AI in Government: Current State and Future Potential with Nathan Manzotti of GSA: Episode Rerun

Welcome to AI Government and the Future, a podcast by Corner Alliance. We explore the intersection of artificial intelligence, government and the future. We work with government to create results. We ignite your agency's mission by helping you to design and implement high impact and innovative federal programs in AI broadband, cybersecurity, public safety and more. Being a government ally is at the core of all we do today. We are welcoming Nate Manzati. He is at GSA General Services Administration for those who aren't up on their federal acronyms. He's the director of Data analytics and AI Centers of Excellence at the Technology Transformation Services TTS there at GSA and also co manager of their community of practice around AI. So Nate, welcome to the podcast. Great to have you on. Yeah, thanks for having me. So let's start. Help me untangle the government acronyms and offices. So how do these Centers of Excellence cops all fit with like the data analytics and GSA and tts? So the Centers of Excellence is a program, it's a fee for service consultancy that is part of Technology Transformation Services which is part of Federal Acquisition Service which is part of gsa. So that's kind of the tree of where the Centers of Excellence sit within gsa. And the Centers of Excellence exist to do consulting engagements with external agencies mainly. Also we sometimes do engagements with other parts of GSA as well. But we're cost recoverable fee for service entity so customers have to pay us to do the work. And there's I believe five centers, Data analytics, artificial intelligence, customer experience, contact center and cloud. Can you give us a couple examples kind of stuff you do for agencies? Yeah, sure. So I've been with the Centers of Excellence for about three and a half years at this point and I can talk about a few that I've been involved with. So in the Data Analytics Center I worked on a project for about two and a half years with the Surface Transportation Board. They're a small agency and that was a data modernization project. So we started with the data maturity assessment. We looked across the whole agency at their data procedures and data inventory and all that kind of stuff and then gave them set of findings and recommendations to work off to start their data modernization journey. A five year roadmap and that was in the first year of that. In the second year we started prototyping some of the top recommendations that we had for them, maturing their data architecture and their systems and adding some things for automation and so on and standing up Data Advisory Board to be underneath the Data Governance Board is more technical group and so we really work across, like, the entire spectrum of. I call it the data ecosystem. So that's an example of the type of project that we do on the data side, on the AI side project that I didn't work on. But it was a colleague of mine, Joe Rotano, who just left us, unfortunately, he built a model for the USDA to grade beef. So it was a image model. We built it using Keras and worked with the meat graders and people at USDA to capture their process and get them to verify the data and things like that and test the model. And it's just a prototype, but it performed fairly well. So that's another thing we do. We do everything from pure consulting strategy work to implementation. Wow, that's cool. So I need that app. So when I go to my butcher and they're trying to tell me this is like, whatever grade, I'm like, no, it's this grade. Pass it over to USDA and then we'll see what they do from there. Obviously, it's a lot different to build a prototype and to test the model than it is to put it in production. Full scale. Right. Support it. Right? Yeah, I had that same thought. I said, you know, this would be cool. It just has a consumer app for free. Like, hey, verified, right? Usda. Maybe if they put that out there, then you can go to the supermarket and test it. Well, it says this is grade A choice prime, whatever. But now the app says, right, Submit user data Catch company. Did you ever watch Silicon Valley that show? One of the guys programmed a app that was supposed to be the hot new AI app. It was a comedy. And it turned out his app was basically, it told you whether a picture was a hot dog or not. And that was the whole app. It's like, is this a hot dog? That's awesome. As dumb as it sounds, narrow, focused use cases for machine learning can do a lot. Yeah, it's true. I think we're going to see a lot more of that too, now that. Particularly now that you have models that can run on mobile. Right. So you're going to see a lot more focused, smaller models that do a really good job at one thing. All right, so that's the coe. Another thing about the coe. So one of the benefits of the COE is that, like, when we approach a customer that has a specific thing in mind, we're not limited to just that specialty. So let's say we're doing a data analytics engagement and they want to know how to modernize their data warehouse and have more real time analytics or something like that. We give a technology recommendation. But then in that second stage they say, well, now we need to teach everyone how to use this and how to transition their