What's Your Signal?

Why AI for Public Safety? Voice Analytics and Much More

• Cradlepoint • Season 1 • Episode 32

Share episode

For public safety agencies, the potential impact of AI is still unfolding. But already agencies are finding eye-opening opportunities such as AI-driven voice analytics. We talked to Kinu Masaki, CEO of Voicebrain, and Jonathan Fischer of Ericsson about ways that AI can help make communities safer and emergency response faster.

www.cradlepoint.com 
https://voicebrain.ai/



SPEAKER_01:

Welcome to What's Your Signal, the podcast exploring enterprise networking trends and the secure connectivity that underlies it all. You're joined by me, Taylor Walker, and my co-host, Mauricio Stephan. How's it going, Mauricio?

SPEAKER_04:

Good. How's it hanging, Taylor?

SPEAKER_01:

It's hanging well. Thank you. So, Mauricio, today we're revisiting a topic that we've covered on the podcast before, but it's got a little twist this time, and I'll give you a hint. Do you remember the movie Robocop?

SPEAKER_04:

Of course, right? Half human, half machine, fighting crime, insane accuracy. I hope you're referring to the classic version and not the reboot with what's-his-name. I don't even remember. It was not that great.

SPEAKER_01:

Definitely the classic, the classic one. And that one felt so futuristic at the time, like something that could never happen. And yet here we are today with public safety agencies that are testing AI applications to improve policing and emergency response efforts.

SPEAKER_04:

Yeah, except for crime-fighting cyborgs, which is a little bit of a disappointment. We're seeing things like voice analytics, data analysis, and even some AI-assisted fire responses.

SPEAKER_01:

That's right. So today we're joined by two guests, Kanu Masaki, the founder and CEO of VoiceBrain, and Jonathan Fisher, the head of global OEM and embedded partners at Ericsson. And these folks have a real pulse on not just AI, but also what it means for public safety agencies. So let's get into it. All right, so Canoe, welcome. And Jonathan, welcome back. We're looking forward to chatting to you guys about AI for public safety. But first, though, especially Canoe, since you're new to the podcast, we'd love to hear a bit about your journey and kind of how you ended up working in the world of AI.

SPEAKER_00:

Well, thank you so much for having me. I'm honored to be So I'm Kinu Masaki. So my background is I'm an MIT lifer. In other words, I did my undergrad, master's, and PhD at MIT and studied AI as an undergrad. But during the AI winters, as they call it, I kind of switched more into biomedical engineering, specifically in the field of hearing. But you know, in the last 10 years, there's been such a huge advancement in AI. And I finally realized, you know what? AI is now ready for the general public. And because my passion has always been the intersection of AI and hearing, The sector I really wanted to help were people who use voice as a main form of communication. And specifically the people I cared about were, you know, people in public safety who uses radios on a day-to-day basis, but they don't have AI as a tool. And that's a group that I wanted to help.

SPEAKER_01:

That's awesome. A very impressive background. And I'd love to hear a little bit more about how your company is kind of leveraging AI for that group.

SPEAKER_00:

Yep. So what VoiceBrain does is we, first of all, capture any critical voice communications, such as radio communication from the police, sheriff, fire, or the TSA. And then we transcribe it, and then we analyze it in real time, and then notify critical key stakeholders when key issues occur that they care about immediately. And then we have this multimodal capability so that we allow people who are on radios and people who are on cell phones to seamlessly communicate with each other on our platform.

SPEAKER_01:

Can do the applications that you mentioned are really kind of focused on these like auditory and then like filtering through applications. text and some of the vocal pieces. Jonathan, what have you seen other applications in public safety for AI? And I know you've got like that background in public safety as well. And so can you tell us about just some of those additional AI applications in the public safety industry?

