Agentic AI for Drone & Robotic Swarming

In this episode of Practical AI, Chris and Daniel explore the fascinating world of agentic AI for drone and robotic swarms, which is Chris's passion and professional focus. They unpack how autonomous vehicles (UxV), drones (UaV), and other autonomous multi-agent systems can collaborate without centralized control while exhibiting complex emergent behavior with agency and self-governance to accomplish a mission or shared goals. Chris and Dan delve into the role of AI real-time inference and edge computing to enable complex agentic multi-model autonomy, especially in challenging environments like disaster zones and remote industrial operations. Featuring: Chris Benson – Website https://chrisbenson.com/, LinkedIn https://www.linkedin.com/in/chrisbenson, Bluesky https://bsky.app/profile/chrisbenson.bsky.social, GitHub https://github.com/chrisbenson, X https://x.com/chrisbensonDaniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaLinks: ROS - Robotic Operating System https://www.ros.orgGazebo https://gazebosim.orgHugging Face Agents Course https://huggingface.co/agents-courseSwarm Robotics | Wikipedia https://en.wikipedia.org/wiki/Swarm_robotics Chris's definition of Swarming: Swarming occurs when numerous independent fully-autonomous multi-agentic platforms exhibit highly-coordinated locomotive and emergent behaviors with agency and self-governance in any domain (air, ground, sea, undersea, space), functioning as a single independent logical distributed decentralized decisioning entity for purposes of C3 (command, control, communications) with human operators on-the-loop, to implement actions that achieve strategic, tactical, or operational effects in the furtherance of a mission. © 2025 Chris Benson Sponsors: Outshift by Cisco https://agntcy.org/: AGNTCY is an open source collective building the Internet of Agents. It's a collaboration layer where AI agents can communicate, discover each other, and work across frameworks. For developers, this means standardized agent discovery tools, seamless protocols for inter-agent communication, and modular components to compose and scale multi-agent workflows.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:03Speaker 00

  Welcome to the Practical AI Podcast, where we break down the real-world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X, or Blue Sky to stay up to date with episode drops, behind-the-scenes content, and AI insights. You can learn more at practicalai.fm. Now, onto the show.

• 0:49Daniel Whitenack?

  Welcome to another Fully Connected episode of the Practical AI Podcast. In these episodes, where we don't have a guest, it's just Chris and I, we take some time to deep dive into some interesting AI topics or explore some things in the news related to AI and hopefully give you a little bit of learning resources or information to help you level up your machine learning or AI game. I'm Daniel Whitenack. I am CEO at Prediction Guard, and I'm joined, as always, by my co-host, Chris Benson, who is a Principal AI Research Engineer at Lockheed Martin. How you doing, Chris?

• 1:27Chris Benson?

  Hey, doing great today, Daniel. How are you today?

• 1:29Daniel Whitenack?

  I'm doing well. Yeah, just took a, tried to take a run before it gets really, really hot here. So hopefully, you know, that worked out well. It was still kind of hot. Hopefully, I don't end up with a sunburn, but...

• 1:45Chris Benson?

  Absolutely. Yeah, it's been hot here in the US, especially in the southern US, where we both are right now. And so, yeah, you definitely don't need you, don't need you collapsing from heat exhaustion right before we get into a show here.

• 2:01Daniel Whitenack?

  Yeah, yeah, for sure. I will definitely say there's fewer people out. There was not in any way a swarm of people running anywhere that I saw. So yeah, I guess that leads a little bit into what you teed me up to talk about today, Chris, which I'm excited about, which is an area of your expertise and interest, which is swarming, which I guess could be related. You know, people might be thinking of animal swarming, robot swarming, other swarming. Maybe just starting there, what do we mean by swarming?

• 2:38Chris Benson?

  So before I get into that from my side, let me set that up a little bit because you actually mentioned several things that knowing me and also just the general topic as we've kind of prepped to do this a little bit, you know, you mentioned animal swarming. And so it's funny, in the AI world, we have been talking about robotics and autonomy and unmanned aerial and ground vehicles for various things. You hear about Amazon drones and Walmart drones are out there now and various other applications. And so even though my background may be in the in kind of military applications because I'm in the defense industry, we're really talking

• 3:21Chris Benson?

  about it in a general say. And one thing I should say upfront, just to cover myself is I'm only representing only my own personal viewpoint and not that of my employer or any other organization.

• 3:35Daniel Whitenack?

  Point taken.

• 3:36Chris Benson?

  Point taken, which is kind of funny because when you talk about swarming, if you put 10 people into a room and ask them each to define swarming, you will end up with about 17 different definitions of what swarming is. It's one of those types of terms.

