?
I think it'll be interesting to hear from Panch because-
It's a good way of getting out to the... answering the question.
I mean, I don't know the answer to that question. That's why we're doing this interview. But I think he's so interesting because he's been there from the very beginning. He's been there through the AI winters as well as the AI summers, which not many people have been, when times were bad, and he's been on both sides within the government, within universities and academia, but he's also had a lot of exposure to these big private companies even before they became as powerful as they were today. So I think he'll be one of the few people who will actually be able to speak on that.
And I think the most difficult thing with a technology such as AI is just the pace of change. Whereas in previous instances, the pace hasn't been so fast that the government doesn't even know what's going on. I mean, it's... I say this all the time, but I can't believe it's only been three years, even less than that, since ChatGPT 3.5 came out, where it was just about writing these really awful poems, but that was all you could really do with it, and now it's doing half the work in most companies, including ours.
It'll be interesting to hear his thoughts on the role the government should play. But you're also a part of this conversation, so what do you think? Public science helped fund and build the foundations of modern AI, but private companies now control much of its speed, scale, and deployment. So who should claim ownership, and should governments really be getting more heavily involved? Tell us in the comments and be sure to like and subscribe. Here now is our conversation with Panch.
Panch, thank you so much for taking the time to speak with us today. We're very excited to have you with us. You oversaw the National Science Foundation as director for almost five years, but we know that the NSF was funding AI initiatives and research for decades, even as Laila put it, during the AI winters. So given that we know that, how can anyone make the argument that AI and the current AI structure that we know today and use today did not come from the public?
Basically, as we have already said, that the last five to six decades of sustained investments by NSF and other agencies, you know, NSF is the agency which has funded 80% of the non-defense compute research over the last several decades, right? So much of the AI of today, which happens to be my own area of research too, is being invested in, and as you rightly said, even during AI winters when people were skeptical about AI, I mean, we continued investing in them.
So what you see today is not just what happened in the last few years or even in the last year, because you hear a lot about this since ChatGPT, I guess, right? But this has been a tremendous amount of work over several decades by a lot of people, and that's why we are here, and I'm thrilled to see this because now it has the potential to achieve a number of things and that makes it exciting for all of us.
So given that, what role do you think the federal government should play? Should they have a role? Do you think that the government should be active in perhaps trying to pass legislation that will put safeguards in place? And if so, what's the first piece of legislation that you would like to see personally? What do you think is the most important right now?
I think first of all, federal government investments continue to shape the future of discovery research as well as some of the high-risk translational research activities. It is very important because as a nation, if you look at the trajectory of science technology innovation, you'll find that we have been in the vanguard and highly competitive because of the investment by the federal government.
Now, of course, you have industry also investing in the area of AI clearly, but that doesn't mean that we have to back off or we should back off, because today's discoveries and future discoveries are going to shape further what we do with AI and companion applications that we'll build and develop with AI, and those things are going to have to be invested in by the federal government. That's the first answer to the question that you asked.
The legislation, I mean, clearly, you know, you want the federal government to play a role in terms of ensuring that we have appropriate safeguards. I mean, look at any technology, look at nuclear technology, look at any technology for that matter. The federal government and the Congress has a role in terms of ensuring that appropriate safeguards and best practices are promulgated, and this is very, very important. And in AI, what we have seen is it becomes even more important for a variety of reasons. So I think this is a very important moment for legislation to be enacted to ensure safe, responsible, ethical, privacy-preserving, secure AI for all of us.
Dr. Panchanathan, you talked about high-risk research, right? And I think everyone right now wants to throw their hat in the ring when it comes to AI, as in it's just so much noise that you can't separate the signal from the noise. You've been there throughout the bad times and the good times. What do you think is the most underrated opportunity in terms of research and just AI innovation right now which people might be missing?
We launched at NSF the last four or five years when I was at NSF, 27 AI institutes, and these institutes permeate all parts of our nation and address problems specific to agriculture or learning or climate and weather, transportation, you name it. So AI in context is what I say. So there is fundamental core AI development, which we need to continue to invest in, both federal government and private sector, but also applications that contextualize AI, leverage AI in the best possible way.
So to answer your question crisply in what more can we do? I really believe, and I'm sure we'll talk about this more too, I really believe that the co-evolution of AI and the human is going to be an exciting future for all of us. It is not just, people talk a lot about AI substituting for human, you know, things of that nature, but actually I find the exciting part is the co-evolution of the AI and human, in general technology and human co-evolution, is something that's very exciting, and I think we should invest more in that realm in my view.
