Wildfires have burned hundreds of thousands of acres across the U.S. this summer. Now, firefighters are turning to new technology to help battle blazes, from AI cameras to vegetation-munching robots.

Guest

Scott Bordenkircher, Director of Forestry and Fire Mitigation at Arizona Public Service, the state’s largest utility.

Patrick Roberts, Senior political scientist at RAND. His latest research report is “Accelerating Technological Innovation Across the U.S. Wildfire Management System.”

Also Featured

Tanner Holt, Resident of Monticello, Utah. He lives about six miles from where the Babylon fire is currently burning.

Jamie Barnes, Utah State Forester and Director of the Utah Division of Forestry, Fire and State Lands.

Arvind Satyam, Co-founder and chief commercial officer of PanoAI, an AI wildfire detection company.

Ford Ainslie, Senior director of partnerships and growth at BurnBot, a remote wildfire mitigation and vegetation management company.

____

Transcript of Full Broadcast

The version of our broadcast available at the top of this page and via podcast apps is a condensed version of the full show. You can listen to the full, unedited broadcast here:

Part I

MEGHNA CHAKRABARTI: Tanner Holt lives in the path of the largest wildfire currently burning in the United States.

TANNER HOLT [TAPE]: So the fire, I think right now is probably oh six to eight miles from us burning on the backside of our mountain west to east over the top of the mountain towards us. Last night, people could see the halo of the fire on the backside of the mountain. So, uh, if you were outside in the dark, you could see the orange glow.

Holt lives in Monticello, Utah, in the southeastern part of the state. That’s where the Babylon fire has been burning for the past two weeks. It’s one of the largest fires on record in Utah, and it’s burned at least 101,000 acres so far, including parts of Bear’s Ears National Monument.

HOLT [TAPE]: This morning there’s a pretty good haze over the mountain. It’s blowing into town. Uh, you can definitely smell it. It smells like a campfire everywhere you go. Right? And the people north of us, the wind normally is blowing northeast, and so communities like Moab, which a lot of people are familiar with, they’ve had the smoke probably even worse than we have. We see ash on windshields in the morning and throughout the day. I mean, yes, it’s all the things that you hear about with wildfires, definitely.

CHAKRABARTI: Holt heard that the Babylon fire started with a lightning strike in a dry field. The U.S. Forest Service hasn’t determined an official cause yet, but Holt says he and his neighbors are not surprised to see the land going up in flames so quickly.

HOLT [TAPE]: This is a tinder box waiting. It’s been waiting for a match for a long time. In the last two years, we’ve experienced significant drought conditions and so highly dry fuels, lots of fuel within our community. We’re farmers, ranchers. People watch the weather. That’s what people do. And so because of that, I think that we’ve, as a community, generally speaking, we’ve been bracing for something like this for years.

CHAKRABARTI: Monticello residents are waiting. Holt says to see if the fire gets roughly within two miles of them. Then he expects fire officials will ask at least some people to evacuate.

HOLT [TAPE]: They use the model. Ready, set, go. Everyone in Monticello and the surrounding — Blanding areas, for the most part, you’re in ready right now, which just means get your stuff ready. Set is, “Hey, have your bags packed. You ought to be at least paying attention.” And go means evacuate.

CHAKRABARTI: This is On Point. I’m Magna Chakrabarti. Wildfires have become an almost year round fact of life for much of the Western United States. In fact, I was in Arches National Park in southeastern Utah just two weeks ago.

Hiking out to the double “O” arch, when I overheard a kid ask her dad why he was taking a picture of what was then just a thin column of smoke winding up into an otherwise flawless blue sky. At the time, it was just a wisp on the horizon, that threat of smoke. But by that night, the smoke had drifted more than 70 miles north and had blanketed Moab blotting out the stars and turning the almost full moon red.

It moved that fast. That’s why early detection is still one of the critical tools for early wildfire containment. Utah State Forester Jamie Barnes says most of the time, firefighters learn about a new wildfire when someone calls 9-1-1 or the wildland firefighter dispatch.

JAMIE BARNES [TAPE]: A call will come into dispatch, say for a smoke check or you know that there’s an actual fire.

They get reported up just like a normal accident would get reported up. And then we have people all over the state. So then we dispatch those people out and depending on the size of that fire, we start to deploy personnel.