SPEAKER_03:

Yeah, thanks. Thanks for having me back, Taylor and Mauricio. So yeah, we are seeing a lot in the space around computer vision and public safety. So computer vision is basically taking cameras and making them intelligent, giving you kind of a real time situational awareness. And we're really looking at doing that at the far edge of 5G, if you will, giving officers the advantage of understanding what's truly happening, perhaps before they arrive on the scene, leveraging cameras, maybe at a location, maybe it's a bank robbery in progress, right? That officer could see exactly what's going on, what kind of weapons might be involved, how many people might be involved in the particular incident. Or it could be a fire department responding to an accident, being able to get into a local camera in an intersection and say, okay, it looks like I have three vehicles. It looks like I may have a couple of injuries. I need two ambulances. I need a rescue truck. I need a pumper because it looks like there's some fluids on the ground or maybe a fire. All those different things that you can gain from that. And it's all being done by a computer and it's telling me. I've been using this kind of concept of a co-pilot. So if I'm a firefighter responding to a scene, it may be talking to me and telling me, hey, firefighter Fisher, on the weight of the scene, you have this entrapment. You have four people in a vehicle, you have fluids on the ground. It's being able to tell you that real time based on what it's seeing and intelligence based on that. The other things we're starting to see is really a smarter dispatch and response, leveraging technologies like voice brain. You'll be able to start doing some triage of 911 calls. So if you have large-scale incidents where you need to be able to understand truly what's happening, AI can start to piece together the scene, can really piece together truly what's happening, be able to deduce that down more than the human mind can and do it much quicker than then you and I could do it. Classify the urgency, send the right agencies out there. I remember from my firefighter days, just even like a big ice storm when I lived up north. And we had to have, there's a car crash here, there's a telephone pole down here, there's a tree down on a house. Where do you go first, right? So this really kind of helps you triage and get to the emergencies and really all about saving lives and property.

SPEAKER_00:

And I always believed video plus audio is one plus one equals three, right? Because, you know, vision AI has always been one step above or faster, I think, than voice AI. But I think now is such a perfect time where both of these two commodities can now work together to really help the officers, the people in the field, really, as Jonathan said, as a co-pilot, to help officers, you know, really help them do their job much more efficiently and be much more safer. And so I think this is such an exciting time in the field of public safety.

SPEAKER_01:

Absolutely. This might be hopefully not putting you on the spot too much to recall, like specifics, but have you heard any anecdotal or like specific stats around like, how much time has been saved or just like some of those really specific benefits of integrating applications like voice brain into

SPEAKER_00:

public safety? Yeah. So one great example is we have a great agreement with the Department of Homeland Security. And so through that, we've been working with the TSA and the TSA officer said, now we actually get to events at least 15 minutes earlier because of voice brain. Right. Because sometimes the officers are very busy. Obviously, there's a lot of different inputs. They don't always some of the, you know, the head officers don't don't always have radios with them. However, when key situations occur, in less than a second, voice brain is able to alert the key officers so that they could all, number one, communicate seamlessly, no matter what platform they are, and then they can act immediately, right? And so they truly said, you know what? We truly believe voice brain will one day save lives just because sometimes in these critical situations, a difference of 5, 10, 15 minutes could be the difference between life and death.

SPEAKER_01:

Jonathan, how does something like an Ericsson Cradlepoint router integrate with something like a voice brain application? What are the different... roles there

SPEAKER_03:

yeah so so we we didn't have this functionality when we first started uh down the path with uh voice brain um but we did have obviously our edge compute on the r1900 um and that was kind of the basis of how we started the discussion um and then it was said hey we need a an audio um uh the ability for the audio to um traverse into the compute space on the device and get that audio back to the voice brain cloud. So actually, our engineering team developed an interface using our USB connection to an audio driver running on the router. And then that audio then goes into the container where that container is able to pick up that audio and then send it back to the VoiceBrain Cloud. Then we're also leveraging our NetCloud Exchange and specifically for security. Obviously, these communications are extremely important and need to be secured end-to-end. So we're actually leveraging our NCX capabilities for that as well. And so it comes from a Lambda radio or a microphone into the R1900 through our Docker container environment on the VoiceBrain application. back to the voice brain cloud where that information is processed. And it all happens quickly and seamlessly because it's gotta be low latency. It's gotta be super, super quick because information and data needs to get back to those officers and that information needs to be acted upon as quickly as possible.

SPEAKER_00:

And I think, you know, having AI at the edge is truly game-changing and critical for these use cases of public safety, you know, homeland security, because, you know, these officers and people are putting their life at risk, right? And so information, you know, having information immediately without any delays is critical. And also keeping, you know, secret information conversations contained and not out in the cloud is also very important. And so, you know, this opportunity to have, you know, Ericsson CreativePoint integrated into VoiceBrain, I think is truly game-changing.