workloads over. We can bring in like the Innovation Adoption Group, which is not a formal coe, but it's a focus area I think is the term for something. We have really capable innovation adoption capabilities. And we bring those guys in, we bring the cloud guys in, we're flexible, so we just flex to meet the customer where they're at. And because we're federal, we don't have to necessarily hold on to every dollar, you know what I mean? So if it makes sense to say, you know what, let's cut this off, you can do your own contract from here with your own procurement authority. Then we get out of the way when it no longer makes sense for us to be there. That's the COEs in a nutshell. So the cop, the artificial intelligence cop, is a cross government community that exists to facilitate communications, share information and provide training for the government on AI. And it's grown quite a bit. So we're at about 3,300 feds right now and numbers going up all the time. Last year was very successful. We did some AI training, brought in Stanford University, Stanford Human Centered AI and Stanford High to do the training, did a six part series and we worked with OMB and that training series was their response to the AI Training Act. And it was really well received, super high quality content. So. So that's an example of the type of thing that we do in the cop. And for this upcoming year we're planning to do a lot more. So it's more like a knowledge sharing coordination group. Not necessarily. You're not doing services for people through the cop. Exactly. So the difference between the COE and the cop, the COE is pay to play. We can come in with expertise, we do assisted acquisitions, we solve your problem. The community of practice is free. Anyone with a.gov or mil email can join. And it exists for knowledge sharing. And we're trying to make it as valuable as possible for the people that participate. So just taking a step. So that's sort of the framework. And TTS also does a lot of the procurement GWACs and stuff like that across government. Right above your level. Well, so TTS Technology Transformation Services is in the Federal Acquisition Service. Federal Acquisition Service. They are the ones that have all the GWACs and all the procurement authority. So there's an obvious benefit for us being part of fast. We have inside knowledge about all of Those things. And what I mean by inside knowledge is that we can help our customers get access to the right vehicles and buy things in a way that makes sense for them. So these are the people to go to if you're a federal program officer trying to think through your AI strategy and procurement. I mean, these guys have all the tools. That's right. And that goes back to what I was saying about we're a consulting organization. We come in and we work with you and help you on yourself. But then when it makes sense, we say, okay, this is going into operations and maintenance, or just you don't really need the assistance of feds anymore. You can hire your own contract team to do this. We can put you in contact with people in FAS that can, like the national account managers and other procurement authorities that can help our agency partners buy direct. So now you don't have GSA in the middle. So taking a step back just from GSA and those programs, I know you don't work with everyone, but, like, what's your kind of take on the AI landscape across the federal government today? Like, what would you say the state of AI is based on what you've seen? Like, are we sort of in the experimental pilot stage? Are we in the. Hey, we're rolling out major AI applications on a daily basis. Like, where do you see the government at this point? I think it varies wildly from agency to agency. But as a whole, the world, probably in general federal government, being part of that, is going through continuous cycles of excitement and fear. And I mean, it's kind of just crazy. I don't think there's ever been a technology in history that has grown this fast. That's certainly true. I mean, ChatGPT alone, right, is the fastest adoption of any consumer app in history. And that's what the driver of it, right? So AI is not new. It's been around for a while, in the shadows, just like every other technology that's out there. People hear about it and they go, okay, that's cool, but they don't really care. What I mean by they don't really care is that if you're not a technician, it's not the most important thing on your plate to go, well, I need to learn about this technology. You interact with it. Anyway, in reality, people have been using AI applications every single day for the last decade because everybody uses Google, everybody has AI integrated on their phone in some way. There's a lot of systems out there that have had AI in them for over a decade now. So people didn't realize it. You use your phone, you use the things that exist and most people aren't thinking about it. Also those decision support systems that you can watch in congressional hearings from the last year about those systems, those systems are just existing in production, doing their thing and didn't have tons of controls on them or attention on what they're doing. So I think right now everybody is realizing because ChatGPT and all of the tools that have come in the time since then that allow