The Future of Communication Between Humans and Dogs with Melody Jackson, Professor at Georgia Tech

(Intro Narrator at 00:00:03) Hello, my friends. Today, Joel is talking to Melody, professor at Georgia Tech, and they discuss how Melody is researching interaction between dogs and computers, sensor vests that enhance communication between humans and dogs, and how this technology can improve the lives of people with disabilities. All of this right here, right now, on the Modern CTO podcast.

(Joel Beasley at 00:00:28) This is the Modern CTO podcast.

(Melody at 00:00:36) So I—

(Joel Beasley at 00:00:37) was curious. Could you just give me a brief backstory of what you've done so far?

(Melody at 00:00:42) Wow. That's a big question, Joel. Early on, I was working in industry with a computer science degree, and I loved that. I was a developer. We did defense systems. I worked for several different companies and really enjoyed that. But I found that I was not getting the creative outlets that I needed. And so luckily, I got an offer from Georgia Tech to come in as a research scientist to work on a development project, not surprisingly, since that was my background. And when I got exposed to what it was like to be in academia, I just was so intrigued with the idea of complete intellectual freedom.

(Melody at 00:01:22) And so at that time, I had my master's degree, and I decided to work on my PhD and got my PhD. And as soon as I finished my PhD in computer science, I met a neuroscientist, a neurologist, actually, who was implanting people with chronic recording electrodes so that they could talk even though they were completely paralyzed and unable to speak. And I just was blown away by that whole idea. My grandmother was disabled. She had multiple sclerosis, and she lost the movement and everything but her left hand.

(Melody at 00:01:59) So I've always been very interested in assistive technology. As a result of that, I started raising and training service dogs for Canine Companions for Independence in 1994. So we've raised and trained five puppies for Canine Companions. And I've just always been really interested in how can we help people with disabilities. And that has turned into an incredibly satisfying and rewarding career. A lot of the time, I just can't believe they pay me to do these things because I just would do them anyway. Don't tell my employers that. But it was just, I guess, luck and a little bit of passion that I wanted to pursue this stream of work, and it's turned out to be amazing. Just amazing experience for me.

(Joel Beasley at 00:02:47) And where do you get your discipline from? Because in order to—a lot of people, you know, they'll be following their passion. I actually don't like the word passion recently because I just looked it up in the dictionary two weeks ago, and I'm like, I gotta stop using passion. Because people say find your passion, and if you look up the definition, by definition, it's a fleeting thing that goes away. And I'm like, you don't want to find your passion. Maybe your purpose. Right?

(Melody at 00:03:12) Purpose. That's a better word.

(Joel Beasley at 00:03:13) Yeah. But to have discipline, to stay focused, and go on this path, it takes work. And so did you get that from your parents or some big life changing event?

(Melody at 00:03:23) Well, I am a proud military brat. My dad was in the Air Force. He's retired full colonel, and that's probably a big part of it. I watched my mom and my dad both be very dedicated to their careers and to their family and to the philanthropies that we wanted to support. And so I think that was just a family tradition that, you know, if you're going to do something, you do it 100%. And so I was always very encouraged to get education, very encouraged to achieve. There were very high bars for us in our family. And, you know, we were expected to get straight A's and things like that because if you weren't getting straight A's, you weren't spending enough time on it. And part of that was a compliment to us because my parents said, you are capable of this, and we want you to do the best you possibly can. So that was drilled into me very early on.

(Melody at 00:04:25) And I actually absolutely agree with it. I've raised my own child that way. And she is now a veterinarian, very proud to say. So I think that whole way of thinking is just sort of part of my family.

(Joel Beasley at 00:04:40) I have no idea where she got the interest in dogs from.

(Melody at 00:04:43) I have no idea. Dogs and horses her whole life. Bless her heart.

(Joel Beasley at 00:04:48) Well, I live on a farm, so that's right up the alley.

(Melody at 00:04:50) Oh, I'm so jealous.

(Joel Beasley at 00:04:51) Yeah. My wife was a vet tech for fifteen years—

(Melody at 00:04:54) Oh, wow.

(Joel Beasley at 00:04:54) —right before she became full time mom. And their vet tech clinic was next door essentially to the Southeastern Guide Dogs. So they worked on a lot.

(Melody at 00:05:02) Wow. We have worked with them.

(Joel Beasley at 00:05:04) Oh, yeah. They're down in Florida. Yeah. So we're a dog family, and we do have something else in common. So my dad was in the Air Force, and that's right.

(Melody at 00:05:13) Yeah.

(Joel Beasley at 00:05:13) Him making me make my bed, I think, is one of the first—you know how they go. So you've been around that. Yep. There's a very specific way to make a bed.

(Melody at 00:05:26) I bet you and I make them the same way.

(Joel Beasley at 00:05:29) Yep. It's amazing how much—you know, a lot of people will ask me, okay, what's the secret to being a great technology leader? And it's really hard to express to people, your environment plays a huge factor. And so understanding how to—you know, self-awareness of where you're at in your environment and how to change it and alter it to get better results. It's not a quick thing. It's a long learned behavior thing.