SPEAKER_01:

Yeah, you know, your comment about, you know, keeping information contained, Mauricio, I think you wanted to ask a little bit about kind of that security piece, right?

SPEAKER_04:

Obviously, right, because I think beyond all the things that AI can do for us, one of the things that's in the forefront of the public's consciousness when it comes to AI is just the amount of data that it takes to train a model, right? And then also the data that's being ingested. So, you know, Obviously, especially when it comes to public sector, both federal, state, local, there's a lot of concerns about privacy and safety. So just wanted to ask a question about how you guys are addressing that in these use cases.

SPEAKER_00:

Well, because we are working with these big companies, Homeland Security agencies and, you know, public safety, obviously cybersecurity is very important to us, right? And, you know, whenever we do work with these agencies, we get vetted tremendously by these agencies, right? And so, you know. You're like, they're at my house right now.

UNKNOWN:

Exactly.

SPEAKER_00:

I got fingerprinted. My background check has all been done. But yeah, so, you know, definitely, you know, security is key to, you know, voice brains. core thesis.

SPEAKER_03:

And Kino, your model, your pre-trained model and neural networks are based upon public safety, right? So, you know, ChatGPT is trained for other reasons. This has been very specifically trained for the needs of public safety and it knows what to listen for and it knows based on its pre-training and long learning lessons that these are the things to look for and to be mindful of. You know, shots fired, officer down. You know, those kind of things are, it's used to listening to that and then taking actions.

SPEAKER_00:

That's completely correct, Jonathan. Thank you for, yes, for bringing that out because that's very important, right? You think, oh, it's AI. Just use, you know, OpenAI's general model and it should work. No, that's not the case, right? For public safety, you know, it's very specific.

UNKNOWN:

We

SPEAKER_00:

specifically trained on radio communication on these different sectors and these different use cases so that we can, number one, transcribe correctly and then we can figure out what are the key situations that these officers care about.

SPEAKER_01:

I'm curious, you've both touched on, you know, there's obviously the need to have the right equipment in place, the right software, the right AI models that have been trained. How How much is AI actually being adopted in public safety? Is there still a lot of hesitation around it? Are there still barriers to entry? What are your thoughts

SPEAKER_00:

on that? No, that's completely correct. It was so funny. I gave a presentation to a big radio conference. And the first question I asked was, how many of you guys are scared of AI? 99% of people raise brain, right? And so, you know, when you think about AI, the first things they worry about is, oh my gosh, is my job going to be taken? Can we trust it? Is it going to take over the world and kill us, right? So, you know, those are some big questions that people care about. But, you know, really our goal at VoiceBrain is really to provide for AI to be a tool to help them, you know, do their job more efficiently and safer. Right. And so, you know, step one is number one, we just capture and upload into cloud and make this database and transcribe it. You know, that's not going to take over your job. Right. And we provide alerts. But we also whenever we do a transcription, send alert. We not only send the transcription, but we also send the actual voice clip. Right, because we know we're not 100% correct, right? No transcription is 100% correct. So we allow the person to choose, okay, is the transcription correct? Okay, because I now believe in this transcription, these are the actions I'm gonna take, right? So we're just giving the tools to the person and not taking over their job. It is true as people trust AI more and more They'll let us do a little bit more. For example, one of the biggest, newest things that people are talking about right now are the AI agents. These are AI tools specifically to do different tasks. So for example, one of the agents that we have is this report writing. In other words, if you have an incident, you can have an agent just, hey, can you quickly just... summarize the incident that happened yesterday with the shooting right but again we're not going to be 100 correct right they could look back and make sure hey this is correct this is not correct then they could fix it right and so you know it's it's baby steps right you start believing and trusting ai gradually and then you know you allow ai to do more and more things But again, at the end of the day, it's the officers, it's the people who make the final decisions of what's written, what's done and the actions taken.

SPEAKER_01:

Absolutely. Jonathan, I'm curious about You know, your input on that as well for applications beyond transcription and beyond maybe some of those AI agents, especially if public safety agencies are using AI tools in real time, whether it's, you know, for video or like some of the fire applications you were talking about, are there some barriers to entry there? Is there hesitation to include AI in some of those other use cases?