• 3:51Daniel Whitenack?

  Yeah, you mentioned the delivery drones around where I'm at. I live right by Purdue University and actually for quite a few years now, we've had these little

• 4:06Daniel Whitenack?

  Little box robots that will deliver you food. So they like drive along the sidewalks and, you know, go through intersections and deliver food to you through, through an app and, and all that stuff. So that one definitely hits a little bit close to home. I think for the most part, I have seen them, you know, successfully navigate the terrain and, and intersections and all of that. Although I will say, I did see one at one point, I didn't see the, the collision, but it was definitely run over on the, on the road. It was in pieces at an, at an intersection.

• 4:47Chris Benson?

  Yeah, somebody didn't get their dinner that night.

• 4:49Daniel Whitenack?

  Yeah, exactly.

• 4:50Chris Benson?

  Somebody called up hungry, where's my food? You know, which, you know, and it was the delivery driver once upon a time. Hopefully that person would not have gotten hit, but yes.

• 5:00Daniel Whitenack?

  Right.

• 5:00Chris Benson?

  Having the, the, some sort of automated vehicle bringing the stuff. So, um, wanted to talk a little bit about, and probably before we get into, you know, talking about what swarming is, kind of distinguishing it because it's one of those autonomy words and, you know, autonomy is another one of those words. And we have things like drone and robot and UAV, UGV, UXV, X covering kind of all the things. And so.

• 5:26Daniel Whitenack?

  I don't know what a lot of those things are.

• 5:28Chris Benson?

  Okay.

• 5:29Daniel Whitenack?

  So.

• 5:29Chris Benson?

  Ground, sky. You got it.

• 5:33Daniel Whitenack?

  UFO.

• 5:33Chris Benson?

  Air. Yeah. Yeah. So like, like UXV would be unmanned or uncrewed is a little bit more modern term for it. The X signifies kind of what the domain it's operating in, air, ground, whatever. Uh, and then vehicle. And if you've seen UXS, that would be an unmanned or uncrewed, whatever system. And so those are really common lingo things and they're not specific to military. You see those in commercial and industrial applications and stuff. And for the most part, we'll probably for simplicity sake, mostly talk about drones to, uh, in this conversation, so we tend to stay away from jargon and maybe robots on, you know, for ground stuff.

• 6:16Chris Benson?

  And that way people are not trying to get through acronyms. While we're talking and stuff, just to simplify things a bit. But we're having, we're definitely living in this age, as you just pointed out from personal experience, where we're, we're starting to see these things, which are, you know, physical embodiments of some sort of AI or other algorithmic driven, you know, movement around the physical world for various activities and stuff. And it sounds like, and I don't know the specific technology for the one that's bringing you the food on how they're, how they're approaching that, because there's a bunch of different approaches, but that is a, either a semi-autonomous presumably, or a fully autonomous, you know, ground vehicle that's bringing you the stuff.

• 6:58Chris Benson?

  Whereas going back to our topic, swarming implies numbers, first of all. And so, but it doesn't just imply numbers, it implies the way numbers are working together and collaborating. And I think that's where a lot of people get in trouble with all sorts of different definitions. And it's super popular to talk about swarming now.

• 7:19Daniel Whitenack?

  Yeah.

• 7:19Chris Benson?

  It's a crazy buzzword these days.

• 7:21Daniel Whitenack?

  Yeah, I almost wonder, like, there could be some confusion. We hear a lot of talk these days about multi-agent systems. And of course, we've had episodes where we discuss agents specifically and what that term agents mean, but as soon as you, as soon as you bring in that kind of multi-agent side, people might be confused, is that what we're talking about? So I guess one question would be like, swarming, does it imply like physical, I guess, physical AI or physical autonomy in terms of things that are operating in the physical world, not the digital world?

• 8:04Daniel Whitenack?

  We've talked a lot about drones and robots in the beginning of this. So is that, is that part of that definition or, or not really?

• 8:12Chris Benson?

  It is.

• 8:13Daniel Whitenack?

  Okay.

• 8:13Chris Benson?

  It is, you know, and we've had across a lot of our episodes recently, especially our fully connected episodes like this, where it's you and I discussing a topic. You know, we've talked about, we're kind of, you know, we've had over the history of, of the show, we've saw, seen the evolution of AI. And there tend to be specific topics that get really hot and and right now agents are really hot. But there's also the notion of agents applying themselves again, you know, in collaboration with various other AI technologies, whether they be LLMs or reinforcement learning or, you know, computer vision. And I think if you take a step farther into this world of autonomy, especially modern autonomy in 2025,

• 8:53Chris Benson?

  agents are a big part of that in having multiple agents collaborating and they're using these other models to get these tasks done. So you have agents operating with LLMs for different purposes and, you know, so you're able to go and grab the right model for the right task and you're putting a bunch of tasks together to go do something which which I might call a mission, but I don't necessarily mean that in a military-esque way. Could be, but it could be something that you're that is what the purpose of your company's autonomy is to do.