(Melody at 00:05:51) Yes. And I have to—the other part of that answer is that I've been part of the Georgia Tech culture since 1986 in one way or another. So I did my master's degree in the late eighties, and then I came onto the faculty as a research scientist working on specific research projects. And then I did my PhD in the nineties. And the culture at Georgia Tech encourages collaboration. It encourages creative thinking. And without my colleagues at Georgia Tech, I don't think we'd be where we are right now. But it has been just an incredible ride to work with some of the most brilliant people in the world. It's very inspiring. So that is another thing.

(Melody at 00:06:32) I think the culture of the organization you're in makes a huge difference as well. And having that support and just being able to sit down with a colleague over lunch and just brainstorm stuff. And we've come up with some of our best proposals that way. Just very informal, see you in the hall. Hey. What about that idea? Oh, yeah. Let's pursue that. So that's been a big part of it as well.

(Joel Beasley at 00:06:58) And then so what type of—what are you doing with the dogs? Mind control devices? I don't know.

(Melody at 00:07:04) Well, we actually, in a nutshell, are allowing dogs to communicate. That's one of the things that we're doing. So I have to sort of start at the beginning. As I just told you, I have raised and trained service dogs for very many years, and so I'm very tuned into that problem space and the kinds of things that they do. You know, I also have very close friends who have service dogs because of disabilities that they have. So I hang out with those folks a lot, and I see what their lives are like. So I happen to be lucky enough to share a lab since 2006 with one of the pioneers of wearable computing, Dr. Thad Starner. He's a very close colleague, and we have done many, many things together. But it sort of shocks me that we shared a lab together for about six years before one of our other colleagues had the idea of, hey, what if we could put wearable computing on a dog?

(Melody at 00:08:02) Because I usually have a dog with me at some point, you know, because I'm raising and training them. Dogs in training have the same public access as a fully trained dog because we have to train them to be okay to be in any situation. So I would have a dog with me pretty much all the time, and it took us a while to come up with the idea that, hey, we could use technology for service dogs. And our colleague was thinking, wow, we could make, you know, pretty lights and make them safety and that kind of thing. And then I sort of had an epiphany. It's like, wow. These dogs could call 911 in an emergency if we put a cell modem and some sort of an affordance on their service dog vest that they're already wearing. We could put a button on there. We could put something that you could tug. And these dogs could do a number of things.

(Melody at 00:08:50) Anything you can do with a computer, these dogs could do. And so that is how the FIDO project started. That was our very first project. FIDO is an acronym that stands for Facilitating Interactions for Dogs with Occupations. So that is how we started thinking about what kinds of things—taking human computer interaction, which is something that I teach, take how do we take the theory and methods of human computer interaction and take it into the animal world? Because we have a perfectly valid user. Doesn't have hands, doesn't have opposable thumbs, but they have nose, they have paws, they have all kinds of things they could use to activate sensors. And so we started to do some usability testing like we do for humans, but with dogs. We started with dogs since they're easiest to get a hold of, and we learned some incredible things. We started out thinking about what are the natural things that dogs do to interact with their environment?

(Melody at 00:09:49) Well, they can bite things. They can hold things in their mouth and carry them around. They can bite on and tug something. These are all things that little puppies can do. They can touch things with their noses. They can touch things with their paws. So we looked at all of these things and said, okay, what's the most natural thing for these dogs to do to be able to activate a computer that's on their vest? So we came up with nine different sensors, nine different affordances. Affordances is just the word we use in human computer interaction. That means it's something that looks like what—how it's supposed to be used. Like, a button is for pushing, a chair is for sitting, a slider is for sliding across.

(Melody at 00:10:27) So what does that mean for dogs? So we had a wonderful time with this paper, this study that we did. And we discovered that dogs can indeed turn around and bite something that's on their vest. They can turn around and tug something that's on their vest. It's really hard for them to use their paws on something that's on the vest. So we kept it to a nose and snout interaction. But the dogs actually started to invent their own ways of interacting with it. So one of the things that we did was a proximity sensor on the vest, which is just like an automatic faucet. You wave your hand over it and it turns the water on.

(Melody at 00:11:11) Well, the dog would wave their nose over it and it would activate the computer to do something, either send a text message to somebody or make a phone call to 911 or say something. We put a little speaker on there, and it would say, excuse me. My owner needs help. And it turned out that with this proximity sensor, the dogs—we thought they would touch it with their nose because that's a very natural thing that dogs do, is touch things with their noses. And the dogs ended up inventing gestures to activate the sensor.

(Melody at 00:11:39) So they would wave their nose down or they could wave their nose up, which made us think, oh, we could have two different inter—two different messages here because they can interact with us in different ways. One dog did it with his ear. He just bent his head over and activated the sensor with his ear. And so we found out that our user is very engaged and has some opinions about how these things need to be used. So it was a fascinating study.

(Melody at 00:12:04) We ended up sending the paper to the flagship conference, International Symposium on Wearable Computing, and we won best paper that year. So that was gratifying because it really made me feel like the community of wearable technology, in particular—a lot of human computer interaction people in there—were taking us seriously, that this was something that was not just a whim and not just a cute little project with dogs, that this is a real field. And we started reaching out to colleagues, mostly in Europe. Actually, there's some folks in Europe that have, sort of about the same time frame, started working on these kinds of problems. And we now have an animal computer interaction conference. We have a whole community that is working in this field. And essentially, we helped create this new field. And that kind of thing just—I can't tell you how satisfying it is to see that we discovered a whole new problem space, a whole new set of things that need theory and methods and, you know, how do we do this best for dogs? And, of course, you know, I also have five horses. So we started looking at what kinds of sensors can we use for horses to wear for things like detecting lameness, which is kind of a black art.