SPEAKER_03:

So yeah, the body-worn camera was really the first piece of technology other than maybe your two-way radio that an officer carried in. officers were very resistant to taking on those body-worn cameras. Only after a few weeks of carrying it, they realized that it was really about protection of the officer than it was looking at what was really going around or really this big brother thing. So definitely became that next kid. I think AI will have that same little barrier of adoption. And today, they're already recording every transmission, you have your recordings into 911, you have your recordings of all your radio transmissions today, but it's making it intelligent and making it be able to kick off an API that says, hey, send another police car because he's in trouble. He just said signal 13 and signal 13 means everybody within the range of my voice come save me because I'm in a bad place. But perhaps the dispatcher doesn't hear that or the other officers in the area don't hear that, but the AI can and it'll be listening continually for that. The other place I see it being used, this kind of AI for good, enhanced transparency, accountability of the officer, accountability of the people and the citizens that they're working with, automatically flagging use of force in incidents for noncompliance, looking for how did that happen. It's not just video, but also the audio components of that. And then really about cross-agency intelligence. So something that may have happened here in Florida Florida, maybe happening somewhere else in the United States, and we can start to build really a global database of activities, right? So when we're talking about that pre-trained model earlier that VoiceBrands created, you know, that intelligence that, you know, if it's something seen here once and it's happened somewhere else, you can start to become intelligently aware of what might happen next, right? So a car hits a pole, right? Does that pole have wires on it? Does it have a big transformer on the top of it? What's going to happen next? So, you know, like, okay, now call the power company because the power company needs to come and cut the wires down before the officers can safely approach the vehicle. All these different things that you can now kick off, you know, that you now know what happens next. And so those AI models can really help really intelligently direct maybe a dispatcher or automatically call, you know, Florida Power and Light and say, hey, there's an accident on 95. Go out there and help remove the wires and shut down the power. Or even better, go to the power grid and cut off just that very small section of power to ensure the safety of everybody responding as well as the folks that are involved in the incident.

SPEAKER_04:

So you said that you guys are focusing on public safety use cases. Are you guys looking at all to bridging to outside of public safety? Or if not, are there like use cases that you would see that similar technologies could be used outside of public safety?

SPEAKER_00:

Yeah. So, yeah, we do not only public safety, but, you know, we work with enterprises in general and then Homeland Security. But, you know, it really is... What's great is voice communication, radio communication is a... glue of society in some sense. Before I thought, oh, wait, people still even use radios. But when you actually open your eyes and realize, oh my gosh, aviation, FAA, they all use radio communications, right? Public safety, Department of Homeland Security, all of these different sectors, also amusement parks, also stadiums, right? All of this is about safety of the people and how do we, you know, correctly react to situations immediately in a smart way, in an intelligent way with correct people, decision makers being part of that, right? And I think that's where voice brain comes in. It really allows for interoperability and accountability and true situational awareness.

SPEAKER_03:

The really cool thing enterprise solution that I think you guys told me about is the one you did with TSA. That's a really interesting one where you can, you know, somebody says something on the radio and you know that that plane might be delayed or you need to do something. Maybe talk about that one a little bit more, Kenya.

SPEAKER_00:

So I think this situation was, you know, a officer that a head of aviation at an airport was headed home and he got a notification on his phone from voice brain saying, hey, there's a plane in distress, right? Using VoiceBrain, he was able to listen in on that radio communication from his cell phone, right? Because his radio wouldn't have worked. And then was able to realize, oh my goodness, my team is in trouble. They're not dealing with it correctly. So he was able to text back on the VoiceBrain platform and talk to the people on site at the airport and say, hey, I'm coming back. Don't do anything. Here are the things you should do. and then was able to turn around, go back to the airport and deal with it immediately. And he said that was game changing, right? Because typically he would have found out about it the next day, maybe the next week, right? And, you know, the effect of not reacting to it immediately would have been very different, right? And so, you know, again, being able to truly have seamless communication amongst all the people, not only within the agency, but in some situations, you need inter-agencies, right? You need the TSA involved, you need the airport involved, you need the sheriff involved. And if voice brain is part of all of that, we allow all these different agencies who use different radios, who are are on different channels to seamlessly communicate with each other. So it really is game changing.