• 9:27Daniel Whitenack?

  Maybe like a goal or an objective or outcome.

• 9:30Chris Benson?

  That's right.

• 9:31Daniel Whitenack?

  Could be synonyms.

• 9:32Chris Benson?

  That's right.

• 9:33Daniel Whitenack?

  I'm not that great at the English language, but.

• 9:36Chris Benson?

  Those are all those are all loose synonyms that could they could apply.

• 9:39Daniel Whitenack?

  Yeah.

• 9:40Chris Benson?

  And so, and we're seeing this, you know, you're we're seeing these individual drones and robots that are starting to do tasks in the commercial and industrial space. Certainly it's been that way in the industrial space and warehouses and robot. I guess the first robot I worked on was 2000 that personally I was doing with two of them that were autonomous in 2018. So this is not new in that capacity. How it's doing it has changed over time.

• 10:07Daniel Whitenack?

  Yeah, maybe it just taking a moment, kind of looking back a little bit because I remember what what year was it that the I forget the year of the Beijing Olympics when they had the it was the first one I had seen where they had a kind of. I guess we could call it a swarm. I don't know. You can correct me if I'm wrong swarm of drones that were sort of in the sky in the, I guess, I think it was the opening ceremony. I'm trying to remember, but that was the first time I had seen that. Of course, I've seen it in other places, you know, after that.

• 10:48Daniel Whitenack?

  But, um, but yeah, what's, what's kind of the, how would you, yeah. Well, how would you, how would you represent the history of kind of coming to the point where we're at now in these, so I'm getting from you, there's these autonomous, so multiple objects in the physical world that are autonomous or multiple vehicles or robots or whatever in the, in the physical world that are autonomous, trying to accomplish a goal or mission. Now, in the Beijing case, I don't know if those were autonomous or not, but yeah, help me kind of parse through maybe a little bit of that.

• 11:27Chris Benson?

  So I should say ahead of time that I don't have any special knowledge of that particular configuration. So I'm making educated guesses on that. And so I would, I would suggest that's not what I, what, and I'll define a swarm in a moment as a follow-up, but I would not call that a swarm. I would call that a large number of individual aerial platforms that are operating in a, in a predetermined coordinated manner.

• 11:52Daniel Whitenack?

  Yeah, it's like synchronized.

• 11:53Chris Benson?

  They're synchronized from the ground control system. So the, a notion in, in flying autonomous vehicles is that you have a ground control station, which might be like a laptop. It might be a bigger thing inside a truck or something, but it has, you know, comms and it has ways of communicating. And, and so for shows like that, you would use, uh, you know, you would use things like GPS and they would have a three, each drone would have a three dimensional path that it's following. That's a little bit different from all the others. But when you put them all up there at the same time, they look like a big coordinated show, but they're not actually coordinated. Uh, it's the human on the ground programming in their path. It is.

• 12:35Daniel Whitenack?

  They are not determining their coordination.

• 12:39Chris Benson?

  There you go. is really important. So yeah, you went right to the heart of it. And so there is that which is really not intelligent at all. It's pre-programming these drones to do something. And the visual impact is that you have this thing happening among a bunch of things, but it's it's sort of an illusion. It's really a bunch of individual things on their path. And then there's a whole, there's a whole continuum of how those kind of vehicles can operate. One kind of popular approach today that I, that, that a lot of people would call swarming and which I don't, and I'll define it right after I say this, is the notion of like giving plays to a group of vehicles

• 13:21Chris Benson?

  that go out and do a task, sort of like some people, the analogy will be a football play. It's go, go to this location and do this, and there might be some communication with the other platforms in your area that you're doing something with, but I also don't consider that be a swarm. So because I think it comes back to the notion of what is swarming. It's, it's swarming is not an autonomous thing. It's a type of behavior exhibited. And we see that in nature. And so I will, I'll give the definition and then I'm actually going to go away from technology for one second and talk about what we see in Mother Nature that's kind of consistent with that.

• 14:03Chris Benson?