(Melody at 00:13:20) Veterinarians are amazing at it, but it's very difficult to detect lameness in a horse and to really pinpoint what it is. So we developed some systems that would measure exactly the way the limbs were moving and things like that, that would help us figure out what's going on with the horse. And now there are actually commercial systems out there that you can buy to do that kind of thing. So it's been an incredible ride. And I'm very proud to say that we are pioneers in this area and that it is a serious research area now.

(Joel Beasley at 00:13:52) Have you written any books about it?

(Melody at 00:13:56) Well, I can't really divulge too much now, but there is a book in the works.

(Joel Beasley at 00:14:01) That's good. Yeah. When you see industries emerge like that and the people that are a part of their emergence, typically, they're the first ones to write books, and then that helps get more people because then we can refer back. I could say, hey. If you want to know more about, you know, what Melody's doing, go check out this book. So—

(Melody at 00:14:15) So—

(Joel Beasley at 00:14:15) —whatever we need to do to dance around this, let me know, and we can come back, and we can have you back on, release the book, or put something on our show.

(Melody at 00:14:23) That'd be great.

(Joel Beasley at 00:14:23) Yeah. Super exciting. Now are you doing anything with the brain? I talked to this guy a couple weeks ago, and for humans, they did—you know, most of the BCIs, they drill a hole in the skull. Right? And that usually loses most people right off the bat. It's like, I don't want to hold my skull if I don't have to. But then they found this electrode stent, and they can put it up to the jugular and release it into the veins of the brain, and then you can control with this link computer, and they built their own operating system for it. So are you guys implanting anything in the dogs today?

(Melody at 00:14:56) No. Actually, all of my brain work—and I am the director of the Brain Lab at Georgia Tech, and actually, Dr. Starner is also involved in that very heavily—all of our brain work has been with humans. And so I did start out on the very first team to ever implant a human being back in 1997 with Dr. Phil Kennedy of Neural Signals Incorporated. And he actually—there's a documentary out right now called Father of the Cyborgs, which is about Dr. Kennedy. And I'm actually in it several places because I was his computer scientist, and he was the neurologist working on that. So that's out now. You can actually stream it, I believe. But we started out with his neurotrophic electrode, which he implants in the brain and then the axons grow through it, which allows the brain to kind of hold it there.

(Melody at 00:15:47) And then with imagined motion, so that we would implant it in the motor cortex after an MRI to determine where this brain might still be active because the folks we were working with were late stage ALS and brain stem stroke and folks like that who are completely paralyzed and unable to speak. And so the idea here is let's give them a channel of communicating with a computer or a wheelchair or the TV that doesn't require muscle movement. So it's only based on brain signals. So this was Dr. Kennedy's idea, and he had been working on this idea for twenty years before that, working in animal models.

(Melody at 00:16:26) So I was fortunate to meet him at the very beginning of the work with humans, and we got an NIH grant, National Institutes of Health, to implant eight patients, so eight locked-in people. And over the next five to ten years, we did that and tried to look at some of the things that would make it work better. And I got very interested in non-invasive methods of doing brain-computer interfaces. And as you said, not everybody wants to go through a twelve-hour neurosurgery, which is what that was. So I actually got interested in things like scalp EEG, just like you would get in a neurologist's office, and functional near-infrared.

(Melody at 00:17:05) And lately, we've been looking at brain signals in an MRI, functional MRI magnet. So we've gotten a variety of different things. And the systems that we've built that are non-invasive work pretty well. There are lots of colleagues that are still working on the invasive systems. And interestingly, Dr. Kennedy ended up implanting himself to continue the research after the NIH money was completed. So that's all in that documentary, Father of the Cyborgs. But we have not done any brain-computer interfaces with dogs. Back to your original question, that is not something we've looked at yet. But certainly with my background in twenty-two years of brain-computer interfaces, it's something that I've done a little bit of thinking about and to see what might be possible.

(Melody at 00:17:52) One of my colleagues, Gregory Berns at Emory University, did a study with awake dogs in an MRI magnet and has done some amazing work with understanding how dogs' brains work. So he worked with trainer Mark Spivak, who taught the dogs to lie still in the MRI magnet, and then they would show the dogs images or a person's face or things like that. And he has sort of proven that dogs actually have a very similar response in their brain to a loved one as a human does. So the hypothesis there is that dogs really do love us in the same way that your family does, that dogs love us the same way, and that dogs have areas of their brain devoted to human language, which is astonishing if you think about it. There's a species with an area of their brain devoted to another species' communication method. Now that makes sense if you know how dogs evolved sometime thirty to fourteen thousand years ago. Dogs decided to be domesticated.

(Melody at 00:18:59) If you look at the work of Brian Hare, he's done some excellent work on showing, demonstrating that dogs probably decided to be domesticated, that humans didn't capture dogs and forcibly domesticate them. Dogs chose to live around us because it was advantageous to them. And then the dogs became more friendly to humans, less sharp teeth, literally, less pointy snouts, lots of really fascinating stuff on the evolution of dogs. That they wanted to live with us, and they still do. So that's been fascinating too, is learning about all the peripheral things that go with being around dogs on a continual basis.

