How technology is helping prevent wildfires from spreading

0:00 These fires just devastate people, devastate communities.

0:04 If we can prevent this anxiety, and this fear, and deep-rooted,

0:10 just concern from people that have gone through this, that's a success.

0:15 First priority is life.

0:17 And the- the more that I can do to engage

0:19 in technology that helps protect my communities and protect my firefighters,

0:24 that- those are no-brainers.

0:26 That's Deputy Fire Chief Zachary Wells and Doctor Neal Driscoll,

0:30 a professor at the University of California San Diego.

0:34 Together, they are two of the bright minds behind ALERTCalifornia—

0:39 one of the most ambitious early warning systems for wildfires,

0:43 anywhere in the world.

0:44 The Microsoft AI for Good team has worked with Zach

0:47 and Neal to use AI and cloud computing to detect fires earlier,

0:52 share intelligence faster,

0:54 and give firefighters the real-time data they need when every minute counts.

1:00 ALERTCalifornia is an enhanced situational awareness platform.

1:03 We have cameras strategically placed on mountaintops to allow us

1:07 to visually identify and confirm fire in its incipient phase.

1:12 We talk about how this technology is already starting

1:15 to change the way firefighters protect lives and communities,

1:19 and what it will take to scale its life saving potential around the world.

1:25 Up next on Tools and Weapons.

1:28 Now, when did the two of you meet?

1:31 2017.

1:32 A Ventura County firefighter said,

1:34 “You have to hear about what they're doing at UC San

1:36 Diego and meet this guy named Neal.” And we've hit it off.

1:40 We've learned how to work with each other from the fire side,

1:43 the university side, and develop tools that can be turned over to firefighters

1:48 for them to use at no cost to them.

1:51 And Neal, what was the inspiration for you that got

1:54 you going down this path in the first place?

1:56 2003.

1:58 What happened in 2003?

2:07 Impacted a lot of people.

2:09 Was the largest fire at the time in California.

2:13 It skirted up my neighborhood.

2:15 Oh, really?

2:16 People—there was like 23, 24 people lost their lives.

2:20 We didn't know where to go.

2:22 So, you called 911 and said, “I can't see.” The whole sky was orange.

2:27 “Where do I go?” They couldn't tell me.

2:31 And this fire jumped all the lanes at Miramar.

2:35 My kids were so frustrated and scared.

2:38 They kept saying, “Are we going to be safe?” I'm still scarred by fire.

2:43 Sure.

2:44 I had to take a pause there.

2:45 Yeah.

2:45 And I didn't lose my house.

2:48 I didn't lose a loved one.

2:50 If we can prevent this anxiety and this fear, and deep-rooted,

2:56 just concern from people that have gone through this, that's a success.

3:01 These fires just devastate people,

3:04 devastate communities, compromise bio-habitats, changing the vegetation.

3:09 We need to use every bit of technology to try to hold them in check.

3:14 And I've been building this network.

3:16 I build instruments to bring signals

3:19 from out in the wilderness back to laboratories.

3:23 We decided to put a couple of cameras and actually

3:27 use the infrastructure to bring in information from remote areas.

3:32 I can have a firefighter from Oregon run my system if I need to import.

3:37 They don't have to know anything about the mountain ranges, the camera names.

3:41 The AI says, “Look at these cameras.

3:44 Something is changed since I've rotated or I've taken the 60 degree image.” So,

3:49 the AI is just saying, “Go look at the change.

3:52 Tell me if it's a fire or it's- it's the marine layer.

3:56 It's fog.

3:56 It's a dust devil.” So, all of a sudden we have a change

4:00 detector that allows watch-stander fatigue to be removed.

4:04 It reduces the noise.

4:06 If fire or an incident comes up, it just hits that screen and you can see it.

4:11 There's nothing else there but that fire or that event.

4:14 And then you can interrogate it and figure out what it is.

4:16 Can you say a little bit about how much territory you cover,

4:20 and then compare the size of that territory with the number of fire

4:24 engines or pieces of equipment and people that you have to cover that territory?

4:28 Yeah.

4:29 So, in Kern County, we cover 8,141 square miles.

4:33 It's the rough size of the three smallest states in the continental US.

4:37 And I protect that area with 47 fire stations.