  So my definition sounds a little military like, but don't take it, don't take it too much that way. It can be, it could easily be applied to commercial or industrial. So, but I think it gets the gist of how I, of how I see swarming. Mine is swarming occurs when numerous independent, fully autonomous platforms exhibit highly coordinated locomotive behaviors in any domain, be it air, ground, sea, undersea or space. Functioning as a single, independent, logical, distributed, decentralized, decisioning entity for purposes of command, control, and communications,

• 14:43Chris Benson?

  with human operators on the loop to implement actions that can achieve strategic, tactical, or operational effects in the furtherance of a mission. Little military sounding, but if you translate those words into commercial and industrial, they still apply. You might use a, you might choose a synonym in some of those areas. But that is a, is, is a pretty high bar right there. And, uh, and as we go forward, we can talk a little bit about kind of what in that definition turns something from a collection of autonomous vehicles into a swarm of autonomous vehicles.

• 15:34Speaker 03

  Well, friends, if you want to build the future of multi-agent software, check out AGNTCY. That's an open source collective building the internet of agents. It is a global collaboration layer where AI agents can discover, connect, and work across frameworks. That means better tools for you, standardized agent discovery, seamless inter-agent communication, and modular components to scale your multi-agent workflows. And they're teaming up with Crew AI, Langchain, Cisco, and many more, dropping real code, specs, and services with no strings attached. Start building alongside engineers who care about high-quality multi-agent software. Learn more at agntcy.org. Again, that's A-G-N-T-C-Y dot org. That's agntcy.org.

• 16:28Daniel Whitenack?

  Well, Chris, I really appreciate your definition. Maybe that's even something we can pull out and put into the show notes for people so that they can can read that if they if they want to. I'm actually reminded as we're talking about this and as I'm hearing your definition of these things, you know, in the animal kingdom. Actually, I'm remembering a movie that I watched with my wife, The Murmuring. I think it was on Netflix and very interesting movie, but kind of at the core of the movie were these, I don't know if I'm using the right terminology, but murmurations of swarms of birds.

• 17:13Chris Benson?

  Starlings are famous for it.

• 17:15Daniel Whitenack?

  Exactly. Yeah. So you can see in the sky, I'm sure many people have seen this, swarms of birds that almost form kind of patterns and synchronizations in the sky and move around together. Now, according to your definition, each of those birds, right? Now, I guess in this sense, you know, there's no connection to a, obviously, a human command and control type of scenario, but certainly they are autonomous, but they're accomplishing something together, which is maybe a little bit mysterious in this case,

• 17:57Daniel Whitenack?

  although I'm sure it's a topic of study. But, but yeah, that's what came to my mind is, is that, is that movie and that kind of phenomenon.

• 18:06Chris Benson?

  Yeah. And that murmuration, if you will, is literally, it's called flocking in that case, you know, which is obviously a bird term, but that is definitely consistent with what I would consider to be a swarm. And if you, if you break that down, and there's, and there are other animals we can talk about, too, that have different types of swarming behavior, but that are consistent with that. If you talk about starlings in particular, you know, they're beautiful. When you see those massive, and there could be thousands of starlings moving across and you see The waves within the system, you know, within the flock as they're doing that. And, and so you kind of go, you know, what's going on there? Well, you have thousands of individual beings

• 18:47Chris Benson?

  that are communicating and sensing each other. They have a unified purpose. They're all going someplace together, but there's no, there's no one master general or CEO starling that's going, you guys are gonna go that way. Everyone do what I tell you to do. They are all subscribing to the mission, if you will, but each one has a, has a, a position, if you will, or a place in that mission where they are communicating. And the way they do their sensing of each other and their communications helps them, A, not collide and, and, you know, fail by, by through collisions and such. It keeps them together, but they also have a way of, of agreeing,

• 19:32Chris Benson?

  if you will, to go do something. They're going to, to move from one place on the earth to another place on the earth. And, and that kind of comes down to the crux of it is that swarm behavior is where you have that, a bunch of fully autonomous beings, in this case, they're functioning in that murmuration as a single, independent, logical, distributed, decentralized, decisioning entity. It's a long mouthful that I use because all of those have a part in that describing it, but they're all working together in that way.

• 20:06Daniel Whitenack?

  Do you want to pick apart maybe each of those words? So, uh, just like, why, why use those multiple words? Kind of distributed, decentralized. Some people might kind of not know why each of those is, is important.

• 20:20Chris Benson?

  Sure. So, so single implies that the joint actions together about how they're communicating and sensing give a rise to one generalized process.

• 20:34Daniel Whitenack?

  Kind of an emergent or combined process.

• 20:37Chris Benson?

  That's a great way of putting it. And that's kind of that, that emergent, you know, we said mission or goal kind of thing is coming out of that. And that process is Independent of any one bird and it's and it's independent of any controlling authority that's saying from afar y'all she'll go do this thing, you know? And so you have that independence, you have that single emergent characteristic, logical meaning it's also kind of it's kind of think about it's arising a an emergent, intelligent notion or a processing notion that's all happening by the unity. It is distributed across the entire flock or murmuration.