CHAKRABARTI: But Barnes says that relying on human eyes is an imperfect system, especially in places where there aren’t that many humans to begin with.

BARNES [TAPE]: Along our Wasatch front area, if you get a plume of smoke, people are there, they’re gonna call it in. But in areas where, say people are only there on the weekends and you get a plume of smoke midweek, it may be different. You know that you’re not getting those calls from people because they’re not as heavily populated.

So lately Barnes has been wondering, is there a better way?

BARNES [TAPE]: We have started to use some AI technology like cameras that can detect a plume of smoke. And I don’t have exact numbers, but right now we have  some test cameras down in one of our highest risk areas, which is the Duck Creek area down south. And there’s a lot of stuff out there that we’re looking at to get better response rate and things like that.

CHAKRABARTI: Now Utah is far from the only state experimenting with AI and other technologies to detect and manage wildfires. At least 17 other states are too. Places like Arizona, California, Colorado, Idaho, Oregon, Washington and Wyoming.

So could it help control fires more quickly, so fewer of them become massive infernos like the Babylon Fire currently burning in Utah? That’s what we wanna look at today. And we’ll start with Scott Bordenkircher. He’s in Arizona. He’s director of forestry and fire mitigation at Arizona Public Service.

It’s the state’s largest utility. And he is with us from Phoenix. Scott, welcome to On Point.

SCOTT BORDENKIRCHER: Thank you Meghna. Glad to be here.

CHAKRABARTI: So, first of all, let me just get some housekeeping out the way. Explain to me a little bit about why the state’s largest utility has an interest in controlling or even mitigating wildfires.

BORDENKIRCHER: A couple of reasons for sure. So a lot of what we do year round focuses obviously on ensuring that our infrastructure can survive any kind of wildfire that’s approaching it, because having power is critically important, especially in times like that. But the biggest reason really is the fact that we are citizens and, and live in this state and in these communities.

And so frankly, the safety of ourselves and our customers is really paramount in our thinking. And so we need to ensure that we can do everything we can to mitigate this really large risk for our state.

CHAKRABARTI: Okay. Makes a lot of sense. I just wanted to ask to see if there was also any other surprise reason in there.

So you just heard Utah officials say that look, most of the time the way that fire wildfires are detected is kind of the old fashioned way that people see a plume of smoke and call it in. What’s the typical method of fire detection there in Arizona?

BORDENKIRCHER: So prior to the use of artificial intelligence cameras that we’ll be talking about, uh, it’s very similar.

So we have one of the largest ponderosa pine stand forests in the nation which people are typically surprised by because they think of Arizona as just being desert, but lots and lots of forest in the northern part of our state. And so while the forest service does maintain some level of watch towers up there.

As stated earlier, when people aren’t there and they don’t see that smoke when it first appears, obviously there’s a big delay in reporting that fire and therefore a big delay in getting responders there. The other issue with that is while somebody may see the smoke, they don’t necessarily know exactly where it is, so you also have the delay to first responders and sometimes just finding where they need to send resources to fight that fire.

CHAKRABARTI: You know, that’s actually an incredibly important point because we’re talking about vast areas of wilderness, right? And you can see smoke on the horizon, or maybe even just a little bit closer, but it still could be within many thousands of acres, right?

BORDENKIRCHER: Correct. Absolutely. People think when a fire starts in town, typically there’s landmarks.

Folks know exactly where the, where it is. You can reference it by street or or corner. But when you think about in the middle of the wilderness, in the middle of the forest, not only is it hard to reference the location, but sometimes you’re not even sure exactly how far it is, and I don’t know about most folks, but I have a little trouble deciding is something a mile away or you know, 500 yards away, or how far it really is. So being able to give exact locations is really, really critical.

CHAKRABARTI: So let’s talk about the use of AI here. I understand that you have kind of a network of AI enabled cameras, Pano cameras out in the forest. So tell me about what they are and, and how they work.

BORDENKIRCHER: We’ve invested to date in about 55 Pano AI cameras, and their chief job is to utilize artificial intelligence to recognize smoke during the day and heat signatures during the night.

So in fact, these cameras are able to detect wildfire both day and night, which is really important. And what happens is when that wildfire is detected human eyes at their control center puts a brief look on it to just ensure it’s not a dust devil or some type of false positive. But then two things happen and that is alerts go to us as the utility in case there are actions that we need to take.