Can I just pick you up on that point? When you say that it's important that it's a brighter future for everyone, AI and humans together. At the moment, there's so much money being pumped into AI. Hyperscalers going to spend a trillion dollars. I've never heard of anybody spending a trillion dollars in a year. How do we make sure that it is for everybody and that what we don't actually see from AI is greater inequality?
That's an excellent question. See, whenever these kinds of technologies achieve their full potential, there is excitement and also there is the opportunity that people want to leverage economically and personally and so on. Industries want to make best use of the moment when people are all willing to take it as part of their consumer experience, et cetera. So clearly there'll be a lot of private sector-based investments that will happen in any technology when it reaches a level of maturity, okay? We have seen that in semiconductors, we are seeing that in AI right now, we have seen that in wireless technologies and so on. The scale that you talk about is absolutely true, is much larger because it is much more expensive and therefore it looks much larger in terms of real dollars.
So how do we do this? What we do is democratization of AI access, as you rightly point out, is exceedingly important. Why is it important? It's important because talent and ideas are democratized, and therefore you want to take advantage of all that we have in terms of talent and ideas across our nation. We also want to deliver progress to everyone in our nation. So both at the input and the output, we want democratization, and that is possible when you have democratized access, and therefore federal government has a role in this.
At NSF when I was there, we built what is called the National AI Research Resource called NAIRR, but we didn't try to do it as only federal government. We tried to partner with all the AI companies so that we could take an approach of a hyper-partnership, as I call it, approach, where we can leverage all of the assets and infrastructure that private sector may want to make happen here, because I think there's the interest in private sector to also want democratization of AI access, okay? Contrary to what people may think, they have interest in wanting democratization of AI access too, and therefore you just band together, bring all the collective investments, and now the Genesis Project, which is talking about in a science discovery with all the partnerships. All of these efforts, I think, public-private partnership-based efforts are very important to ensure democratized access.
And on that note, what... I mean, you've obviously spoken about democratized access, but what does democratized access actually look like? Is that people having access to the tools and the infrastructure, or is it more than that, and is it education and an understanding as to how best to use the tools? Because most people have access to things like ChatGPT 3.5, a frontier model for free, but they are not reaping the benefits of those tools, whereas other people really are. So how do you do that?
I mean, as with, I mean, you remember when, if you recall, not too long ago, we used to talk about the digital divide, right? Broadband access. This is not new again. We are just reliving all of this maybe at a different scale, okay? But the problem is the same. Broadband access was not available for people. We said, "Oh, there are going to be communities that are going to be left out. You know, more metropolitan areas will benefit from this. Rural areas may not benefit from this." We have had these kinds of challenges in the past. So this is also a similar challenge.
So here is to answer your question, we need democratized access in the form of education, training, awareness. The most important thing I always tell people is when public is skeptical, you know why they're skeptical? Because we have not done a good enough job of talking about the benefits in addition to talking about the challenges and acknowledging them, addressing them, and then all of this has to be much more transparently done. And so I believe that when I talk about democratization, it is about all of these things, not just availability of ChatGPT as important as it is, but training, learning how to use ChatGPT. Of course, access to models, applications, algorithms, and hardware, everything, all of that is part of what we mean by democratization of access.
My final thing on that note is the thing that scares me, I guess, is just the speed compared to previous examples that you gave like broadband, as in it's only been three years and just because it's moving so fast, if you haven't engaged in it until now, there's just such a big learning curve. And I've just never seen anything which has been so unprecedented in the scale and speed at which it's going. So it's very, very hard to catch up if you're not already locked in.
No, I would share that with you, Laila. I'm not disagreeing with you on that front. Things are moving really fast. I agree with you. And which means that, that's why I keep repeating this hyper-partnership term, and we need to really band together. It is not going to be just one sector, whether it is the federal government or industry or the state governments or the local regions and so on doing it. We have all got to do it, but we have all got to do it in a synchronized manner.
Let me give you an example if I may. When I came to NSF, I launched the Technology Innovation Partnerships Directorate, a new directorate, the first new directorate at NSF in 31 years. One of the exemplar programs we launched was called the Regional Innovation Engines, okay? The fundamental core belief in the Regional Innovation Engine program is that every area of our country has got innovativeness in context, place-based innovation. But such innovation cannot be just that federal government invests in it, or only the state government invests in it, or a foundational philanthropy invests in it, or the industry invests in it. It's got to be everybody coming together.