• 21:20Chris Benson?

  It is decentralized. There's no one special bird or group of special birds that are the birds that are in charge of everybody else and you guys have to do what we say. And finally, it's a decisioning entity that by by this emerging thing happening, choices are made collectively for the whole thing and yet no one bird is saying everyone's going to go do this thing. So it's all these kind of characteristics about that emergent. I may add the word emergent into the definition that I have there because I think that's really a great way of describing that. But it's that unified collection to do this. Now, that's a far cry from what we're currently doing in the autonomy space at large.

• 22:04Chris Benson?

  I'm not saying that doesn't exist in little pockets, academic pockets and other places, labs, whether they be government labs or university labs and stuff like that. But you don't see widespread deployment of drones or robots that are acting in this manner as we sit here recording today.

• 22:24Daniel Whitenack?

  Okay, that that makes sense. Yeah. Yeah. Part of my, what I'm thinking and my questioning in my mind, which maybe you're going to get to as you kind of get get through some of these things is obviously, we have a kind of inspiration or a model there from the the animal kingdom. There in the case of, let's say, robots in a warehouse or a manufacturing facility or autonomous vehicles, we have in probably things that people are familiar with, the idea of self-driving things or autonomy. So like self, like the Waymo cars or that sort of thing.

• 23:05Daniel Whitenack?

  And there's certainly obviously AI at play there in terms of computer vision and maybe reinforcement learning and

• 23:16Daniel Whitenack?

  Decision making. So assuming that people might know at least or be exposed to that, in terms of the connection to AI specifically, what about this kind of swarming? Or, you know, robots or these entities working together to do a task? What is necessary for AI to do in that swarming scenario? That's not, that's not necessary in just a single like self-driving car, like a Waymo car or something like that?

• 23:50Chris Benson?

  That's a great question. There's a lot of nuance there. So, so the self-driving car is having to navigate an environment in which it's both sensing and to some degree communicating potentially, although that is evolving over time, with, with other actors in that environment around it. And there is emergent behavior on what to do. There's kind of the classic cases of, you know, kids playing on the sidewalk and the ball bounces into the street, you know, and, and reacting to various unexpected events like that. But at the end of the day, that vehicle is still, which may be fully autonomous, is still only having to decision on its own behalf.

• 24:34Chris Benson?

  And so it has an onboard inference system, the onboard computers, the onboard models, or potentially that in combination with some remote that it's doing over some sort of connection, radio connection or cellular connection. But it's still decisioning for itself. The difference, according to my definition of a swarm, is that the emergent quality of, of the decisioning is arising from participation in many of the platforms that are participating in that unified mission or activity. But it has to decide for all of them and it may have to decide on platform A. We're looking at something, we're evaluating it, we're sensing it,

• 25:17Chris Benson?

  we're processing based on reinforcement learning and our agents are then saying, "Platform A, you go do this activity. Platform B, you go do that one. Platform C, you go do that one." And those are all slightly different activities as those platforms collaboratively work together based on the sensing and inferencing that's happened across the swarm to get an activity accomplished.

• 25:43Chris Benson?

  Does that clarify a little bit?

• 25:44Daniel Whitenack?

  Yeah, that that helps. And I'm trying, maybe this is the decentralized part of what you're talking about. But part of maybe what is a challenge in my mind to think about is how such a thing would happen without a central kind of so like, so to say like, oh, let's just take a very simple. Maybe it's simple, you know, in my mind, it seems complex as well, but we have 20. We have 20 robots in a manufacturing facility trying to do something.

• 26:21Daniel Whitenack?

  In one model, I could see where those, each of those robots sent detects certain things on a number of sensors, maybe cameras or temperature sensors or force sensors or something like that. All of that is communicated back to a central processing system. And it's really the single entity, the kind of master brain that makes the decision and communicates out a next step to all the, all the robots. But in the decentralized way, you talked about how there's not this kind of central, there may be a central command where maybe there's oversight or something, but things somehow happen in a decentralized way.

• 27:04Daniel Whitenack?

  That's what I'm trying to connect in my, in my brain, maybe.

• 27:07Chris Benson?

  No, that's, that's a great nuance that you're pointing out there. So in the warehouse, that's not really, that's definitely a centralized environment and you don't even have to have onboard compute for that because your robots can have a highly reliable system of communication over wifi, you know, with a master server that's there and you're in the distances involved are close enough to where you can have them interact with each other without having to do that. And according to the definition of the swarm that we're using here, that would not be a swarm, even though you may have many platforms working.