So maybe we need to de-energize lines or do something for the safety of firefighters. But then the more important thing is those alerts go directly to dispatch centers for the first responders. And so we’re seeing that a dispatch center notification might beat a 9-1-1 call by five minutes or up to 25 minutes, or in the best or worst case, depending on how you look at it, maybe any 9-1-1 call, because not one comes in at all.

The other thing that those cameras do, and I mentioned this earlier as being important, is they triangulate the direct and exact location of where that fire is. And so first responding resources are getting notified of the fire start. They’re getting the location of where that fire started, and they’re getting images of what that fire looks like from a terrain perspective and a size perspective.

CHAKRABARTI: Well, we actually also talked to the co-founder of Pano AI. So when I said “Pano” cameras, I should be specific and say Pano AI is a company that makes these cameras. Arvind Satyam is the chief commercial officer and co-founder, and he told us that their AI wildfire cameras are trained on some 4 billion images.

ARVIND SATYAM [TAPE]: So these are two cameras are continuously rotating. They’re doing a full rotation each minute, and this gets stitched together into a 360 degree panorama. So think of the concept of a fire lookout tower, and then we apply artificial intelligence to that imagery, and we have multiple models. We have a daytime model that’s essentially looking at smoke and not smoke, and then at nighttime it’s looking at heat or not heat. And just as you would with self-driving cars or other applications of AI, we have a very large training set that we continuously update.

CHAKRABARTI: So, Scott, let me just quickly ask you about, have about 30 seconds before our first break here. I’m trying to visualize what these cameras look like. How big are they? Are they up in towers? Would you notice them if you were hiking by?

BORDENKIRCHER: So, so they look spherical. They’re kind of like a ball, maybe the size of a of a volleyball or so. Yes, they’re typically up in high locations, so we wanna make sure that these cameras have good views of as far as they can possibly see.

So these are not ones you’re gonna see like on a light pole or, or kind of deep in the city. You’re gonna see them up on mountains, up on radio towers, television towers, things like that, so that they get a very, very high encompassing view.

CHAKRABARTI: Well, you’re listening to Scott Borden Kercher. He’s in Phoenix, Arizona, and we’re talking about what essentially is hopefully a good news story about AI technology and how it may help detect and even fight wildfires early.

Part II

CHAKRABARTI: I wanna just go back to something you had said a little bit earlier. So these Pano AI, that’s the company that makes these spherical cameras, as you’re saying, you put them up in towers or you know, on mountains to have a, a high broad look over large tracks of land.

And you said that the speed with which the cameras can detect smoke or heat and then triangulate their location and notify first responders, that sometimes those cameras have alerted first responders faster than a 9-1-1 call can come in and they’ve done it by five to 25 minutes faster.

Now, help me understand, is that a lot of time? Because on the one hand it might be like, well, what is 25 minutes in fighting a wildfire?

BORDENKIRCHER: So depending on the weather conditions and the fuel conditions, and I think we all know certainly in the western part of the United States and even as drought continues to work its way across the U.S., those conditions worsen.

And so even the five minutes, especially when first responders know where they need to go and what they’re up against is hypercritical. A fire in a good wind can get out of control very, very quickly, and so being able to get on scene and on scene in the right spot means the world in terms of the difference between a small fire that is easily contained and what can become a catastrophic fire very quickly.

CHAKRABARTI: Got it. So it’s not just the notification time, that’s faster, but because the location is determined, all the personnel and equipment can get there even faster. So does that mean that it’s basically the time saved is probably much more than 25 minutes?

BORDENKIRCHER: I would guess that it probably is, absolutely.

CHAKRABARTI: Okay, so I understand that you’ve had these cameras in place in Arizona for what, about two years now? How much is this costing the utility or the state?

BORDENKIRCHER: So we look at it from the standpoint that the investment is really to save lives and property. And so we don’t look at costs from the perspective of what an individual camera costs.

We look at it in terms of what’s the value. And the value here is really in safety and in reliably being able to deliver the service that our customers depend on, frankly, to live.

CHAKRABARTI: Okay, well, let me bring in a new voice into the conversation. Now, I’d like to introduce Patrick Roberts. He’s a senior political scientist at RAND, and he’s been doing a lot of research on technology and wildfire fighting.

