AI in Hearing Aids with Dr. Seth Dobrin

In this summer rerun of one of our more popular episodes, Dave interviews Dr. Seth Dobrin, an expert in artificial intelligence (AI) and a hearing aid user. They discuss Dr. Dobrin's journey to getting hearing aids, the stigma associated with hearing loss, and the positive impact that hearing aids have had on his life. They also delve into the topic of AI, including the potential benefits and challenges it presents.

Dr. Dobrin emphasizes the importance of responsible AI and human-centric design, particularly in the context of healthcare and the workforce. He predicts that AI will create more jobs in the future. Dr. Dobrin also shares his wishes for the future of hearing aids, including seamless translation capabilities and the integration of health monitoring features.

Link to full transcript. 

Transcript

SPEAKER_00 0:01

Welcome to Starkey Soundbites. I'm your host, Dave Fabry, Starkey's Chief Hearing Health Officer. I am really excited about today's conversation. We're going to dive into artificial intelligence with a world-renowned expert on the topic of AI who also happens to have a hearing loss. Dr. Seth Dobram was IBM's first ever global chief AI officer. And in September of last year, he founded Quantum AI. Seth, thank you so much for joining us here on the podcast today on Starkey Soundbites, and I'm really excited about our conversation.

SPEAKER_01 0:37

Yeah, thanks for having me, Dave. I always enjoy chatting with you.

SPEAKER_00 0:41

Well, let's I don't even know where to begin. I've got so much that I want to fit into this time, but um let's start with your hearing loss. Um we've talked about this in the past that uh your wife Tabitha had been telling you for some time that you needed uh to consider using hearing aids. Why is it that you waited? Because I think you waited almost 10 years, didn't you?

SPEAKER_01 1:03

Yeah, I waited quite a while and probably more than that. We've been married for over 20 years, and more than half our marriage, she's been telling me I needed hearing aids. So it's it's been a while. Um, you know, it was well, it's the stigma, right? I mean, there is a stigma associated with hearing loss. It's I think it's gotten less and less, or hearing aids gotten less and less over the years, but um, you know, it was it was really the stigma of wearing hearing aids. You know, my my great uncle wore hearing aids, and it was so long ago that he'd take his glasses off and he'd say, I can't hear you. I have my glasses off because they were attached to his glasses. Uh, but that was who I always saw as people needing hearing aids. So I just resisted.

SPEAKER_00 1:43

Yeah, and it's it's your your journey is not an unusual one for most people in the US. Uh it's about a seven to ten year period from the time someone either suggests that you get a hearing test, or maybe you should you should go and see someone and talk about it, to when they actually get hearing aids. And so um it's not uncommon. And I and like you said, I would agree. I think traditionally the stigma surrounding hearing loss and especially use of hearing aids, um, has uh really been a big barrier to people in their mindset. And in many cases now um as hearing aids have gotten smaller and and uh I think I I'm biased, I think they've gotten a lot cooler. But um, you know, I I think that we've still seen that the traditional generation, those born before World War II, has some of that stigma involved with them. But what we've seen with the baby boomers and younger uh is that they may be less stigmatized, but they have higher expectations for what hearing aids can do. And I think you fall squarely into that camp.

SPEAKER_01 2:48

Yeah, they have very higher expectations and uh and and for the most part they they live up to them, right? And and just to be clear, I'm not a baby boomer. I would I'm I'm I'm the Gen Xer.

SPEAKER_00 2:58

I know you are. That's why I said I'm the baby boomer, and uh, and you're definitely a Gen Xer, I know. And you know, the the issue, you know, uh I've been fortunate to really um sort of know you as you've been on the more recent part of your journey, and and we've been through several sets of devices that uh you've been kind enough to um provide feedback to me uh as to where we are and where we're going. And the thing I I I really have to say, first of all, is thank you for the radical candor that you provide about where we're uh where I'm doing well in terms of the way that I'm using the technology and where the technology that Starkey has is going and where we where we succeed and and uh where we have opportunity to do better. But before that, let's talk about your experience when you first started to use hearing aids. When when did you first get them?