4:41 So, each station has a fire engine with three firefighters.

4:44 About 170 firefighters in our department is on every single day.

4:48 Okay.

4:48 But 170 spread throughout 8,000 square miles is a big area to cover.

4:53 We live in a time when I think most of us

4:55 are used to having a smoke detector in our homes.

4:58 And the obvious principle is if you detect the fire early,

5:01 that's how you best put it out.

5:04 This is like in effect— an AI-driven smoke detector for this entire region.

5:11 Yeah.

5:13 What is the impact of early detection for somebody like you and your team,

5:18 as you're responsible for putting out these kinds of fires?

5:22 So, what I've experienced already,

5:25 the ability to detect fire and confirm it in its incipient phase,

5:30 respond in an effective firefighting force.

5:33 The first alarm of engines, and chiefs, and all the crews that we need,

5:38 and extinguish that fire without ever receiving a 911 call.

5:42 Going back to the protecting life, property,

5:46 and the environment— first priority is life.

5:48 And the- the more that I can do to engage

5:51 in technology that helps protect my communities and protects my firefighters,

5:56 that—those are no-brainers.

5:58 Running a fire department, we don't have an R&D budget.

6:01 We don't have time other than our priority is responding to emergencies.

6:04 We're not going to stop every single fire.

6:06 You know, it's- there's over 10,000

6:09 wildfires that happen in California every year.

6:12 So, seeing a program like ALERTCalifornia and figuring out

6:17 that we can invest our people and our effort,

6:20 opening up our radio towers to add a camera

6:23 on a tower that exists for our daily use,

6:25 we feel that we can make a dent in the universe and make an impact.

6:29 We already have an action plan and a system that works really well,

6:33 but we want to make it better.

6:34 We want to be able to engage those that are

6:37 the boots on the ground that are doing the hard work.

6:40 And obviously, a big part of your success is sending the right

6:44 people and the right equipment to the right place at the right time.

6:48 It's hard to predict where an incident is going to occur, but using AI,

6:52 using these cameras that are always looking and finding and, and observing

6:57 and then getting that signal through all the noise to the right people,

7:01 the dispatchers, the firefighters, so that they can make those decisions.

7:05 That's where we've found success.

7:07 That's where we'll continue to find success

7:09 as technology drives this to more advanced levels.

7:12 One of the things I find fascinating about the progress the two of you

7:15 are making together is that you are

7:17 making the technology better and cheaper- Yes.

7:20 -at the same time.

7:21 Which is, of course, the story of computing.

7:23 Right.

7:23 We also, with the technology we're developing, we share that with other people.

7:29 And so they cost them nothing.

7:31 So here we share that R&D, which is huge.

7:35 Look at this.

7:36 Look at this area.

7:37 Look at how much area is covered by this one camera.

7:41 Every day I look at all cameras and then

7:43 I look at cameras that have been recently moved.

7:46 Active cameras.

7:47 Because that tells me somebody that's been trained by us- -Right.

7:50 -onboarded, deliberately moved a camera.

7:53 So, there had to be some reason, And so you look at first and you go,

7:57 “Huh, that doesn't look good.” And then they, you see them triangulate.

8:02 And so here the public can see that.

8:04 You can go on and say,

8:06 “Click the button that says ‘active cameras’ and see where,

8:10 where firefighters are looking.

8:13 That's such a benefit.

8:14 And we have layers with maps from ESRI that-all the street maps are in there.

8:19 So you can look at where your house is.

8:21 You want to know where your house is,

8:23 and then you can see the perimeter of the fire.

8:25 And you can say, “You know,

8:27 I don't need a firefighter or a sheriff to knock on my door.

8:30 I'm right in front of that fire that's coming at me, and it's moving north,

8:34 and I live right here.” It's a benefit of having

8:37 an open system and having the imagery available to the public.

8:41 And we connect with other technology

8:43 providers that provide information, including CAL FIRE, that hosts the imagery,

8:48 through our system on their site when they're posting official information.

8:53 When you're dealing with emergencies to this magnitude,

8:56 you don't have the time to dust off username and passcode.

9:00 You just need to be able to see it to make decisions very quickly,

9:04 that- for your department, for your community, for yourself.

9:08 And that's where being an open system has greatly benefited us.