His latest report is titled “Accelerating Technological Innovation Across the U.S. Wildfire Management System,” and he joins us from Washington. Patrick Roberts, welcome to On Point.

PATRICK ROBERTS: Thank you.

CHAKRABARTI: So first of all, give us a broad, sort of overall picture of what is it about wildfires in the U.S. that makes them ripe to really be a good target for technological innovation?

Like what is it about fires that make technology such an important potential tool?

ROBERTS: AI and wildfire and technology is really mostly about seeing earlier mapping risk, better predicting spread, and really helping people figure out how to prioritize action and what what to do.

So it’s not just that the AI will put out the fire. But AI and other technologies can provide decision support for, for the utilities, for fire chiefs, for land managers, insurance companies, emergency responders, to figure out how to take action when there’s a fire, or how to reduce risk before a fire.

CHAKRBABARTI: So tell me more about the how to reduce risk part.

ROBERTS: The goal is, as Scott said, really reducing the incidence of catastrophic wildfire, wildfires that really harm people and property. Some fires are a part of nature’s cycle of renewal, but some really threaten people and property on high winds days, they can be quite risky and so figuring out how to put out fires where they need to be put out, or reduce risk, either protective actions in a community, in a home, home hardening, or maybe, thinning vegetation that that is likely to catch fire. That’s an important risk reduction activity and AI and other tools can help figure out where do we prioritize that, that vegetative thinning, other risk reduction activity.

CHAKRABARTI: Okay. So can you tell me more bit more about how that happens? Because I mean, who’s making the decision about where the priorities should lie? The AI is just recommending?

ROBERTS: AI at this point is decision- support, and helping process a lot of unstructured data for people who make decisions. They’re the communities. They might be forest managers in forests or national parks. They might be state land managers. They might be fire chiefs in a particular county or in a community. So there are a lot of different people in this wildfire management system. There might be property owners all involved in thinking about how to manage risk and reduce their risk.

CHAKRABARTI: Mm-hmm. I guess what you’re describing is that there are so many different facets to why wildfires are these sort of overwhelming and complex problems in the United States. It’s everything from just managing a fire once it starts to, as you talked about, hardening. Can you give me some more examples, Patrick, of places in which AI and technology are focusing on improving management throughout this complex system?

ROBERTS: Sure. I think there are some technologies, AI technologies used before a fire to reduce risk. And some cameras can be used for that. But also, satellites, are being sent up to help understand ecosystems and understand fire behavior and where is a fire more likely to spread, how is it likely to spread and then where are the risks. And in really processing all this complex earth and human systems data to help people figure out what they should do, where they should maybe put in fire breaks and evacuation routes and really where they should thin vegetation and prioritize areas that are most likely to burn and threaten people.

CHAKRABARTI: Okay. Scott Bordenkircher, let me turn back to you on that. Are these sort of additional, other than fire detection, that you’re using with the Pano AI cameras, are these additional tools that AI and technology can provide?

Are they being used by the Arizona Public Service, the utility?

BORDENKIRCHER: They absolutely are Meghna. So we have a whole suite of software that allows us to model fire, allows us to see where a fire might burn towards or allows us to prioritize then the efforts that we put into potentially designing and or hardening our grid, our own electric infrastructure.

We also have a, a wide system, a network of weather stations. And so using, uh, starting to use artificial intelligence and just general computing power behind the scenes to crunch that information and provide our meteorologists and our fire scientists with, again, as Patrick said, really decision support. So where those areas that are particularly risky throughout our state, where is that risk may be growing or changing as forest health starts to decline or continues to decline. And so those systems are very much in use throughout our entire wildfire mitigation program.

CHAKRABARTI: Okay. I wanna go back to one thing. A little earlier, Scott, I asked you about the cost of these cameras and I’m seeing here that the Associated Press, they did a story about this and they’re reporting that Pano AI charges about $50,000 a year per camera that they put out into the field. But that $50,000 per year also includes the sort of the data analysis, the risk analysis, and that 24-7 intelligence center where the information comes in.

And how many cameras does Arizona have out in the field right now?

BORDENKIRCHER: So we currently have 55 out in commission in the field.