SPEAKER_01 3:50

Uh so 20 mid-2020. So it was right after COVID and the masking started. Uh so you know it's been three, a little over three years probably now. Um, and one of the drivers for it, well, there are two main drivers for it. So, or maybe three. So one, the first is we talked about already, my wife has been telling me you need hearing aids for a long time. The second was masks. I, you know, I always knew that I read lips, I just didn't know how well um and how important it was for me to, you know, to hear, quote unquote hear people. Uh, and so I really was having trouble hearing people when I was when I was out and about, you know, the mask plus the barrier. So the barrier that was between everyone, all the plexiglass and stuff made it even worse. And then spent a lot of time at home. I travel, you know, most of my time, especially when I was at IBM, I was traveling around the world. And so, you know, spending a lot more time sitting next to Tabitha on the couch. And, you know, one day she looks over me and she says, You're either ignoring me or you're deaf. If you're ignoring me, we're getting a divorce.

SPEAKER_00 4:55

And and so thankfully you made the right decision. Yes.

SPEAKER_01 5:00

So uh she didn't quite say it that strongly, but I think that's what she meant. Um, so I went to my audiologist uh and did our my test. Um, and you know, when when Frank came back and said, you know, gave me the results. At first I turned to Tabitha, I'm like, see, I told you so. And I'm like, oh shit, I need hearing aids. How can I say that? I need hearing aids. Um, and uh, and so you know, it was kind of mixed emotions that, you know, I I wasn't, I mean, I didn't think I was ignoring her, but you know, sometimes you do things sub subconsciously. Um and and I tell you, I I got hearing aids and a friend of ours owns a restaurant, and he was having a party for his a birthday party for his girlfriend, and we're sitting at a table, most everyone's wearing masks, like 12 people, and I'm in the middle of the table, and I turn the tablets and I'm like, oh my God, I can like hear everyone. Like, and she says, Yeah, I know. And I went, Mike, no, I can hear exactly what they're saying. And she goes, Yeah, that's what you're supposed to do. And so, like that now, that was the night I got them, and it was instantaneously, you know, almost life-changing instantaneously over the last three years, it's definitely been life-changing. You and I have spent quite a bit of time talking about that.

SPEAKER_00 6:09

We have. And um what would you say has been the biggest impact? Maybe that that social situation where you're in a restaurant and able to hear that that's certainly one of the top drivers is performance in background noise. But is that the biggest success that you would say over the last time?

SPEAKER_01 6:28

I would actually say no. I would actually say the way I function overall is better. I'm not tired as much at the end, I mean, I'm not exhausted mentally at the end of the day, the way, same way I used to be. Uh, and and I attribute that a lot to really trying, spending a lot of time, especially when I was out with customers or or or my team, you know, talking to people, it was very exhausting for me. Uh, and and so I don't really have that anymore. I mean, it's still tiring when I'm six time zones off or something like that, but not the same, same level it was. So that's that's the probably the biggest you know impact. And then I've seen some little things that were surprising for me. Like, you know, I think I talked to you about this, and I, you know, I thought I was kind of off the wall. You know, I used to have trouble driving at night because I couldn't see. And, you know, about six months ago, I'm like, I I can see at night now. I can drive at night, which I had wouldn't even attribute to that, but you said that's fairly or not uncommon, I guess. I don't know about fairly early.

SPEAKER_00 7:26

Yeah, it's it's it's you know, we talk about the top drivers for people that are considering getting hearing aids for the first time, and it's certainly hearing soft sounds and hearing speech and noise, making sure loud sounds are never uncomfortable, and spatial awareness. And I think that's really what I attribute your observation to is that now you're hearing in that automotive environment and you're hearing more completely as you're driving and you're integrating in the same way that you mentioned vision and hearing is important uh uh for you know your lip reading without even your subconsciously pe people will sometimes say, Oh, I'm losing my hearing, I need to learn to lip read. And I said, You've been doing it your whole life, but now it's just been more important to that. But I think similarly, the spatial awareness uh we say you don't hear with your ears, you hear with your brain, and it's that integration of multiple modes, if you will, that's contributing to that sensation that you're seeing better at night because of the spatial awareness. That's interesting, yeah. Um so what what hearing aids do you have now?