(Joel Beasley at 00:19:40) Yeah. It's definitely mutually beneficial. At the origin, you hang out with our tribe and alert us, we give you some of our meat, you know?

(Melody at 00:19:47) Yes. Well, you know that wolves and coyotes don't bark naturally. They howl, but they don't alarm bark, because in nature, that would get you killed.

(Melody at 00:19:58) So dogs evolved to that. That is a brilliant example of one of the things that dogs evolved to do that made them useful to humans. Yeah. Coyotes and wolves only bark when they're in distress. So that's a, like, help me, I'm being killed by an elephant. I don't know. Whatever. But the barking, guarding, herding, all of those things are behaviors that dogs learned to do to be more useful and therefore get to live with humans. So that whole thing is evolutionary. Anthropology is fascinating as well.

(Joel Beasley at 00:20:33) So you mentioned functional MRI. My brother and stepmom are both physicians, so I hear them talk a lot. Not an expert at all. I just am around it sometimes. And the way you mentioned it almost made it sound like portable. Is that what it is or no?

(Melody at 00:20:46) No, we wish. No. It's a big giant full-size magnet. Okay. But what we're hoping to learn though on the brain side of things is where in the brain are the activations happening? And is that something we could detect from a scalp electrode, either a functional near-infrared or an electrical signal from an EEG? So those are the things that we're studying with the MRI. The MRI is never going to be practical for communication, but it is something that we can learn a lot about what's actually going on in the motor cortex. Fortunately, the cortex of the brain is right under the skull. So it's pretty easy to get to it with external methods without having to do an implant. So that's what we're doing is just learning, mapping it out, figuring out where things get activated.

(Joel Beasley at 00:21:35) Oh, that's way easier socially too. I mean, that resource is super easy. Everyone's thumbs up with that. So just out of curiosity, are dogs' brains from one dog to another, is there a greater variance? Are they more different than from one human to another? Are they about the same?

(Melody at 00:21:54) That's a great question. So the organization of the human brain, with a few exceptions, tends to be pretty uniform. Now you've got the differences between right and left dominant people. The organization of the brain in a left dominant person can be very different than a right dominant person. But the basic pieces are all pretty much the same. And that's actually true of dogs as well, except that with selective breeding, you have very different head shapes.

(Melody at 00:22:24) So you've got a German shepherd with a long pointy nose, a border collie with a long pointy nose, and then you've got the little pug with the brachycephalic, the smashed face of the bulldogs and things like that. The same structures are in there, but the shape of the head is completely different. You've ever seen the skull of a chihuahua? It's like a little apple. It doesn't look like a wolf skull, which is probably what a border collie skull looks like.

(Melody at 00:22:51) So some of the organization of the brain is going to look a little different, although all the features of that brain are still in there. So a large part of the dog's brain is olfactory bulbs. So a large part, not surprisingly, of a dog's brain is devoted to smelling. And that's how they understand their world best, is by sniffing and smelling, which is why they're always doing that. But the olfactory bulb can be very smashed into the brain in a pug, or it can be more of the original, what we call the wolf brain, a little bit more of that organization.

(Melody at 00:23:30) But it's still about a third of the dog's brain. So it's an important part.

(Joel Beasley at 00:23:35) And so they're generally structured the same between humans as far as the location of the motor cortex and memory and different things like that. Are there similarities between the dog's locations and the human's locations?

(Melody at 00:23:46) Actually, there are. There are similar parts to a dog brain and a human brain. And of course, the dogs hear sounds and process sounds. And Greg Berns and Mark Spivak discovered that they can actually process human language in something very similar to the human Broca's area, which is where we process language and sounds, Broca's and Wernicke's areas, and they do have a similar type area. So it's fascinating to find out that the mammalian brain does have a lot of similarities between species, even though we are very differently built than dogs.

(Melody at 00:24:23) Our brains do have some similar components.

(Joel Beasley at 00:24:26) Very interesting. Thank you for indulging me on my—

(Melody at 00:24:29) Yeah. Sorry. I can digress all day. So you just—

(Joel Beasley at 00:24:33) So I'm an entrepreneur. Right? I've built companies, and I really like the market and understanding value and things like that. But I'm curious, where is this technology that you're working on? Has any of it been commercialized? Is it all still in research? Does anyone have these sensors out in the wild?

(Melody at 00:24:51) Well, that is a great question too because I am very interested in making sure that the work we've been doing for the last ten years does become available for the people that need it. So we have developed prototypes of, for example, a medical alert vest that if someone is having a seizure or maybe a diabetic crisis, the dogs can be trained—they've already been trained to do this kind of stuff. They can detect thirty minutes before you're gonna have a seizure. The dog knows it, if they've been trained to do it. And some dogs do it naturally.

(Melody at 00:25:25) And so what those service dogs are trained to do is to say, okay, you're about to have a seizure. They'll push you up against a wall or into a chair so that you don't fall when you have the seizure. And then when you start to have the seizure, the dogs are trained to lick the person's face and to try to rouse them. And so these are the things that the dog can do is stay with the person. But what if that dog could reach around and press a button or pull a tab that calls 911 with your GPS location?