CHAKRABARTI: 55. Okay, so we’re talking about maybe two and a half million dollars a year. Then, Patrick Roberts, let me turn back to you because I am conscientious of the fact that we don’t wanna speak about AI as this, all knowing, all seeing perfect tool because in every use case that we’ve ever talked about on this show about AI, there’s always issues of, well, is it coming to false conclusions? Is it hallucinating? Is it providing bad data based on biases built into the the vast amount of code that creates the AI? I’m just wondering what are the potential risks of using AI? False alarms, perhaps, that these cameras — just to stick with them as an example — that the cameras could deliver?

ROBERTS: Yes, there are always risks and false alarms or false positives that they say happen. That’s why the key is in all AI uses, not just to rely on the tool to tell you what to do, or not just offload your thinking onto the tool, but really incorporate the tool into your work process, into your operational process, into your decision making process.

As we heard about Pano AI, they incorporate the camera notifications into human review. And then once there’s a review, there’s a decision, and a particular distinction, about what do you do about a fire. Again, not all fires are bad. Some are part of nature’s cycle of renewal. There might be a rainstorm coming and maybe it’s not that serious, or there might be high winds coming and it is serious and they have to figure out what to do, where to send people, and so that’s really part of fire management. This is one tool to provide additional data more quickly. And in serious, potentially catastrophic fires, time is of the essence and minutes matter.

CHAKRABARTI: Okay. So Scott, I’m wondering how the continuous training works with the cameras that Arizona is using from Pano AI. Are there things that just sort of happen as normally in Arizona climate that could lead the cameras to the occasional false positive?

BORDERNKIRCHER: Yeah, absolutely. So, Arizona has lots of strong winds at times. Obviously we do have a lot of desert sand and dust, and so we get dust storms, we get dust devils, which tend to look like columns of smoke.

We have quite a bit of agriculture here in the state of Arizona. And so, something as simple as a truck driving down a dirt road or a tractor can certainly put up dirt and dust and debris that looks like smoke. And so part of what Pano AI constantly does is look at those images and continues to train their AI so that what doesn’t happen is that alerts don’t get sent out in those circumstances. So they’re really working diligently to ensure that what gets sent out is truly the positive positives, and that those other things are filtered out.

CHAKRABARTI:  So then, let me follow up because I think Patrick said a little bit earlier that AI in the case of wildfire management is a decision making, it assists decision making rather than makes a decision in and of itself. So when an alert comes in through the Pano AI system, before firefighters are dispatched, or decisions are made around what to do, is there a human check in there somewhere?

BORDENKIRCHER: Absolutely for us. I mean, we’re looking at the images again to decide how far that is from our infrastructure. What do we need to do? And then I think Patrick said it very well from a firefighting perspective, when they get those alerts, again, they’re doing the same thing. They’re certainly looking at the conditions on the ground, what the winds are doing. Is it that type of fire that they, frankly, can let burn? Because some fires, again, are very, very good for the health of forests and good for the landscape. Or is it something that they need to immediately  jump on and and attack? So I think humans are making those decisions on what to do with that information, but this allows them to, again, know the location and the size and where it is, and to be able to make that decision very quickly.

CHAKRABARTI: Got it. So basically,  people are seeing the alerts come in and they’re doing pretty knowledgeable evaluation about going back to whether it’s a false positive or not, or if it’s not a false positive, what do we do? Who should we deploy, et cetera. Patrick Roberts, I was wondering if you wanted to just add more to that about the continuing centrality and importance of people running these systems.

ROBERTS: Absolutely. And in talking to fire managers, fire chiefs, I don’t see them as, as too trusting of AI or too credulous. They’ve been fighting fires, managing landscapes for a long time and have a really rich, deep culture of how to do that. I do see a need to get land managers and fire managers and even utilities ready to use these technologies and ready to incorporate them into their work systems and operational teams, and that’s something we try to help with in our research. Figure out how to link these great innovators and innovations and technologies coming on the scene with the users, the fire managers, fire chiefs, land managers that have been doing this managing fire for a long time, and how to help them do what they do, just a little bit better.

CHAKRABARTI: Okay, so we’ve been spending a lot of time talking about cameras, because those are actually out in the field and Scott has some firsthand experience with how those Pano AI cameras work. If we’re seeing AI or just sort of more sophisticated digital technologies pretty much everywhere, I’m wondering are we seeing that being used in say, drone technology for surveilling and detecting wildfires, Patrick?