SPEAKER_01:

Can you, how do you envision some of those applications evolving over the next several years? Like what, I don't know how much you're allowed to say, like what's on the horizon for voice brain, but what do you see

SPEAKER_00:

coming? You know, what's so exciting to be being in the AI field is it's like, Changing an exponential rate. I mean, it's probably faster than exponential, right? New AI tools are made on a day-to-day basis. And, you know, once voice brain is able to take your radio communication and upload into the cloud, now we could take advantage of all these amazing tools that exist out there. Right. So, for example, we just kind of previously mentioned these idea of agents. Right. Jonathan mentioned this ability to maybe predict what's going to happen if if we listening into the radio. I think there's there might be a riot coming soon. Right. Or there's, you know, oh, my goodness, I think somebody's trying to shoot the president. Right. Or, you know, so that's predictive analysis. Right. The ability to now make agents to do these different tasks. Hey, maybe, you we could figure out where to place the officers in efficient way so that you're able to deal with key situations faster, right? Maybe it's a certain part of the city has issues all the time versus others. Maybe let's add more officers there, right? So AI can do so many things and I can't even imagine all the new tools people are gonna come up with with AI. But what's great is once we're able to capture all this radio communication, we can leverage any and all of these tools specifically for these public safety sectors. It sounds like the more

SPEAKER_01:

you use it, the better it can get.

SPEAKER_00:

A thousand percent,

SPEAKER_01:

yes. Which is pretty cool. Jonathan, how about you? Any thoughts on kind of where some of these AI applications for public safety are headed in the next several years?

SPEAKER_03:

Yeah, I think it's really the combination of all these different data points coming together, right? So you have voice, you have video, you have sensor data, right? Being able to pull all of that into a centralized location and have the AI be able to tell you exactly what's happening, right? So I was a hazmat officer for almost 20 years as a firefighter. if I could know exactly what methyl ethyl death we were rolling up onto at a, you know, at a semi truck rollover, and I would know that that is this particular chemical and I need to wear this type of protective equipment and I need to get, you know, Five miles worth of homes evacuated outside of the area. I can know what direction the wind is coming from. I can know when the next precipitation is coming. I can know there's a creek nearby. All of that information takes time to gather. Now, if I had all of that coming in at once and I can see that as I'm responding, I can make sure that citizens are safe, my team is safe, that we're responding in the most effective and efficient way, that we're containing that chemical in a way that's appropriate to that chemical. And ultimately, we save the environment, we save our people, we save our team, we all go home safely. So I really, that's where I think this is going. I think it's really about the culmination of so many data points, right? Those data points are oftentimes have to be connected via a cellular device because they're all over the place, right? They could be on a highway, on a smart sign up above the lanes. It could be a camera system where there needs to be connectivity to that camera system because it's out on a pole somewhere remotely. It's all those data points coming together to truly paint the picture and doing it in a way that if I had to do it manually and look up in a book, I remember we'd use these ERGs, emergency response guides, look through that guide. And then go to the NIOSH manual to know that this is what what I need to put on for protective equipment. And, you know, this is the plume direction. I need to know that I have to evacuate people this this direction for so many miles. And this is automatically calculating that real time. I mean, that takes hours for us to actually do that. And before we can even go out to start containing that chemical in some way. So, you know, I'm excited about it. And there's a lot of we think about the police side of things. But I think safety, rescue, and just general public safety is a real area for advancement.

SPEAKER_01:

You know, I think the only downside for me of having these conversations is my partner's a firefighter and then I go home and I'm like, are you guys using this? Why aren't you using this? Why hasn't our city adopted this? I need you to be safer. But on that note, we have taken up a lot of your guys' time today and I really appreciate you taking the time out to join us to kind of dive into the implications and the applications of AI for public safety and beyond. So thank you so much for having this conversation with us.