• 27:42Daniel Whitenack?

  Multiple devices.

• 27:43Chris Benson?

  Collaboratively, they're being controlled by, by a local centralized agent. And in swarming generally, according to this definition, you're really.

• 27:53Chris Benson?

  Looking at an environment where that's not possible. You're looking, you know, what we would traditionally call edge or, you know, far edge is another term that will be able to use. Basically where you can't count on either local or cloud environments to provide the compute. And so you have to have the compute onboard, including all the activities that will arise from that compute.

• 28:16Daniel Whitenack?

  Yeah, I'm almost imagining moving, it's still kind of in my mind, living in the industrial world. If I think more to like oil and gas and like certain of the environments that are part of, part of working in that industry. Whether it be deep under the ocean or in very disconnected, remote places where you might not even want to, want to, or even could expose humans to those environments.

• 28:49Chris Benson?

  That's right.

• 28:50Daniel Whitenack?

  But they're also very hard in terms of connectivity and, and that sort of thing. You might want to accomplish tasks in, in that sort of environment. Am I getting to the right kind of scenarios?

• 29:01Chris Benson?

  You are. And like one good one that is commonly used is the notion of disaster recovery. Where you might go into an area that is geographic, you know, it, it may be geographically, you know, remote or, or may not be. But either way, the infrastructure has likely, things have happened with the infrastructure to where things have been torn down. You don't have cell towers, things like that. And you don't have the level of connectivity necessary for centralized control. And so if you're thinking about disaster recovery in a swarm context, you, where you do not have that infrastructure in place to guide those.

• 29:40Chris Benson?

  Then you could put drones or robots on the ground that have their own inference capability. And if you're approaching it with that decisioning entity that is decentralized that I've described. Then that could occur where they're actually working collaboratively to save lives. You know, if it's, if there's buildings in rubble, they're, you know, One robot that has that ability is maybe pulling rubble off because it has that capability. While others that specialize in getting down between the rubble and they're smaller. It might be, in other words, not all robots may be the same. You might have a heterogeneous mix of robots with different capabilities. But they're able to dynamically go, here's a rock and it needs to be moved.

• 30:25Chris Benson?

  While you go under the rock, you other platform go under the rock looking for a survivor there. And so you can have a collection of robots that each has an optimal function. Yeah, where they're working together to do something and ultimately save a life. And you don't have, you know, maybe you've had a hurricane come through or tornadoes. And you don't have the normal digital infrastructure that we're all so used to today.

• 30:52Daniel Whitenack?

  Well, Chris, that is really helpful to break down maybe some of the use cases or scenarios that maybe compare and contrast multiple agents or robots or systems that are working in an environment that maybe are in a swarm or are not in a swarm. Maybe a part of, so that's helpful on the use case front. My mind is still kind of wondering on this side of the AI side. What unique AI problems are unique to the swarm environment that aren't encountered elsewhere?

• 31:35Daniel Whitenack?

  And how do those map to kind of maybe open challenges that are, that are open? Or kind of maybe known models that are able to handle certain of these things?

• 31:45Chris Benson?

  Yeah, so there's, I mean, I think you just kind of alluded to the complexity involved. And I think the reason we haven't had swarms before now, and that they are still something that, that is emerging within what is being developed is because not all of the technologies and models that you would need have been mature enough up until this point in, in time to be able to do that. So there's a lot of pieces. And some of that is on the AI side and some of that is on is operating in a physical environment. And so on the AI side, you know, over the years, we've talked about these different architectures, you know, of different models with different purposes.

• 32:29Chris Benson?

  And swarming ends up taking quite a few of those and making them available to agents. So that you can, within a complicated physical environment, different agents on platform are working with each other to accomplish the on-platform parts. And then be able to communicate and coordinate that with other platforms that also each are multi-agent, multi-model configurations. And so, and then being able to do that.

• 32:58Daniel Whitenack?

  A swarm of swarms?

• 32:59Chris Benson?

  Yeah, sort of like if you think of all the various technologies that go into just a single platform able to do that. We're only now really getting, you know, in the large, getting to where these kinds of things are possible. And where the compute is powerful enough for edge devices and that the models are being reduced. We've talked a lot over the last year or two about how models are getting smaller and you go to hugging face. And, you know, the vast majority of models there are very small models that are useful in a lot of specific tasks. And so you have to match up small models that can run on edge inference or even CPUs to do various tasks for specific agents and coordinate that with sensors and comms.

• 33:43Daniel Whitenack?

  Yeah.

• 33:44Chris Benson?