(Melody at 00:25:54) Also, at the same time, it texts your family and says, we're at Starbucks on 5th and Spring, and she's having a seizure, so that your spouse or your children or somebody responsible would know that this was happening to you. But in the meantime, 911 has been called with some sort of a probably pre-prepared message that this is a service dog from a person with a medical condition, epilepsy or whatever, that causes them to have seizures, and the dog has now alerted. And so this could save lives. So, yes, we would love to have someone commercialize this. I'm a scientist and a researcher and an educator, and so that is my purpose.

(Melody at 00:26:35) I won't use the word passion because we just had that discussion, but that is my purpose. And so I am not an expert in entrepreneurship, but I would love for someone who is interested in commercializing this type of technology to talk to folks that might think that there's potential in this. And now, obviously, the disability market might not be that big, but if we could get insurance companies to cover some of this stuff, I think that it would be really valuable for people with severe medical conditions. There are a lot of other uses besides medical alert. I have a colleague at Canine Companions who has a muscle disability.

(Melody at 00:27:17) So she uses a wheelchair, but it affected her ability to speak. So she can't speak very loudly. She signs. Her dog understands American Sign Language. And so one day they were in a dog park. She was playing with her dog, tossing a ball for him or whatever. And she got her manual wheelchair stuck in a field, in a rut in the field, and she couldn't move the wheelchair. Well, she couldn't call for help either because she can only whisper. So she gave her dog the speak command to bark, which she's trained to do, but they were in a dog park, so nobody thought that was weird, that there would be a dog barking in a dog park.

(Melody at 00:27:55) So she sat there for a while, like, a couple of hours before a human being finally walked by, and she was able to wave them down and say, help, please help me get out of my wheelchair. So we actually built her a system that allows the dog to walk up to a person, activate a sensor on his vest, and a speaker says, "Excuse me, please follow me. My handler needs your attention." And so that would allow the dog to go and get help, because unlike Lassie, films and movies and things like that, when a dog runs up to somebody, they very rarely will follow the dog.

(Joel Beasley at 00:28:32) Usually if it's barking at you.

(Melody at 00:28:34) If it's barking at you, you might just go, oh, I'm afraid, I'm gonna run away because this dog is gonna bite me. But this allows the dog to literally go up and communicate with someone. Now, when we tested this, and we were on the Canine Companions campus when we tested this, so these are people who are expecting service dogs. But we taught the dog to go find a human, go find the nearest human. He went up, he activated his sensor. The speaker said, "Excuse me, I need—please follow me."

(Melody at 00:29:00) And the person jumped back about six feet because they'd never had a dog talk to them before. So this is something that we would like to become more mainstream, but this is just an example of another way this can be used. A third way this could be used is that Canine Companions also raises and trains hearing dogs for people who are deaf. And so the way a hearing dog works is they're trained on specific sounds like the baby crying or the sound of somebody calling your name or the doorbell ringing. And so what they'll do is they go up and they alert, poke the person with their nose or however they want them to alert, and then they'll have the person follow them to the source of the sound.

(Melody at 00:29:40) So that's great if somebody's calling your name or the doorbell's ringing. The dog just gets up and takes you to the front door. What if it's a tornado siren? The dog has no way to lead you to the source of the tornado siren, but that's really important information to the person that's living in that house. They need to know that. So we developed a vest that has several different affordances on it that allows the dog to give several different messages.

(Melody at 00:30:04) One of them might be the baby's crying, one of them might be the doorbell's ringing, and the other one might be there's a fire alarm or there's a tornado alarm. Any sound at all, the dog can be trained on. And it actually sends a text to the person's cell phone, or it can send a message to their head-worn display. So Dr. Starner, Thad Starner, was actually the lead developer on Google Glass.

(Melody at 00:30:28) So he knows a lot about head-worn displays, and so we've certainly built some systems based on that. And this is one of them that the person would just have a message flash up in their head-mounted display that would say, hey, there's a tornado siren going off. Maybe we should get to the basement. So the dog can talk much more clearly. And sometimes I'm taking a nap. I don't necessarily want to answer the doorbell. So it would give you a lot more information about what's actually happening without having to get up and go follow the dog around to see what is happening. So once we started looking at this area, we realized there are so many different ways that this type of communication can be used. And one more story. You mentioned the guide dog. So we had a student who's graduated now, but he was completely blind, and he did his undergrad and his master's at Georgia Tech.

(Melody at 00:31:20) So he was on campus a lot with his dog from Seeing Eye, which is a group we worked with up in New York. So he had a Seeing Eye dog, and they were going between classes on a very familiar path one day. He'd been at school for a while. He knew where he was going, and he knew what the hazards might be, et cetera. And then all of a sudden, in the middle of the sidewalk, the dog just stopped and sat down.

(Melody at 00:31:45) Now, the guide dogs are trained with something we call intelligent disobedience. So if the dog perceives there's something dangerous happening right in front of you, they'll stop you from moving, and they won't let you go into danger, like walking into a street or falling into a hole or something like that. And so that day, the dog stopped in the middle of the sidewalk, and my student said, "Dude, what's the matter? Let's go." He pulled out his collapsible cane to make sure there wasn't like a branch on the road or something like that.