SPEAKER_02:

Thank

SPEAKER_01:

you for having us. I like this one. I like talking to Canoe and Jonathan because I think the combination of Jonathan's in-depth experience in public safety and Canoe's super in-depth experience in AI just come together really seamlessly. And I was glad that they addressed... One of my biggest concerns going into this conversation was like, are we talking about something that maybe is not really being used yet? Are people hesitant to do it? And when Canoe was... you know, kind of going through the transcription through voice brain and how that's a benefit. It started to make me think like, oh yeah, I use AI like that, you know, just to transcribe a meeting and then have it spit out like a recap for me or spit out like an outline for me. And it's a very similar thing. type of application. And it's helped give me some, I guess, more confidence in AI and makes me more likely to use it for other things, which I think is probably the case for the public safety audience as well.

SPEAKER_04:

Yeah, absolutely. I think also one of the things that both her and Jonathan addressed was that people tend to have this pessimistic view of AI, right? And it's typically based off of movies, right? Robocop, Terminator, you know, where AI is run amok or AI is as, you know, taking everyone's jobs and AI has done all these things, which, you know, I'm not saying that those things aren't a possibility because you never know. But what I am saying, though, is like right now people look at AI that way because But what they don't realize right now is that AI is just a tool that is available for everyone to better, you know, do your job or better do other activities, right? It's like a prime example that you just gave, right? It's like, I don't know about... you or our listeners but you know a lot of times i'm double and triple booked for meetings and can't attend all the meetings that i need to attend to and so ai has been a great way to get that summary of meetings that i couldn't attend and and not necessarily not that anybody does on purpose but you know if you rely on someone else's notes they usually have their own spin to them right they have their own their own what they heard what they understood and that comes through in their notes where ai is just really doesn't have that slant on the points that it's giving you. That's a real good thing. Then the other thing I think this is a great example, especially for public safety, but also for just enterprise in general is like, relying on voice communications does have the ability for the listener to miss things, right? Whether intentional or unintentional, you know, just as an example, you know, I was in a meeting with someone and this just happened where they missed the cue that it was, you know, their turn to talk. And so, you know, voice brains could spit out that alert that, hey, you know, you're you're you missed this important thing or and in the case of public safety right is like they use codes or you know um you know someone who's in trouble or an emergency can maybe only quickly say something and they can't repeat it and the person on the other end misses it and so you know that could cost seconds or minutes and so just having that ability that ai doesn't need a bathroom break. It doesn't need to go get water. It doesn't get distracted by their phone. It doesn't get distracted by the dog barking or the guy in the background mowing the lawn. Like it doesn't get distracted by those things. So it's always listening, always paying attention. And as it gets trained better, we'll start to always understand that the difference between something that's a real emergency and something that, okay, this can be summarized and thought about later. So, you know, all of these things are great in the context of public safety, but have broad applicability just across all enterprises and sectors.

SPEAKER_01:

I also liked the, you know, points that Jonathan was making about data aggregation and how that can help things. And even just like on a real time scale, you know, being able to, aggregate conversations that are happening across radios and that AI, whether now or down the road can maybe say like, there's a lot of stuff happening like in this part of the city, or with this particular type of emergency and it can, you know, maybe spit out some hypotheses of like, what could be going on so that public safety agencies can better prepare for something that's maybe more large scale than a bunch of small things. They can see that bigger picture more easily.

SPEAKER_04:

Yeah, 100%, right, is going back to kind of a US national tragedy, right, is the September 11th, the Twin Towers. One of, you know, there was a huge commission That that was like, OK, how could this have been? How could they not necessarily like, you know, understanding the terrorist attacks and how we can prevent terrorist attacks, but more understanding the response and how can we make the response better and how can we make that? And one of the key issues that was identified was interagency communication. One of the things that came out of that for anybody listening that's not aware is FirstNet. So seeing FirstNet was one of the key things that came out of that. But I think VoiceBrain can help address some of the other things is that being able to correlate different agency conversations about the same event to an incident commander who can have that overall situational awareness. And then on the flip side of that is being able to maybe prepare adjacent agencies that may have interagency agreements for something that they might have to respond to that the other agency hasn't asked them yet and so you know that age or when they do ask them they can share the voice brain information with them so they can quickly get up to speed without having to take the time of doing you know person-to-person briefings and those kind of things so there's i think there's a lot of potential there um just in things that we've seen in the past to to help address that ai can help address

SPEAKER_01:

I agree. Well, thank you for listening. If you've enjoyed this conversation about AI for public safety, please share it with a friend. Then rate and subscribe to What's Your Signal wherever you listen to podcasts.