9:11 You have to use it because in an event when fire’s at your back door,

9:16 trust me, it's a— you're so anxious and you're not making good decisions.

9:29 When you think about the decade over which this work has been going forward,

9:33 I mean, a lot has changed.

9:35 Cameras have gotten sharper, connectivity has gotten faster,

9:38 batteries have gotten smaller, computers have improved.

9:43 What has been the impact in the last year or two of AI?

9:47 I look at the metric that the AI is providing,

9:52 reducing noise that allows the dispatcher to have

9:55 situational awareness and, and observe how things are changing.

9:59 And it's telling me where I have change,

10:02 and now I want to have it tell me where I have change faster,

10:06 with better clarity.

10:07 And what Zach has taught me- he says, “Neal,

10:10 when I can look at a fire in almost real-time,

10:14 like video, I gain so much knowledge in that first four to five minutes.

10:20 It's critical.

10:21 I can tell so much from that little time slice,

10:25 especially if I get frame rates higher.” We're never going

10:28 to take the subject matter expert out of the loop.

10:31 So, when people ask me, “Well,

10:32 is this going to take jobs?” No, it's going to create hybrid jobs.

10:36 AI is not meant to replace firefighters.

10:40 It's meant to enhance firefighters, to allow them to do their job.

10:43 But through AI, being able to see the signal

10:47 of an incipient fire and be able to respond,

10:50 that's something that I can do something about.

10:52 We can contain and confine those fires when we know about them.

10:55 What AI does is reduces the potential for large,

10:59 complex fires and allows us to mitigate them at the smallest level,

11:04 which means my firefighters get to return home

11:07 to the areas that they respond to, medical aids,

11:10 and traffic accidents, and structure fires.

11:13 And those resources are then prepared for the next fire.

11:17 AI has a way to help keep us safer and notify us so that we can do our job.

11:22 And Juan, as the person who's brought the AI technology to all of this.

11:26 It's sort of a story of the firefighter, the professor, and the data scientist.

11:33 From the data science side,

11:34 what's the key to working effectively in this kind of team?

11:38 For us, what is critical is the fact that we need to get very good quality data.

11:42 So, these cameras not only provide an amazing data,

11:45 but when you have great people understand what

11:47 is a fire or what is not the fire, to label the data and indicate to the AI

11:52 models like that's- that's the key for our success.

11:56 Firefighters labeling data.

11:57 That's critical for us.

11:59 So, the next part is the fact that you have these these models

12:04 to have the ability to control the cameras and actually zoom in.

12:07 Because that will provide much better, like, much better quality data.

12:11 That's another thing that is key for us.

12:13 So, I've had one dream fulfilled with this project.

12:16 What's that dream?

12:18 That dream is that I wanted to have, I could drop a pin on the map,

12:24 and I could see every camera that could see that pin illuminated,

12:28 and I can go look at it.

12:31 That's what I wanted.

12:32 That was the dream.

12:33 So I can know what cameras have an unobstructed view of the fire.

12:38 Okay.

12:38 And that was hard to do.

12:40 Okay.

12:40 But now with the data quality and, and what Juan's talking about is how

12:45 do we take these data and get them so they go into the model.

12:52 All of a sudden we're reaching what we believe is

12:55 a better targeting on the fire than we had before.

12:59 And I think that we're in a total agreement that we have to have awesome data,

13:05 and we have to have enough of it that we can, you know, populate the model.

13:09 This is very promising.

13:11 The next step is if we are in places where we have multiple cameras,

13:14 is that the model will detect the camera.

13:16 You have two- as long as you have two cameras that can look at the same fire,

13:20 you actually know the location of a fire,

13:22 because then the model will basically indicate the location of a fire,

13:25 that's critical for you.

13:26 Yeah.

13:26 Well, think about the decade that you all have been on this journey.

13:31 You started ten years ago.

13:33 Think about eight years- how far you've come.

13:36 Think about 2035.

13:38 What would you hope would be the state of the art in 2035,

13:43 not just in terms of what technology can do,

13:46 but what technology actually is doing, not just here in California,

13:49 but everywhere where these wildfires are such a hazard?

13:52 Right.

13:53 My dream is that the camera, after it sees what's a potential fire,

13:59 it moves all the cameras around it automatically,

14:01 points it at it, and they zoom in automatically.