SPEAKER_01 8:33

So uh well, I have the the new ones in right now, so the the Genesis, right? And then right here on my desk, and not just for this call, I have the the previous, the evolve. Um if you notice they're both colored.

SPEAKER_00 8:49

Yeah, and that's um what I wanted to go to.

SPEAKER_01 8:52

Whoops, there it is. And I do and I do that so that you know, when you know, I spend, you know, we talk about me traveling, about half the time I'm traveling, I'm on stage uh giving presentations. And I almost always start out the presentation talking about and talking about AI, and I'd start off by saying, you know, this is very personal for me. Are we hearing aids? And most of the time I'd pull them out and and you know, and they're filled with AI. And I talked about a lot of the ways that, you know, a Chin and I have talked, and you and I have talked about all the AI that's that's in them, uh, you know, like sampling the environment every six milliseconds and and adjusting and and and some of the translation that it can do, albeit slowly now, but hopefully eventually faster. Um, and I talk about that, and this is personal, and why it's important that I trust the company who's doing my hearing aids because they hear everything. They hear my personal conversation, they hear you know, proprietary conversations I have internally with my company and with my customers. And so I talk about it all the time, plus because they're colored and they're not normal ones that go behind the ears every time I go through security or every other time, they're like, take your ear your AirPods out. I'm like, Well, they're not hearing aids. I mean, they're not airpods, they're hearing aids, and then they get all the apologetic, and I'm like, don't worry about it. But I mean, so I I intentionally, you know, I for me it's I'm trying to reduce the stigma by letting people know first that I wear them, and I'm you know, not not uh not not old, but I'm not young either. Um, and kind of sharing some of the cool things they do and how they leverage technology, like you know, streaming directly from phone calls from them, watching videos. I mean, it's I don't even you know, acknowledge half the things they do anymore.

SPEAKER_00 10:30

Yeah, I uh well, first of all, for those who are listening to this rather than watching it on our YouTube channel, Seth um showed me that he's got a pair of black ITC custom rechargeable Genesis AI devices that he's wearing. And then he has another set that is all white. And it was where I was going next, I'm glad glad you brought it up because I really appreciate and and and love the fact that you wear the ones that are not matched to your skin tone because you want to, you're you're not trying to hide it. You're wearing custom devices, which many people assume people want to hide uh the devices, but you're sort of celebrating it in the way that uh you hearables uh by uh drawing attention to it, so that then I would imagine, like me, you then nerdjack the conversation with some millennial and tell them about all of the features that they have.

SPEAKER_01 11:22

Yeah, yeah. In all fairness, though, the first pair I had were skin tones. Yes. Um and and I still have those there, they're my my backup, backup pair. Uh and so, but I went from, you know, I think you and I had a conversation and you said with the white pair, would you mind being more, you know, you're already talking about them, would you mind, you know, see, you know, pair white ones or something? I said, sure, I don't, I probably wouldn't do red, but I'll do I'll do other colors.

SPEAKER_00 11:45

No, I love it. And um, so then let's talk a little bit about some of the features that you like best on the latest product. We've talked about the fact that you travel constantly. And yeah, we're in the with with the Genesis AIs, we're we're in the second generation of rechargeable devices. And the first generation, uh I think in our industry, you know, miniaturizing these batteries, transitioning from replaceable zinc air batteries where the end user is in control of replacing the batteries. But you know, you think about people, older people or people who have manual dexterity issues or arthritis, um rechargeable is easier, but then in that first generation, we can sometimes see people develop range anxiety, where especially for someone who's flying across the ocean or traveling long uh distances would run out of uh battery at the end of the day.

SPEAKER_01 12:42

Yeah, so so uh, you know, I think a couple things, the other thing, the other type of person that doesn't want the battery are lazy people like me, I don't want to have to have to worry about. So I look at it the other way. If it's battery, I have to worry about it, right? So I think the uh the evolve would last me somewhere around 18 to 24 hours nonstop. And and fortunately, you know, pre uh pre-pandemic pre-supply chain, uh, you know, the cases had battery in them. And so when I got on a plane, I would just take the case out and set it down next to me um and um and could recharge them. But there's some times where I'm working through the flight and I've been up all day and and I've had them last 40 hours.