  To be able to accomplish stuff. So it's quite complicated. And that kind of, and I use this kind of in air quotes, that kind of miniaturization of the technology. I don't mean physical miniaturizations, but kind of getting everything to where it can run on a remote edge device that is likely battery powered is a bit of a challenge. So I think as we look forward right now, there's, you know, we're really at an inflection point in history in terms of being able to do this.

• 34:14Daniel Whitenack?

  Yeah. And what? Just at a very practical level, I'm sure there's a great diversity here. But in some of what you're talking about, there's a unique communication element and goal-setting element of these systems. And if I look at like AI models that are running on any particular platform within the swarm, what, at a very high level, what kinds of models and tasks need to be accomplished that are associated with the swarming versus just like sensing and that sort of thing?

• 34:51Daniel Whitenack?

  Are these models that are, you know, the small SPMs or something that are making a kind of binary decision or gradient boosting machines or neural networks or Gen AI models? Like, what is the diversity you see and what are the kinds of tasks? You just, by way of example, a couple of them that might need to be done?

• 35:14Chris Benson?

  Yeah, it's a mixture of different models, some of which can be more traditional AI models, like what we most often talk about here on the show. Some of them are more of the classical data science models because, you know, we've talked about this a number of times where you don't need to go to a bigger, harder, you know, more expensive model to use if a smaller model will work, but it's very task-specific. And then, and it's bound within the software systems of each platform, but also the software systems which govern sensing and comms. And then, so you need the ability for different nodes, if you will, of what, you know, different platforms to maybe, on behalf of the swarm, to take on some of the computation because

• 35:56Chris Benson?

  you can't distribute a model inference across all of the things. But what you can do is say, if you have a swarm of a dozen, I'm just obviously making this up off the top of my head, have three of the, you know, of the platforms running inference on a particular model for a particular task while three others do another thing. And then you can, you can have things like election algorithms which decide which one you're going to take. And so you can, and you can have a valuation of what, who's getting Getting the best sensing information and who has the best comms reach for the entire swarm to ensure that data gets around. So you can have some redundancy in the computation and all

• 36:38Chris Benson?

  those things. And then by election, pick and election is just one mechanism. That's one common algorithmic mechanism that that one can apply. But by election within the swarm, choose who is issuing, you know, the results of the various types of inference that the agents are putting together and deciding upon a task so that you can then go and assign specific platforms to do different parts of that task where not every platform is doing the same thing, but they're all contributing to the larger goal.

• 37:07Daniel Whitenack?

  Gotcha. Yeah, that's very helpful. Now, I'm sure there's some people out there that are maybe excited by this, disturbed by this topic, maybe, I mean, I mean, in the, in one sense, just again, thinking about things outside of the robotic world. We talked about the birds, which are beautiful and you see the the patterns which they which they form. I also think about, like, in the human context, there's interesting, maybe swarm behaviors that are somewhat disturbing. Like, if you think about a mob mentality, you often hear about people, you know, maybe a mob

• 37:50Daniel Whitenack?

  of people all accomplish a very disturbing thing, like a destructive thing. And then afterwards, you know, individuals that were in that swarm say, I don't know what, you know, the individuals might have not ever have done what was done by the mob, right? But because of this emergent behavior, things happen that maybe are outside of the bounds of any individuals, you know, moral compass or that sort of thing. So if you if we take this then to the, you know, robotic or drone or industrial side, some people might say, well, how do we, how do we push one

• 38:31Daniel Whitenack?

  direction and not the other direction? And what is the oversight related to this?

• 38:37Chris Benson?

  So it's a great question. I'm glad you brought it up. In my enthusiasm for the technology, it's very easy to lose sight of this. And we had a a very recent show where we talked a little bit about, you know, the anthropic. I would refer people back to that and the fact that you had, you know, models that were coming out with what we deem to be bad outcomes. And that can certainly apply here. And so you have to have some form of guardrails. And so there's two, I want to start with, there's two phrases, one of which I mentioned before, I said, human operators on the loop. And human operators on the loop versus human operators in the loop are two distinct things. Right now, in the, in the military space,

• 39:21Chris Benson?

  based on current regulations here in the US and stuff, you, you most often are talking about in the loop interactions between humans and stuff. But as you go forward, and this is not a military specific thing, and this is something that you're going to see in commercial and industrial as well, as you start talking about swarm capabilities with emergent, you know, emergent properties happening, then the challenge is that there are some times you want a human to be able to step in the loop directly into that processing and say, yes, no, or make a selection or something like that. But there are other times when you're talking about a larger swarm where you need a human where the, it may be, there

• 40:05Chris Benson?

  may be too much going on, and there may be too many of these platforms flying around or on the ground for one human to track everything at every moment and interact. But you can have a human on the loop, which can say for the overall goal or for the thing the swarm decides approve or disapprove, you know, if there, an emergent behavior comes up and the swarm makes a decision based on what it's seeing in real time, then you can say, no, that's not going to work. And that's what a human on the loop does is you can, it gives you that guardrail of saying, that's not the emergent behavior I want. You know, it's time to go back or maybe do a full recall and shut down. So I think if

• 40:47Chris Benson?