14:06 So, I don't have to do all that.

14:07 All that work is done because that's kind of tedious.

14:10 It takes time.

14:11 And then the firefighter just can look and say, “Yes, that-this is an issue.

14:16 Is it a small issue?

14:18 Is it a big issue?

14:19 Is it a controlled burn?

14:20 Do I have to address this?” All

14:22 these things are running through the dispatcher’s mind.

14:25 And if we can give them high-quality data and enhanced,

14:29 you know, situational awareness, the dispatcher’s going to be able to make

14:34 a data-driven decision that is more complete.

14:37 Every fire is different, and every fire has a chance to get out of the box.

14:42 And that's what we're trying to solve is: let's not let the fire out of the box.

14:47 And the good thing here is also because

14:48 you have so many cameras and you have, unfortunately,

14:51 given the amount of fires that you have every year,

14:53 you have every fire, you can learn from the fire.

14:56 Yeah.

14:57 Right now the focus has been on detection.

14:59 The next, I think the next also phase is prediction,

15:01 it’s what's going to happen with that fire.

15:03 Correct?

15:03 Unless something that- yeah like yes every fire is different,

15:05 but if you have enough of these fires,

15:08 you can train models to see what's going to happen in the next hour,

15:11 in the next- And so we give our data

15:13 to people that model fires and fire behavior.

15:16 And they use MesoWest, the wind data, the weather data.

15:19 So they can say that we start- we see a fire,

15:22 we confirm it by firefighters saying, “Yes,

15:25 that's a fire.” The dispatcher then dispatches according to that.

15:29 While that's all happening we're having a model,

15:32 Technosylva and WIFIRE, they're taking the pin we gave them,

15:36 they've got the weather conditions, they've got the topography,

15:39 they got the fuel loads, they got the fuel health, and they make models.

15:43 Where's the fire going to be in five hours?

15:46 Ten hours?

15:47 And then we can see if that's correct, and we can iterate on that.

15:50 And so we're learning about how topography controls where fires go.

15:54 Yeah.

15:55 Wow.

15:56 And what would you say, Zach?

15:58 Ten years from now, what would you hope would be the state-of-the-art?

16:00 In the fire service we have a saying, “Everybody goes home.” You know,

16:05 and I talked to you guys about the priority of life safety.

16:08 So in 2035, I'd love to see that this system is driving

16:12 down the loss of life and driving down the injuries to firefighters.

16:17 I would love to see the technology generated here

16:20 expand to other areas so that other people can benefit.

16:24 Because we have areas that don't have the economic means to, you know,

16:29 develop the technology, but through a sharing model,

16:33 and training them to do what we've done in partnership with Microsoft,

16:37 we see the ability to expand this.

16:40 That only drives the economy of scale up,

16:42 which drives the cost of operation down,

16:44 which makes it easier for my firefighters in California to use.

16:48 We can do this anywhere.

16:49 That is what is amazing to me.

16:51 Because you all have perfected the technology, brought down the cost,

16:55 and made it something that is affordable to scale,- Right.

16:58 Right.

16:59 and so, you know, now you go to like the, and you know,

17:02 the Greek government, the Italian government,

17:04 the Canadian government, the US government.

17:07 I mean, the the level of investment

17:10 compared to the costs that are being incurred.

17:13 -Yeah.

17:13 It's huge.

17:14 Well, it's an inspiration.

17:16 I mean, it's, it's first of all extraordinary to see

17:19 the inspiration that you had that led to this work

17:22 and this partnership between the two of you to see

17:25 the impact it's having and to see the future.

17:27 Speaking freely, working with people like Juan,

17:30 and the AI for Good Lab has been huge.

17:33 We couldn't be happier with the partnership.

17:35 And, you know, partnerships need to be win-win, and so that we can work together

17:40 to accomplish something bigger than us individually.

17:43 Well, thank you for the privilege,

17:46 because it really has been a privilege that you've

17:48 given to us at Microsoft and to Sai and Juan.

17:51 And, on behalf of others I know who are part of this team at Microsoft,

17:55 it's easy to see why everybody is so enthusiastic about

17:59 just the opportunity to work with the two of you.

18:01 So, thank you.

18:01 The honor’s ours.

18:03 Thanks.

18:03 I'm going to get misty.