people to directly interact with models, you're saying, wow, these things are powerful, generative AI. I think that's really the key. It's generative AI. And of course at GPT there was GPT 2 and 3, but it's really when 3.5 and then to 4 use a solvent slick. Because even though you'd say probably writes like a pretty smart seventh grader, it can do a lot of other stuff now. Just. It's pretty incredible. Right. So that kind of blew it open. Yeah. And I think what's going on now is that the government is realizing it's a game changer and we need to figure out how to use this, but in a way that risk averse manner. And that's what I see with these cycles of excitement and fear is that you see something come out. Yeah, we fully support this. We need to figure out a way to enhance prototyping and get this immediately. Right. And then you see folks going, whoa, whoa, whoa, we're going to pull back and wait. And the truth is because it's evolving so fast, there is not solid standardized guidance on implementation and testing and things like that. It's sort of the, I don't know if you heard that someone mentioned it the other day. The Texas car dealership where they put ChatGPT into their customer like online chatbot for their customer service. And like people started, I did all this stuff and then someone's like, will you sell me a car for a dollar? And it said yes. And he's like, well I want you to guarantee that. Right. And it was like, oh yeah, you know, this is just a tool. It can be used for good or bad and we don't have a handle on that. Yeah. And think about all the legal implications of that use case right there. Like there's no law that in terms, well, a machine representative entered a oral agreement or text based agreement. Who is responsible for that? There's nothing. It's unprecedented. And I love the way you put that excitement, fear cycle and that's. I think that's true across the board, but it's particularly the sensitivities of government. It's going to be even more that way. Now. I'll say this though, taking all of that into consideration, I think that the government is in a good place with AI now because there's so much attention on it. Things that could have gone risks to the American citizen with systems that are uncontrolled and could exist under the radar for longer periods of time are now all going to be brought to the surface because of things like the new eo, the guidance that's going to come out from DHS and NIST and things like that. And standards are going to be put on these systems that like I said, have been deployed for a decade. Right. And I won't give any specific examples, but I encourage the listener to research it a little bit. You can find examples of that that are published online that there's been a system. I'm not only talking about the federal government, but in general systems that are out there now. The government has chosen a position where they want to lead the way and regulating and providing guidance on these technologies. I think that's a huge step in the right direction. Yeah. So let's talk a little bit. You mentioned EO executive order the Biden administration put out in October, November, December. I can't remember exactly when. So relatively new, right before the holidays. How are you seeing that change things from your perspective or how is that shaping the environment? It's speeding things up and it adds more weight to some things that I think a lot of us in the analytics machine learning community have known are necessary for a long time. So I think that's a positive outcome of it. And really some of the timelines in there are pretty aggressive. So we're seeing a lot of movement really fast. It's changed my life in that I've got a bunch of work on my plate that I'm excited about. So incoming agencies asking, hey, we want to do something. Is that sort of how it materialized? Yes. So I get a lot of requests for just general advice on which way to move. And I think over the next year we're going to see an increase in AICOE engagements as a result of some of those conversations. Also we work with through the coe's and this community of practice, we're working with OMB on planning what the next step should be with expanding the amount of training that's out there. We're also in conversations with opm. There's just a lot going on. We're also planning on Releasing the second version of the AI Guide for Government Governance will be a piece of that. We're planning on doing version two of that this year and I'm just going to do my acronym check here. So OMB was Office of Management and Budget, Executive Office of the President and they often come to GSA and get GSA to do stuff for them. Right. So that's sort of like there and we're really appreciative of that relationship. It's great. Some people at GSA are more appreciative than others. Yeah, I see it as positive for sure. We work with them and we're able to include their priorities in our planning for the COP and some things that we focus on and that just in my opinion helps to like expand, make it bigger. So I guess what I would see is sort of like this generative AI. My take would be it's speeding up but it's not like a day to day thing baked into federal operations at this point. Would you agree with that? And that sort of like that's what's coming next as you