I recommend people to go look at the Regional Innovation Engines portfolio and you will find, I'm very excited that everybody comes together and does this, and that actually really has true impact, and that would be true for this also. We need to do this fast. We need to act on this in a very agile manner. I would agree with that. It's happening in certain places, but we need much more of a national approach to a comprehensive approach to this.
All right. We're going to keep on the topic of speed when we come back, because coming up after the break, we're asking Dr. Panchanathan whether the government can ever move fast enough to set meaningful safety standards for AI.
Continuing our conversation now with Dr. Panchanathan, former NSF director, I want to talk now about the government and whether you think they have enough in-house technical expertise to try to evaluate frontier AI systems properly and keep up with what Laila was talking about, just the speed at which things are progressing.
I often tell people that we sometimes underestimate the kind of capacities and expertise that is there in the federal government. Having worked in NSF, I will tell you, I will dispel the notion that somehow the expertise to do all of these things does not exist. I'll tell you why that exists. At NSF, for example, we have unbelievable program officers who essentially nurture all these ideas working with the academic sector. We also bring people from the academic sector, industry sector to work at NSF for short periods of time. This is true of other agencies. There's also true of administrations who bring in unbelievable expertise to the administration from the private sector. So therefore, the expertise is definitely there and therefore, I'm not too concerned about, "Oh, well, are they capable of doing that?"
But expertise alone is not enough. I go back to the earlier point that I made. In these kinds of things, you want to be partnering with the people who are in the frontier, you use the word frontier, in the frontier, and they are the people in the industry. And I can tell you, I've had many conversations with leaders in industry. They are only too happy to partner with the government in helping shape all of this future together in an unselfish way, not necessarily just only advocating for their technology or their kinds of applications that they're engaged in, but they're willing to do this because they all see it as part of the collective good that we need to be engaged in. So my experience having spent five years in the federal government and also being in the advisory board is that there is that collective expertise that is coming along.
So actually, I want to follow up on that because now you are at Arizona State University, and I want to know if you think that these public institutions and universities can remain independent and credible sources for research if you've got a lot of that capital flowing in from industry.
So it depends on what kind of problems you're working with industry on. So let me pose the following question back to you, if I may.
I would love that.
Let me ask you a question because this I ask to people all the time. In the area of AI, okay, would you want the talent of the future to be trained by only academic professors? We acknowledge the fact that the field is moving so fast, and we also acknowledge the fact that a lot of work is being done in industry. Some of the pioneering work is happening in industry. So the model of even education, that's a discussion for a different day, but I want to plant that in your minds here. The model of education has got to change to accommodate for this convergence of academic faculty working with practitioner faculty from industry, all of us working together to train the talent for the future generation. So that's the first point I want to make.
Now back to your fundamental question that you asked, will industry partnership essentially enable the kind of discovery research that needs to be undertaken by public universities? The answer is absolutely possible, yes. Because I will tell you, at NSF, we pre-wired with the industry certain pre-competitive research problems that the industry felt was important to solve, and they were willing to work with NSF. And that would be true of industry investments in projects in the university too. So I'm not saying it's a given; it's got to be shaped. It's got to be shaped by the public universities working with industry, and it is not without intentionality and intensity. It has got to be both when working with industry in that manner.
All right, I'm sold.
All right, Dr. Panchanathan, we want to ask you a couple of rapid-fire questions that we pose to all of our guests here on "Agents of Tech." Mine is: where do you think people will draw the line on bringing AI into their personal lives?
People generally look for, what do I get out of this technology? Where do my benefits accrue? Okay? Now, if the benefits are significant for the individual, they will adopt and adapt to the technology. Where they will draw the line, however, is, "I do not want my privacy to be compromised in this process. I do not want to have detrimental effects either for me or my family or my community." Those are lines, hard lines that people will draw. They will say that I will not accept. I mean, I will not compromise just because I have the benefits that is accrued, and therefore be lackadaisical about the fact that we need to have proper protections, appropriate protections, safeguards, particularly safeguarding people who are vulnerable.
And so you brought up this issue of vulnerability through education, Laila. So I'm coming back to that because I think it's very important to understand that people care not only about what they benefit, but they care about what kinds of safeguards is important for them, their families, but also the community and the people that are vulnerable. And so this is where the lines will be drawn, and I think they're already seeing those lines being drawn gradually.