(Melody at 00:32:14) He pulled out his cane and he felt all around. He's like, I don't—there's nothing here, buddy. I can't feel anything here. So he said, "Let's go." And the dog said no, and he sat.

(Melody at 00:32:24) So finally, he was late for class. So he's like, all right, look, you got to quit fooling around. I don't know if you found a hamburger on the ground or what, but we got to go. Let's go. So he gives him a little correction, a little bit of a punishment. Like, let's go. You're being bad. And the dog said, okay. And they both stepped into wet cement.

(Melody at 00:32:42) So the dog knew there was something wrong with the sidewalk, probably because it smelled funny. He knew there was something wrong with it, and he had no way to articulate to his owner, his handler, that something was wrong. So we actually developed a guide dog harness with buttons on it and proximity sensors. We tried all the different sensors that allows the dog to say, there's something in our path. You need to go around.

(Melody at 00:33:07) Please follow me. So, again, yet another use for this kind of thing, and that's just with service dogs. So we thought about military applications, police applications. Dogs can actually tell the difference between the smells of different explosives and different accelerants. And so we actually have done some experiments, not with explosives, but with just essential oils and things like that, to prove that dogs can actually tell the difference between, say, birch and clove scents, and they can either activate something on their vest to tell you which one it is or even a touchscreen.

(Melody at 00:33:44) We've actually have a project where we have dogs activating on touchscreen to tell us what scent they just smelled, and we've shown that that does work. So the dog could tell you, oh, that explosive that I just found is C-4, which is pretty stable. You could bump it and it's not going to go off. But if it's something really volatile like TATP, if you mess with it at all, it's going to explode.

(Melody at 00:34:06) And so that's something the bomb squad wants to know, as well as the canine handler. So we even are looking at military applications where dogs would be able to communicate something like that. Again, could save lives. Absolutely.

(Joel Beasley at 00:34:20) Yeah. I didn't fact check this, but I was listening on Joe Rogan podcast of this doctor. I don't know how to say his name exactly, but he's like a Neil deGrasse Tyson type. Okay. Yeah.

(Joel Beasley at 00:34:29) I just don't know how to pronounce his name. So people are probably yelling at me right now. But I've seen him around. I've listened to him talk hundreds of times. But this last talk I listened to a few weeks ago, he was talking about dogs and how significant their ability is to smell and that to the dogs, the world looks very different. It's a bunch of smells. And then he had also said, this is the fact check part, that dogs have a higher accuracy of detecting cancer in people than most of the cancer tests. And I was like, whoa. That's crazy. We need doctor dogs. You know?

(Melody at 00:35:03) Well, my colleague Cindy Otto up at Penn Vet, so University of Pennsylvania, that school is doing exactly that. They are training dogs to detect cancer. And these dogs are extremely reliable with very, very tiny samples as well. So I think they're looking at cervical cancer. They're also looking at detecting COVID. The dogs can tell if you have COVID or not. So those studies are underway and very, very fascinating stuff, very, very important.

(Joel Beasley at 00:35:35) I have a Berner, Bernese Mountain Dog.

(Melody at 00:35:38) Yes.

(Joel Beasley at 00:35:38) And he's about eight years old. So, you know, he's towards the end of his life, but he's still very happy and things are going well. So my wife went into labor about two days ago. But from the moment, maybe a day or two before the labor, that dog will not leave her side. Usually, he's cool with her going downstairs and going to bed, and he just kind of lays around, but he's anxious. I was watching him on the security cam pace back and forth the whole night until she came back up. And Michelle, my wife Michelle, she was like, he knows that the baby's coming and it's going to come soon, and that's why he's so anxious and wanting to be right next to me and all this stuff. And I was like, that's crazy. That was amazing. With all three of the pregnancies that we've had.

(Melody at 00:36:20) So he's a pregnancy service dog. That's awesome.

(Joel Beasley at 00:36:22) Pregnancy service dog.

(Melody at 00:36:24) Well, I have had friends whose dogs wouldn't quit smelling, you know, like their stomach or something like that, and they went in and, sure enough, they had a tumor.

(Joel Beasley at 00:36:32) Wow.

(Melody at 00:36:33) And how the dogs can detect that? I know that there's some really subtle smell that you probably give off if you have cancer, but the dogs were smelling the exact body part that had the tumor. How do they know? It's fascinating.

(Joel Beasley at 00:36:46) My thoughts are it's like infrared light. Right? In the fifties or whenever before we discovered it, if you were to say that there's this magical light that's around us that can't be detected, you would think the person's crazy. You know?

(Joel Beasley at 00:36:57) But these dogs have this ability to detect this, and it's just up to us. Earlier you mentioned something about putting the dog's brain under the MRI and them being able to see that they can love. We've known that for a long time.

(Melody at 00:37:11) Right. Anybody who has a dog knows that.

(Joel Beasley at 00:37:13) Yeah. So we just use science later to further prove what we already know to be true. But there's definitely data out there that we're not picking up on that we don't have the censoring technology to pick up on that they are picking up on. And about three years ago, a company tried to recruit me to come be a tech leader at this company and what they were doing, they were extracting somehow the sensory cells from dogs that could detect bombs, and they were putting them into this organic microchip. Their plan was to sell them with TSA. So you essentially have the bomb dogs or whatever. We have that technology, but without the actual dog because they could do it with these cells. I don't know if they're still doing it or if it ever made it to market. But when I saw that, I was like, man, we live in a really cool time, don't we?