  If you really want to learn more on these kinds of things, you're starting to see I it's I'm not sure that I can think of a specific swarm. I'll have to do some research and see if there are any specific swarming classes. A lot of the kind of drone and robot systems that you can get out there in open source. For instance, there's Ross and Ross two, which is the robotic operating system, first and second version. Ross two takes a slightly different approach to Ross one. So there are some people on each version. I know Ross two has some swarming capabilities in there to build on. It doesn't instantly give you swarm capability, but it has some tools in there that you can start building on.

• 41:31Chris Benson?

  And that might be a good place to start to learn about the topic. But in general, kind of staying tuned into learning sources, including this podcast, where you can learn about multi-agent environments and as we increasingly are talking about physical world deployments of AI technologies, those are all good places to to consume because this is a very cutting edge area to be focused on.

• 41:57Daniel Whitenack?

  Yeah, it seems like a lot of, a lot of things are developing and I personally am very excited about some of the things happening in robotics. Obviously, we've maybe some of us, the more we see chat interfaces. It's like we just want to see some other application of of AI technology, which that is only, that's such a small segment of what it, what is possible with AI. Even Gen AI, there's so many more ways that that will, I think, become embedded in our physical spaces. If you just want to search for kind of this idea of physical AI.

• 42:40Daniel Whitenack?

  I know that's a, a big topic of kind of AI embedded in our physical spaces. Maybe not through chat interfaces, but through other interfaces. And one thing that I've really enjoyed seeing is the stuff coming out of Pollen Robotics. P O L L E N. That's a company that I believe if if I'm not mistaken, was acquired by Hugging Face. And they're releasing a lot of, well, I don't know a lot. They're releasing robotic systems that are more geared towards.

• 43:16Daniel Whitenack?

  Experimentation, open source development, integration of open models, integration of apps and exchanging of those apps even within hugging face spaces, but for actual, physical systems. Actually, Prediction Guard, our company, we got one of the, or ordered one of these, I think it's Ritchie, many robots. Looking forward to playing around with that in our office. That should be really fun. So I think also there's an increasing number of ways that people can access even physical systems or robots in a much more accessible way than before,

• 43:59Daniel Whitenack?

  where maybe everything was coming out of whatever it was, Boston Dynamics or wherever, very, very expensive, millions of dollars systems. Here, there's ways to access these kinds of systems, whether it's drones or robots or that sort of thing, in a much more accessible way, which is really encouraging.

• 44:19Chris Benson?

  It's not expensive anymore. So in the maker space, there is a lot of resources. A lot of it is based on stuff you may already know, like Arduino or Raspberry Pi. And, you know, adding a Jetson in there, carrier boards where you could basically get some basic stuff. So we're going to see more and more of this going forward. And as people really kind of get even some of the more advanced topics figured out, it's going to be in all of our lives in the years ahead. So we're going to see a rapid introduction into daily life of these things. And so if this topic interests you, definitely, I would encourage you, just as I do, is to go out and jump into the maker space. And for very little money,

• 45:02Chris Benson?

  you can have some pretty interesting experiences building some of these things.

• 45:06Daniel Whitenack?

  Yeah, and it's occurring to me that after this, we'll put up maybe a webinar where we could talk about these things a little bit more, maybe even demonstrating some things with robotics. So don't forget to check out practicalai.fm/webinars for upcoming times that it will be live where discussions can happen. And yeah, just really appreciate you helping us deep dive into these subjects, Chris. It's been really good.

• 45:38Chris Benson?

  It's a fun topic. Happy to do it. Thanks a lot, Daniel.

• 45:41Daniel Whitenack?

  All right. See you soon.

• 45:42Chris Benson?

  See you later.

• 45:50Speaker 00

  All right. That's our show for this week. If you haven't checked out our website, head to practicalai.fm and be sure to connect with us on LinkedIn, X, or Blue Sky. You'll see us posting insights related to the latest AI developments, and we would love for you to join the conversation. Thanks to our partner Prediction Guard for providing operational support for the show. Check them out at predictionguard.com. Also, thanks to Breakmaster Cylinder for the beats and to you for listening. That's all for now, but you'll hear from us again next week.