And my question is, what is something that is universally accepted in your specific field of expertise that you actually disagree with?
You know, I'll tell you, you asked my field of expertise. My field of expertise has changed in so many ways over the years, which is a good thing. That's what you want, isn't it? So let me address this from the academic sector point of view, because now I am in the academic sector, right? So let me address it as, okay, this is my field, academic sector. I deeply care about the future of education, higher education, education at large. I would in fact even use the word learning, the future of learning, right? It's lifelong, we all know that.
So where I have a difference is what I alluded to earlier, this belief that somehow learning happens in the campuses and four walls by the faculty that are going to be training you, the sage on the stage, all of these kind of models that we have looked upon as the sacred models of the past. I don't agree with that anymore. I think the learning is now much, much more broad engagement of people across sectors to be able to come and provide the appropriate learning for our future generations.
So I know that this is, this will be something that higher education sectors will be working on. It's not that they are not aware, but it's a question of awareness being converted to action and accelerating that at the same time, all the pace that you need to have. Because I think, I'll tell you, if we don't do this, our future generations will not look upon learning in the traditional sense as something that is valuable to them.
Yeah, totally.
And that's not a good thing.
So my final question is, and this will not be a difficult question for you as we've come to realize over the last half an hour, is what are you optimistic about?
I am very, very optimistic about what technology and the human can do to unleash the latent potential of humans. We have not seen anything yet, my dear friends. We have not seen the full potential of human beings being unleashed. The creative elements of the humans are not fully released yet. You wait and see. I do not know whether you'll live to see this or not. I hope technology will keep me alive for much, much longer. So if that is the case, I wish to see this, because I'm telling you, decades from now, the generations of people, the human potential will be unleashed at a scale, and the creative geniuses will be fully tapped in a scale that you can't even imagine now.
I'm very excited for that. I'm very excited for my grandchildren, my great grandchildren who are not there yet. I have grandchildren. I am very excited for the future. I'm very optimistic, and I know we have to take care of certain things along the way to make sure that we make that future possible, and we will do that. As a human species, we always know how to, at the end of the day, preserve ourselves and our future generations. We may go wrong here and there, that's okay, but we self-correct. We will get back on track if we have derailed in a few areas, but I'm hugely optimistic about this.
Awesome. You know, he talked about his changing area of expertise and changing career path. Maybe AI ambassador is what's next for you, because I've never seen somebody speak with so much enthusiasm about this arena. Dr. Panchanathan, thank you so much for your time today. This has been really enjoyable. Thank you.
Thank you so much. Autria, Stephen, and Laila, thank you so much. It was truly a pleasure talking to you this morning.
So I think he did a really good job of laying out the mutual benefit of having the future of AI research and implementation be a joint venture between the public and private sector.
Yeah. I think he made it clear that, look, things are going to change. And the change is inevitable.
It's coming.
And we've got to, like, think of new paradigms across the board in terms of collaboration, how we lease out infrastructure when it comes to private sector infrastructure versus public sector infrastructure. And also, most importantly, the education structure, which as in people graduating are like, "Well, what's it all for?" And so you can't just stick by what has always worked because those standards came from the Industrial Revolution where you're working nine to five, five days a week, your education was typically STEM. The way that you learn, as in it's so dated and it actually doesn't suit our current life.
Not working very well these days.
I mean, I'm with him totally on, we ain't seen nothing yet, and the huge amount of potential I'm truly with him. Where I'm not with him is that we're all in this together. I don't see any sign of that, and I don't see any sign of the-
That's the optimist in him.
Right, and I don't see any sign of the public sector catching up here in terms of how you fund the research, how you put guardrails on, but your point of education, who's going to do that, right? Who's going to actually redesign the whole education process? So I don't, I see the potential, I see the speed, I see the money. I don't see, I see the inequality widening, and I don't see the inequality narrowing.
But what does it take to shift that direction? A couple of people at the helm?
It's that opportunity, isn't it? And understanding how we bring opportunity to everybody and not make it about money, you know? So it's about equality of opportunity, that's what, to me, I think we need to be working on.
Well, a huge thank you to Dr. Panchanathan for joining us today. What would you like to see? Private industry continue to captain the ship, or should government and federal research institutions regain a bigger role? Weigh in by dropping your comments below. Be sure to subscribe, and we'll see you next time right here on "Agents of Tech."