see it. I mean it's almost going to get forced in a way where that's getting embedded into productivity software with the office suite and you guys use Gmail and or the Google suite and it'll be in there as well. So it feels like next year is going to be the year it kind of comes into everyday operation so the rubber's really going to hit the road. That's an interesting question because I agree. So this year everybody's trying to figure out what to do to create some constraints around how to use it to make sure that people don't put sensitive information in there. There is going to be sensitive information. It's in a walled off controlled environment with all the security controls. That is all being figured out now. And I agree with you, next year is when like the full production versions, integration into everything is probably going to really kick off heavy. And said everybody's using it's powerful and so you can get so much done especially in the right hands. Somebody like me, if I'm working on something like developing some guidance for benchmarks for AI testing and quality control, of course I'm going to use generative AI to flesh out my answers and then I research those. Right. But knowing how to use it, I say okay, I can't just take this blanket response from the AI, I have to verify it and all that. But it's just an accelerant progress. So whether in my opinion what I think is going on is if agencies are blocking it, probably people are still using it. So that's one of the reasons why I'm excited that we have an opportunity to contribute to training the whole federal government in AI. Because I think that training on multiple different levels is very necessary right now. Including like training on kind of like cybersecurity awareness training. Do this with AI, don't do this. Baseline understanding of the risks and how to use it. And then in addition to that also more role specific training. Like someone who's an acquisition professional. Well, how can they use it? What should they be aware of? But even just that use case or going back to OMB rulemaking comments. You know, they go through these comment periods and it strikes me that I'm hearing from other business owners, maybe outside of the government, their quoting doing a proposal for work went from six weeks to six hours with generative AI. So you can imagine your colleagues in fast getting deluged with or you know, they're the people they're doing acquisition for getting deluged with contract proposals or on the comments side getting deluged with comments on a rulemaking. And so it really strikes me that federal government has no choice but deploy AI to like deal with that AI. And so the, the forcing function is coming. No one can really bury their head in the sand and avoid it. Yeah, I mean you're going to see a period of cat and mouse, right? Yeah, of course. Let you guess which one is going to stay ahead. So I think the private sector is going to be ahead and it's going to force government to play catch up on it because I think government's just not going to be as agile in deploying solutions. And one thing that I think about a lot, and I talk about this too, and it's an emerging area of research and product development. You know, in the private sector I think there's a lot of power in using generative AI to structure unstructured data. What I mean by that is you take something like ChatGPT equivalent through an API and give it a context that says hey, look at this humongous trove of PDFs and then extract all the objects into this schema and if you know what a JSON object is, put it in a JSON. I would say extract all the people from this and take the page number, their name, the contextual information, et cetera, et cetera, et cetera. And then you do a bunch of iterations on that and now all of a sudden you have a database that you can analyze and Doing that and then also testing to make sure that that model actually did so. Running it through a couple different iterations that are slightly different to get the same information allows you to evaluate how good that model is at extracting that information correctly from the PDFs. Not only do you get the structured database that now you can do analytics on more traditionally and as more easy to verify and things like that, you're not just going to AI for everything, but you can also test and make sure that it's accurate. You're kind of hitting on what I think is probably the most profound thing the government's going to do beyond regulation. I mean, the treasure trove of data within government is immense. There are billions of treasures in the attic here. And so you look at open data and all that stuff, now that you can do it with these tools like you're talking about, and those are getting better. This is a national asset that could grow our economy at huge. It's really going to be pretty transformative. It kind of seems like what generative AI is making possible is kind of what the promise of big data was back in like 2012, 2013. Now it's going to be realized because like back then it's just like there was advances in like paralyzed compute. You had the hadoop and stuff that came along, but still to analyze that data, it was technically complex and sometimes cost prohibitive. Right. And a ton of it was just unstructured. Right. So it was