ROBERTS: Yes, we’re seeing autonomous vehicles, drones for, for surveilling,  for situational awareness, figuring out what’s going on in remote areas where it’s hard to send humans. Satellites too are being used for that purpose and also for gathering data on fire behavior. Fires are very complicated, so this new satellite data could help us understand and help support decisions.

A nonprofit consortium with California and others are sending up satellites built for wildfire, fire sat, just recently. And Canada is sending up purpose-built satellites for wildfire in 2027. I think that’s also part of the future, and it’s not either or cameras or satellites. It, it’s all of that giving different kinds of data to really help us understand fire and, and make decisions more quickly.

CHAKRABARTI: So is the use of satellite technology is not necessarily that new? I mean, I feel like that’s many, many years old because of the fact that they’re up there and they can see millions of acres at a time and they, I guess the heat sensing technology in particular has been pretty good. Or am I wrong about that, Patrick?

ROBERTS: No, exactly. We’ve had satellites for a long time. What’s new is purpose-built satellites for wildfire and for observing wildfire at the right resolution, the right level, in the right locations, uh, at the right time to really follow fires, monitor fires and understand them.

And then the new and AI or machine learning algorithms that can really help us understand all kinds of data so we can really intervene and burn fewer hotter, really destructive, fast-moving fires and maybe burn some cooler, easier fires that will help regenerate ecosystems and really manage the land. I think that that’s part of the future of some of this data in AI.

Part III

CHAKRABARTI: Patrick is in Washington, D.C., and he’s written about another company that I want to take a second to learn more about. It’s a California-based company called BurnBot. And BurnBot has created a whole squad of robots that are designed to help prevent wildfires. Ford Ainsley is BurnBot’s, head of partnerships and growth.

FORD AINSLEY [TAPE]: We have the world’s largest fleet of remote masticator. I like to think of those as chew bots. They’re about the size of a golf cart, but they’re unmanned and remotely operated. They’re rubber tracked machines that have teeth on the front of them that really create mulch and chew up vegetation, which is really important for reducing your fuel density, your fire intensity, and your flame length.

CHAKRABARTI: Then Ainsley says there’s the BurnBot RX2, a big remote controlled vehicle that does actually burn things.

AINSLEY [TAPE]: Visually, the RX2 appears similar to that of a Zamboni that you see smoothing the ice, but instead of smoothing ice, we’re actually putting prescribed fire on the ground. So what the RX2 really is, is a mobile burn chamber that applies prescribed fire in highly controlled and near smoke-free way.

It’s output. And what it’s creating is a black line, which is really a pre-burn strip. That then can effectively act as your fire break and fire barrier. We’re able to run it back and forth and build the exact width needed that that prescription calls for for effective fire break and fire barrier.

CHAKRABARTI: Now that fire break, you may have seen pictures in the past of firefighters actually out there in the forest creating them with, you know, there may be a fire burning around them, but they create a fire line using flames of their own in order to chew up and burn up all those, that forest material so that a fire has no place to go Now, BurnBot’s doing that with its robots, and it was founded in San Francisco in 2022. And since then, Ainsley says the venture-backed startup has expanded to New Mexico, Utah, Montana, and Texas. And here’s how it works. Clients contact burn bot, tell them what kind of vegetation they want turned into mulch, or what areas they need to be burned with those prescribed burns. And then burn bot employees bring the robots to the job site and operate them.

AINSLEY [TAPE]: AI helps us identify the highest risk fuels where to prioritize treatments. All of the precision robotic operations are assisted by AI, how we adjust propane flow, air flow, foam water application. AI helps determine our appropriate operating speed, burn intensity, treatment consistency in the environmental conditions.

CHAKRABARTI: Since its founding about four years ago, Ainsley says BurnBot has treated over 20,000 acres in California alone.

AINSLEY [TAPE]: Our largest client is the U.S. Forest Service, where we help treat thousands of acres for them every year and are on some very large projects on the, the smaller end of that spectrum. Some of our clients are HOAs, schools, wineries. And then in the middle there, we work with a lot of cities, county governments, a lot of fire departments. It’s really anyone with an asset, whether that’s a home, a factory, whatever it may be, or a roadway or land that they’re trying to protect from wildfire risk and trying to reduce that wildfire risk.

CHAKRABARTI: Ford Ainsley, he’s head of partnerships and growth for the California based startup BurnBot. So, Patrick Roberts, tell us more about why BurnBot caught your attention and what it’s doing that maybe is significantly different from what human firefighters have traditionally been able to do.