(Melody at 00:38:00) Yeah. That's really interesting work. I do know that there's never been a machine built that can detect odor as well as a dog. Yeah. So we're still pretty far behind. But I guess if you take elements of the dog brain, that might be the only way to achieve that. But I'd rather have the whole dog.

(Joel Beasley at 00:38:20) Me too. Me too. I like we employ those dogs. Right? I'm pro dog jobs. That's a lot of them. I'm running on, Melody.

(Melody at 00:38:28) Exactly. And, you know, some animal rights activists say, oh, dogs should be free. But dogs chose to be with us for a reason. Dogs love to work, and I'm a lifelong dog owner. And dogs that have jobs are happier. I can just tell you from my own personal experience that, of course, they have to like their job, but all my dogs love their jobs. They absolutely love their jobs. My dogs go to work with me a lot. They drag me into the lab because they want to do the studies.

(Joel Beasley at 00:39:01) Yeah.

(Melody at 00:39:01) They just think it's the most fun thing in the world to learn how to activate a sensor. My border collie can drive a robot, and she thinks that's the funnest thing in the world. They just love it. So I actually think that some of the folks have got it wrong, that dogs don't necessarily want to be free. They want to be with us, and they want to be doing things with us, and they want to be learning, and just like a human, they want to have something to do. And doing something that is cool and interesting is just as great for a dog as it is for a person.

(Joel Beasley at 00:39:32) So I'd be interested to see dopamine studies within dogs. That'd be pretty cool. Well, this was great. I mean, is there anything that we didn't cover that we want to get out there to the world?

(Melody at 00:39:42) One thing that I wanted to mention while you had said, what else would we want to have commercialized? We'd love to partner with somebody that's interested in doing that, that knows what they're doing in the manufacturing and marketing and all that kind of thing space. And this is another reason that I love research. We were looking at some data from one of our bite sensors for the FIDO vest, and it was three different dogs, and it was a bite. And the study was that we asked the dog to bite 10 times, and we wanted to see, you know, how hard were they biting, that kind of thing.

(Melody at 00:40:13) And so they showed me the graph of three different dogs, and I said, you know what? Looking at this data, I can tell you which dog this is. So they weren't labeled, but I said, this top one, that's like a metronome. So, you know, I said 10 times, you know, get it, get it, get it, get it. And it looked just completely even. That was my border collie, Skye, who was perfect in every way and was a service dog, could have easily been a service dog, passed the service dog test. The middle one was a dog we had raised that had failed Canine Companions because he was too soft. And so you see, we told him to bite and he'd kind of bite, and then we'd tell him to bite again. He'd bite a little bit, and it just looked very, you could see he was trying, but it wasn't up to threshold. And then the bottom one was my friend's agility dog, who was super high drive, over the top, and his graph looked like this, you know, it was up and down, up and down, up and down.

(Melody at 00:41:03) I said, that's Blitz, that's Schubert, that's Skye. And they said, yeah, you're right. And I said, I wonder if we could tell something about a dog's temperament by giving them a bite sensor and just looking at the data. And so this started another project, which we call from a company called DogStar, which actually the funding came from the military, but we were interested in looking at patterns of activation in an instrumented dog toy.

(Melody at 00:41:31) So we started out with an instrumented tennis ball, something that was the size of a tennis ball, a silicone ball, and an instrumented tug toy. And I just wanted to see, can we tell the difference between the dogs that are going to make it as a service dog and the dogs that aren't going to make it as a service dog. Since I had a released dog that, you know, Canine Companions usually lets the puppy raiser keep the dog if they fail, so we kept him. And so I thought this was really, really fascinating. So we did a two-year-long study with Canine Companions where we went in and all of their dogs that had just come in for advanced training, so these dogs are about a year to 18 months old, we tested them with the ball. Five minutes with the ball, throw it 10 times. Five minutes with the tug toy, you know, have them tug on it 10 times. And then we followed their careers to see how did they end up. And it turned out that we could predict with almost 90% accuracy which dogs were going to graduate and which ones weren't. And Canine Companions graduates only about 40% of their dogs that they raise, then train, and breed specifically to be service dogs. But that's how perfect the service dog has to be because this dog is going to be given free of charge to somebody that has a disability.

(Melody at 00:42:42) And they probably don't have any motor strength. They might not be able to speak. These dogs have to be perfect. So being able to release the dogs earlier that aren't going to make it, even if we're wrong 10% of the time, because I think it was 87.5% accurate. Even if we're wrong, we still could save that organization, which depends on donations for their operation. We save them $5 million a year if we can release those dogs early. So that was something that was a big, it's like, wow, we can do quantification of temperament. So imagine how that might be for a breeder of agility dogs, say. Which ones are going to be the best agility dogs? We might be able to tell just from five minutes with a ball.

(Melody at 00:43:30) What if, you know, you're looking at nose work dogs, detection dogs, is this dog going to be a good bomb dog or a better cancer dog or, you know, something else? We're actually looking at those kind of things with Auburn University right now. We are partnering with Auburn's Canine Performance Sciences group down there, and we have funding from Homeland Security to work on that. Also, we did a longitudinal study with my perfect dog, Skye, the border collie. We said, well, let's see how this goes over time. So we threw the ball for him, you know, 10 times every day for a year. And then we'd take his data and run it through our machine learning classifier to say, you know, did he pass as a service dog or not? And every single day, he passed as a service dog. Until one day, about eight months in, the students ran the data through and they said, he doesn't pass today. And I said, what?