very difficult. Exactly. So now I think we're going to see that kind of reality come to bear, which is an interesting concept in itself because that's something from the past, and said, wow, this is going to be huge. And then it took a different technology to really realize the benefit of that. What is being promised now that in reality is not going to be possible with AI that everybody's talking about. And then it's going to take, who knows, maybe quantum computer, whatever the next thing is. And I think that's often the case. Like I think about that in like biotech. I think it's a really good point. Like everyone talked about genetic medicine starting and whenever the genome was mapped back in 2000, whatever. And now it's like CRISPR and even other technologies and AI that are actually going to make it a reality 20 years later. So that's often the case. But I think you're right. I think at this point the kind of big data, open data, government thing is going to be a real. I think that's just a whole area we haven't Even like thought through the scale of that yet. And I think agencies don't know what they have. I have a friend who's kind of working on AI with some pharma companies and even pharma doesn't know what they have. They've done tons of research and clinical trials that failed and all sorts of stuff that's stuck all over their organizations. They don't even know what they have, much less like the massive federal government. There's just so much stuff out there. Yeah, I mean some of these agencies have data just stored on magnetic tape that used to be in paper archives going back to 1910 or something. What kind of crazy and interesting analytics is going to come once all that data is analyzed? So you kind of hit on one of the big focuses of the executive order, which I really credited, I thought the executive order called out very well was the challenge of educating and hiring, educating, retaining federal workforce who has expertise on machine learning and AI. So tell me a little bit how you're seeing that. Are you seeing some efforts by the government to really try to, I mean you're, you're obviously doing some of these trainings, but I can just imagine anyone who gets really qualified in these technologies is going to get picked off by the private sector. There is definitely a pretty significant pay gap between the private sector and the federal government. Well, they're talking now like entry machine learning engineers are getting 500 to $800,000 a year in salary. Really? Where is that link? Yeah, no, it's like the kind of people that are going to get hired by OpenAI, by Google, by Microsoft, by Meta, those people. And I'm sure they have to have certain very specific parts of that. So I'm just thinking through like what's an entry level machine learning engineer at that level level they're probably a PhD, computer science. That's probably true. Le Masters. Yeah, Le Masters. Well, they're saying that the number of PhDs and Masters in those fields are thousands, not like tens of thousands. So I mean there's just so much demand. Exactly. And the way that that market is kind of evolving. So those guys, yeah, they're going to go off, make a million dollars a year, they're going to build the next generation of technologies like the GPT 6 and 7, 8. But the rest of us really need to know how to access, integrate those technologies into more like specific applications and then add controls around them and test them, which is engineering. So the AI researchers, they're always going to be in super high demand. But I think now the pool and the need for like engineers that understand enough of the computer science and statistics and AI specific machine learning, specific methods and stuff, but also understand security and engineering principles and software development principles and data principles and all that kind of stuff, that pool has gotten bigger and even that. So we'll go a tranche down. Not the people making 300, $500 million a year, but in that space of people who understand all of this stuff, there's still a huge pay gap with the private sector. Right. Like everyone knows that. The interesting thing is we have a couple of programs. Like one program that TTS has is called the US Digital Core. And the US Digital Core is its purpose is to bring in more junior talent to the federal government, get them interested, junior technology talent. Founded by a friend of the pod, Lena Trudeau. Yeah, Chris, exactly. Yeah. So I've been involved with the hiring on that, like evaluating resumes for that and I've been really surprised with the quality of applicants that come through for such a low paying job. And so that indicates to me that there's interest in working for the federal government in the technology space. Even though some of these people are way more qualified than I would expect to see come through this pipeline. And I think that's because of working for the federal government in this space is a little bit unique. You have a chance to impact the lives of U.S. citizens. And you know, not to go down this rabbit hole with it, but you know, we do impactful work and people come to us for that. And so the Digital Core is one thing. As part of the executive order, there's an AI talent search. TTS is involved with that in a couple different ways. And