ROBERTS: BurnBot is a great example of a company that came on the scene as you said in 2022. They’re a group of technologists in Silicon Valley and elsewhere working on self-driving cars and apps, and they noticed the, the growing wildfire crisis and it really posed a threat to California, but, but really the nation and said, “Why can’t we use our skills, our technological know-how to really address this problem of wildfire and reduce or eliminate catastrophic wildfire?” And BurnBot is one of those companies that really made it from an idea, from an innovation through prototyping and testing and actually deploying in the field and multiple states as you said.

So it’s a great example of a company that made that transition into use. And it’s also interesting and there’s a lot of attention to suppression or putting out fires, understandably, but they’re on the before, the risk reduction side and there’s a lot of potential there and a really, a huge payoff.

If you don’t have to pay to fight a fire and there’s not a risk there and not damage, wow. It’s a big payoff  from doing this risk reduction act activity.

CHAKRABARTI: Mm-hmm. Well, Scott Bordenkircher, let me turn back to you because  I’ve been hearing that in recent times. It’s been, and correct me if I’m wrong, but it’s been harder to recruit wildland firefighters.

The job is inherently very, very difficult, obviously, and also risky and human beings need to take breaks, whereas, you know, something like a BurnBot robot could, I don’t know if they run them 24-7, but I’m just wondering what you think of the advantages or disadvantages of this robotic technology are over using people out there trying to fight these fires.

BORDENKIRCHER: So BurnBot right now is not a technology we use, but certainly we’ve seen many, many instances where even with having clear utility right-of-ways, which is kind of the responsibility of the other part of our forestry and fire mitigation organization, those right-of-ways are used very often as burn-off points or back-burn points by firefighters. And so I think anything and anyone that is responsible for cutting those fire lines or making sure that there are some of those fire breaks already in place, definitely is critical in the fight against a fire becoming out of control and turning catastrophic.

CHAKRABARTI: Now,  I understand that you may not have direct experience on the fire line yourself, Scott, but there are folks on your team who are or have been wildland firefighters. Is that right?

BORDENKIRCHER: That is a hundred percent correct. So that’s not my background by experience. But our set of fire mitigation specialists all come from a wildland firefighting background, with a couple sprinkled in there city fire departments, et cetera.

So yes, those folks have an amazing skillset and an amazing long line of experience in dealing with this situation.

CHAKRABARTI: Yeah, so that human experience then coupled with this new technologies, it seems really exciting. I mean, what do they think about using AI, using the cameras, for example?

BORDENKIRCHER: The whole team is extremely excited about technology and what it can do in general. So clearly they see, especially given their backgrounds, the benefit of early response and early detection. If you think about the other side of the organization which is our meteorologists and fire scientists, those technological advances in those tools are hypercritical to them. Being able to make good forecasting, be able to judge risk appropriately and really allow us to kind of prioritize and decide what tools are the best tools that we can bring to bear and invest in and to ensure that we’re combating this risk.

CHAKRABARTI: So Patrick Roberts, I’d like to take an even sort of higher altitude look at this because we’ve been talking about mitigation efforts and detection efforts around when fires are about to start or already have started, particularly in the American West.

There’s also the question of can AI or is AI being used to help just overall forest management because there’s this long running debate about whether like the U.S. Forest Service has made the right decisions over many decades in terms of when to let fires burn, or how much woodland to cut down, et cetera.

Can AI be used to help improve how we manage our forests so that maybe that can over the long run reduce the dangers of these near-constant wildfires?

ROBERTS: Yes, absolutely. It can help in at least two ways. One, better understanding fire behavior and ecosystem health and prioritizing where to reduce risk where to thin vegetation or create fire breaks, et cetera.

I think satellites and cameras and processing and structured data can help there. It can also help just add extra, not literally hands,  extra tools like the BurnBot, which can thin vegetation or, or burn it very safely on hillsides where it’s hard for humans to go or in areas that may not have a full-time burn crew or maybe the burn crew is busy fighting a fighting fire.

So, this is some  additional help and there are people involved in the BurnBot and managing it just doesn’t drive off on it on its own like a self-driving car. There, there are people right there remote operating it and making decisions about when is it safe and when is it practical for a community to thin vegetation there, so it’s part of a governance system.