(Melody at 00:44:20) What do you mean he doesn't pass today? Is there something wrong with the ball? So we took it apart. You know, we looked at the hardware. There was nothing wrong with it. Did you guys change some code? No, we didn't. Anyway, eventually, for some reason, Skye rolled over. His whole underside was flame red. He was sick. He had a staph infection. And

(Melody at 00:44:34) oh, wow. I couldn't tell. I, who was his owner and knew him like the back of my hand, couldn't tell he was sick, but the ball could tell. So we started thinking this could be used for medical monitoring. So we currently have a project with the Georgia Aquarium where one of my master's students, Josh Terry, built a toy for the sea otters because they have some geriatric sea otters. And we said, we think that we can build toys that will quantify the behavior of these animals and be able to tell when they're not right. So that, you know, usually otters are very destructive. They like to shake things and bang them against the wall and things like that, and they do that. But we can tell that today they're not banging it as hard as usual, and there might be something going on with this otter.

(Melody at 00:45:25) So the Georgia Aquarium has been wonderful working with us, and we're interested in taking it into other species like maybe the beluga whales or maybe, you know, dolphins and things like that, lots of animals that might be willing to play with a toy or interact with some sort of a tangible object that we'd be able to measure that interaction and tell you something about that animal's health. And that, I think, has huge commercial potential. We think we might be able to take our smart toys, is what we call them, smart toys, into a shelter and assess dogs in a shelter for temperament. Is this going to be a good dog to be in a home where, you know, needs to kind of be a couch potato? Or is this the kind of dog that's going to need two hours of hard exercise every day? We might be able to tell that. So these are things we're still exploring. Again, they're prototypes, but we do think that there's some pretty tremendous commercial potential, and we're pursuing patents on these things right now.

(Joel Beasley at 00:46:17) That's so, it's like a smart toy platform that can do everything from the, all the examples you gave me, they all have a similar underlying technology?

(Melody at 00:46:25) Exactly. They have inertial measurement units, IMUs, sensors such as barometers that could tell if something's being squeezed, proximity sensors that could tell where it is in space, gyroscopes, magnetometers, all those things we've incorporated into some of these essentially toys for animals, smart toys.

(Joel Beasley at 00:46:44) You got like a puppy operating system.

(Melody at 00:46:48) Yes, yes. I used to in a former life, I was an operating system developer, so you're absolutely right about that.

(Joel Beasley at 00:46:54) So just out of curiosity, so you make these discoveries, you find these things, and you move on to the next experiment? Like, there's not an arm behind you that's like, we can commercialize this, we can take it to market, there's not that portion that follows the research?

(Melody at 00:47:10) We are indeed focused on research, and that's what I'm trained in. That's what I love, and that's what we want to do. We want to have the wild idea and go off and prove that something works or find out that it doesn't and figure out how we should make it work. That's what we do. Georgia Tech does have vehicles for commercialization. They have some programs, incubators and things that support innovators. But I don't want to be the person that runs the company. I want to be the one that has a wild idea and goes off to get the next best thing. So that's why I would love to find somebody that is interested in commercializing some of these technologies, because the technologies do need to be out there. I'm just not the right person to run the company, because I don't know anything at all about running a company.

(Joel Beasley at 00:47:54) Yeah.

(Melody at 00:47:54) I'm married to somebody that does. My husband is a CEO, but not me. So we would be, even if maybe some of the larger animal-related companies might be interested in licensing some of this technology, we would be very interested in talking to some folks about that kind of thing.

(Joel Beasley at 00:48:13) That is very cool. And then that technology, it's like ready for market, or it just has to be designed for scale?

(Melody at 00:48:20) Yeah, it would need to be designed for scale. Everything that we do right now is a prototype, so it's not necessarily the most beautiful or the most robust technology. So it would need professional designers and marketers to make sure that it fits the kind of demographics that you might be targeting and to make it work reliably and things like that. But we've built most of the things we've built with off-the-shelf components.

(Melody at 00:48:48) And we do design our own circuit boards and things like that. But a lot of it is off-the-shelf components. So it's not a complicated thing. It's just that the way that it goes together is what we've invented.

(Joel Beasley at 00:49:01) That's crazy though, because I mean, if you can save a company millions of dollars, that's the discovery.

(Melody at 00:49:07) Yes, absolutely.

(Joel Beasley at 00:49:08) Well, I'll put you as our resident animal expert.

(Melody at 00:49:12) I'd love that.

(Joel Beasley at 00:49:14) And this has been really fantastic, and we made a podcast. How do you feel?

(Melody at 00:49:18) Fantastic. I appreciate very much that you guys are shining a light on this work, letting people know that it's happening, that it's a real thing, and it's a possibility to change lives literally for humans and dogs. And we keep finding more and more things that we can do. So we appreciate feedback and input, and anybody that's interested in maybe taking these things forward, we'd love to talk to them.

(Joel Beasley at 00:49:44) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email: [email protected].

(Joel Beasley at 00:50:02) Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.