I just heard that there's going to be some AI specific hiring actions. So I'm standing by and listening too. I'm not sure what's going to happen if they're going to authorize some kind of special pay or something like that, that probably would be something that would make it less of a difficult choice to stay or join the federal government. So I'm going to say a couple of things that I'm sure you're probably not going to be able to comment on because they're policy decisions. But what I see is a combo of first, I'd really like to see a company like Anduril come into the commercial, the civil space, so more of a company that is providing the building blocks for government outside and building the solutions rather than to what they see as the problems, rather than the government saying, hey, this is what we need, we're going to give you 5,000 requirements. So I think that Anduril approach is one first step and that's a huge culture change for the government. So no need to comment on that. Interesting. I'm just generally curious about that approach, what that looks like. So Parker, he founded Anduril and sort of the IDOD creates these huge platform technologies like Joint Strike Fighter or some battleship or whatever. And they define, they do these exhaustive lists of requirements. And a lot of us have been through parts, versions of that process. Anduril said that's not the way anything gets done in Silicon Valley. There's a problem at a major Fortune 500 company that probably exists in multiple of them. Someone develops like minimum viable product, goes out and tries to sell it to them. So it's like, I'm not taking your requirements, I'm solving your problem. Right. So that approach, I think that that has to spread to the civil sector. But again, huge culture change. It's interesting that you mentioned that, because something that I've been thinking about, but. So this is just like a potential thing that we're going to do in the future in the communities of practice is we have other higher priority items that we have to tackle for first. But I think there's an opportunity to create a process where we take the use cases. And now I'm talking specific AI, but really this could be for any common problem that exists across the board of agencies. Find one use case that kind of generalizes that requirement as best. If you get the 80% solution and then do like a contest where create a product and then at the end of that contest, there's a procurement authority that's made available to any agencies that want to do a prototype or something like that. Sort of like a prize challenge mixed with a phase three sbir almost. That's kind of cool. I like that kind of similar concept to like what bucket hiring is, but for solving problems that exist across government. Yeah, we have some clients. I don't want to talk specifically about our client, but they've kind of done something similar like creating innovation mechanisms for a big program that's doing the actual service. So they kind of create this set of prize challenges, grants, cooperative agreements, and it's kind of like an accelerator to then do stuff with this bigger agency as like their provider. And it's a really cool model where government's partnering with like, hey, I've got innovative acquisition authority. I might be able to hire different kinds of people because I've got some hiring authority. And I can help enable you to do a whole lot more stuff. It's a cool model. Kind of like what you guys are doing, right? Yeah. I mean whether that will become a reality sometime in the future or if it's even possible, I don't know. But it's just something that we're thinking about that's really cool. And I think the other thing that's going to have to happen is sort of an opening up of this US digital core with a lot different hiring authorities. I mean NIST and DOE and all these people have way different hiring authorities than others. They're able to hire nuclear engineers, not have to go through all the processes of the government. We're going to have to do that. And I think the government's just got to get used to the fact that people are going to be coming and going and so it should be much easier to bring in mid career professionals for two, three years, maybe adapt some of the DARPA model and people coming in and out of government. It's just going to have to be the case because yes, they're going to have to pay people more, but they're never going to be able to pay people as much as the private sector and there has to be that exchange. I just don't see how the government's successful with that. Exactly. If there was some kind of mechanism to incentivize people to just join the government for a period of time and to incentivize companies to let those people go for a period of time. Yeah. I mean to be honest, the digital core people are probably coming in, they get a couple of years is good on their resume, they know the government, they can then hop out, start a company that serves the government, work for Microsoft or OpenAI. So it's kind of like, I mean people call that a revolving door, but I think that's the way you get stuff done. Even myself. Right. Look, I've gone from being a Fed to being a contractor, back to being a Fed. And in the current environment I've considered going back the other direction. Right. But