CHAKRABARTI: Mm-hmm. I’m thinking back to the fact that, look, reality check, this does cost money. And I’m wondering if state budgets are already stretched pretty thin here but fire is  a matter of life and death of the economic health of states across the American West.

Patrick, do you have a sense as to sort of the cost benefit analysis or, or how states should go about trying to to pay for these new technologies?

ROBERTS: Yes. That’s something we found that there’s a challenge for this field, but people are stepping in to try to solve it. Some of the technologies are expensive satellites, you know, that, that’s not gonna be cheap.

And some of the camera systems, there are some big players who can make the investment, the federal government, utilities and insurance companies and big states. Some of these technologies and the data can filter down over time. I think that’s likely to ha happen. We all cameras in our pockets now and in our phone but there are some simpler AI technologies also available to local departments that may not have big budgets. For example, alert and warning systems for, for different languages. You might need to warn people and instantly translate into a different language in your community. You know, AI tools can, can help with that as well.

CHAKRABARTI: Okay. Scott, did you have any thoughts on that?

BORDENKIRCHER: I think the hard thing is this is a national problem, right? So if you think about even from a utility perspective in the United States, utilities equipment start somewhere between two to 5% of, of fires, which means you’re talking about the, the 95 to 98% that are lightning caused and unfortunately, sometimes, careless humans or misbehaving humans. And so, really we need to be able to bring investment and finance from a variety of sources, and part of that is federal government. Certainly, I think states are starting to really step up in areas where they recognize the declining forest health and just the worsening conditions from a fire weather perspective, but this is gonna take a group effort. I think, Patrick states the idea of having sources of funding from private, public governmental partnerships. I think all of those things are gonna be necessary to combat this.

CHAKRABARTI: So Patrick, let me come back to you on that because both of you have mentioned something that I want to just make more clear.

Fires being what they are and how they burn, we have a lot of, just politically speaking, overlapping areas of authority, right? Local, state, national, different agencies, different local departments, et cetera. Is there any possible way that, or anyone that you know, that’s working on AI tools that could help coordination between these groups because sometimes it can get a bit sticky?

ROBERTS: Exactly right. AI can help the welfare problem, but it’s not just a question of whether any one technology or tool works. There’s this challenge that the welfare management system and technology innovation system is fragmented and we have a lot of really promising technologies out there.

But in our report, we found we need more pathways to really test them, to fund them, to validate them, to create standards for operability that operators can really trust when they wanna use them, and really bring them to scale across the full cycle of prevention, response, and recovery. And there is some lessons from other innovation agencies in defense, intelligence, health, et cetera, that we provided to the welfare community. And there are new organizations, offices, groups being created to try to really bring these technologies from the innovators to the users. These are being created at nonprofit organizations in fire agencies and really focusing on fire technology innovation and linking the technologists and innovators to the users as its own kind of activity.

I think that’s one way forward. And linking to funding from governments, pool funding, philanthropy has been important as well.

CHAKRABARTI: Well, we’ve just got a couple minutes left. And Scott, if I could, I’d love to bring this back to the level of people who are living with the reality of a near constant wildfire season.

At the very, very beginning of the show, you heard that Utahans say that they have this “ready, set, go,” system where they’ll know immediately if they have to evacuate. I grew up in Oregon and I’m in touch with people there who say that they basically always have a wildfire escape plan. They always have a go-bag ready in their trucks.

It’s just kind of become one of those challenging facts of life. Can you tell me from your experience there in Arizona, how have wildfires changed how you live on a daily basis?

BORDENKIRCHER: So I think there’s several things that have influenced that. I mean, certainly we all appreciate the beauty of nature and especially our forested areas. And so more and more citizens in the state of Arizona are moving into that wildland urban interface that the trade-off there though means that they need to be aware of the circumstances that can cause a catastrophic wildfire and the fact that they have that risk now near to them. So, so certainly in Arizona, we use the “ready, set, go” system as well.

And folks are constantly encouraged to help themselves. And so we talk a lot about in our forested neighborhoods, in our wildland urban interface neighborhoods about defensible space, ensuring that you don’t have that heavy brush and those high level of fuels directly up against your house or your property.

And so making sure that just folks are taking that little bit of time and that effort to give themselves the best opportunity for safety that they can is something that we continue to encourage and something that kind of just comes now with living in that area.

This article was originally published on WBUR.org.