I've chose to stay because I'm excited about some things that we're doing but in the future have to reevaluate of course as well. And what I'd like to do eventually is get the contractors to not just be warmed over versions of the government with higher pay scales like have them actually be innovative places that look at government specific problems and say we don't need to create another CRM for government that's been solved. But There are a lot of government specific problems that other private entities don't have. So, like, have a much more innovative community outside. And there's a lot of barriers there. You got all these cyber regulation, you got FedRamp, you got Blah, blah, blah. But I think over time, if we can streamline that, I think there's an opportunity for finding more because there's widely different missions across agencies, federal government. That said, there's baseline commonalities too. And so grouping those baseline commonalities together and just getting more efficient with sharing things, that I think is a goal that would be a good one to pursue. And I mean, that said something that I've learned from working in federal government and working with the private sector is that, you know, the federal government is not. The private sector shouldn't be. Right? Yeah, exactly. It shouldn't be. It's intrinsically government functions is what the government is supposed to do. So there is a pretty big difference and it makes sense. These rules that are in place that are oftentimes slowing down work that we're doing in the centers of excellence and things like that, and it's a frustration in the short term. They exist for a reason. They're there because of sometime in the last hundred years or so, there's been probably some kind of catastrophe that led to those rules being created. So I think it's important to remember that too. For technicians working in the government is like, yeah, it is frustrating, but there's a reason. National security, even things affecting people's benefits and stuff like that, can have massive negative impact if we're not careful and deploy systems. I totally take that point. And it's true. I mean, I always said, like, if you wanted the federal government to act like a private sector entity, like the military might not protect Vermont because they'd be like, it's not. We're not getting an ROI on that. So it's like government has a different kind of mission and obviously it has a lot of stakeholders and anytime a mistake gets made, these rules get created. So I think about this, the interaction between government, industry, academia, nonprofits, I kind of like conceptualize that in my mind as like one big machine that's kind of like the machine of our collective society. It's important, I think, especially with AI, because it's moving so fast for that machine to work together and they all have different functions. That's an interesting perspective. I like that. Yeah, I wrote up a little article about it, but I've kept it to myself. I'll have to get that Maybe we could put it in the show notes. So that would be great. Where did you put it out? I didn't. I just wrote it and I've been expanding on it. But if you're interested in that, like, I could share the little graphic with you and some explanation about it. Basically, the idea is, what's the core function of each one of those groups? The government, they regulate things. Government has the authority, industry has the money, academia has the brains. Nonprofits have the ability to facilitate things. So I call that support. I love that. So I've thought a lot about that. So it's almost like government has a certain amount of convening authority. It has the authority and then, yeah, the money, the brains. And then nonprofit can represent interests. They can convene specific interests the way the government can't necessarily. They can also facilitate interactions between industry and government in a way that allows those conversations to happen without direct conflicts and things like that. Like, I think nonprofits serve an important role in facilitation and things like that. And a good thing to end on that. I can go into from that in this next year. In the community of practice, our goal is to kind of be the one place that people go to access all those different groups across what I call the machine for AI. So we are partnering with academia, other federal communities in the nonprofit space, like with ATARC and actiac entities like the Wilson center that does training industry groups as well. So trying to be a connector of everybody. So I encourage people to join the community of circus. Just Google GSAAI Community of Practice and you can find the link to join. Excellent. Well, thanks for being on the pod today. It was great having you, Nate, and we'll talk soon. Hey, thanks a lot for having me. I enjoyed this. AI, Government and the Future is brought to you by Corner Alliance. To find out more about Corner alliance and how we work with government to create results, visit our website@corneralliance.com and then make sure to search for AI government future in Apple Podcasts, Spotify and Google Podcasts or anywhere else podcasts are found and click subscribe so you don't miss any future episodes. On behalf of the team here at Corner alliance, thanks for listening.