SPEAKER_00 13:22

Yeah, and the genesis.

SPEAKER_01 13:24

So the Genesis, and that's not just in my heat ear that's you know, taking phone calls, that's listening to music, that's you know, so so you know, I think that is completely amazing when you consider that a pair of AirPods, you know, can't even last eight hours. Um, and then the other thing that I find very interesting that most people who wear hearing aids probably don't know is you know, the technology and how you pass the the connect the the hearing aids over your head and not through your head because water, which is 90% of your brain, yeah, right, slows down and fat, you know, the fat in there is even worse, slows down any kind of waveforms going through your head, plus it might not be healthy. So the technology to actually pass it over your head is completely, you know, amazing for me and and and really how does that impact you for people that aren't familiar with it?

SPEAKER_00 14:17

So when you have user controls on these devices, for one thing.

SPEAKER_01 14:20

Yeah, so I have user controls on them, so I can, you know, I can turn the sound up and down. If you're watching the YouTube, I accidentally turned it down, so I turned it back up. Uh, I can start and start and stop music and phone calls by tapping. Um, and then and so that's some of the things is that the user controls on these these in-the-ear, uh, in-the-ear custom devices have. And so, in terms of how the over the head works, so the evolve are the Genesis are better than the evolve than that. And I can tell because when I use the evolve, if I have my phone in one pocket and then I move it to the other pocket, it it gets kind of it kind of breaks up a little bit. You can kind of tell there's something, something wrong. Um, that doesn't happen as often with these new ones. So that's that's one thing that's gotten better. Um, the other thing that I like is that you can there's a mic that I can take when I talk, take phone calls, I can use the mic in the hearing aid. So if I'm in an airport, I can keep my phone in my pocket and still talk to my wife or or whomever. Um, and so you know, so that's that's one thing I like. The other thing that when you first sent me these pairs and you said, you know, you wouldn't tell me what was different, right? And you said, tell me what you said to me, turn around to tell me what's different after a week. And the reduction in background noise is amazing. Um, and because of the type of hearing a lot, hearing loss I have, I didn't really have an echo problem, or at least I didn't think I did with the in-the-ear devices, but I noticed that actually got better. I was hearing myself less. At first, I was hearing myself more. I told you you made some updates, and then I actually started hearing myself less. Um, I think those were the biggest changes that I noticed. Yeah. Um, in these in these new ones.

SPEAKER_00 16:08

Uh the other one that a lot of I think we have talked about in the past and a lot of other patients who've switched from, let's say, a previous generation, Livio or Edge or Evolve to Genesis is how quiet they are. When the in the in a low ambient environment, yeah, they're they're very quiet. And I think we've talked about that.

SPEAKER_01 16:27

Yeah, yeah. So that, yeah, I'm yeah. I think um I didn't even think about it. I don't even think about that.

SPEAKER_00 16:32

Yeah, that's cool. Well, you know, the issue with the battery life and the rechargeable, I think providing that option for still having replaceable batteries for those people who want to be in control, but then for those who want the ease of use, and I'm the I'm in the same camp of rechargeability. And I think, you know, there's a little bit of a environmental benefit. Uh, you know, we still have to dispose of those lithium-ion batteries eventually, but during the time you're not disposing of batteries uh uh uh you know as frequently. But I think our real goal, and Achen and I talk about this along with the RD team, is those the best technologies are those that are ubiquitous and they just are sort of part of your life. You just set them and forget them, and you just put them in and you don't have to think about it, and you don't have to worry at the end of a long day. If you've got 41, 40 hours uh you know of use, including streaming time, that's where we are today, but you're still gonna have all day use three to five years from now. And and so we want to take the battery life worry, the range anxiety off the table.

SPEAKER_01 17:35

Yeah, yeah. So I'd like to see some more compute heavy functions, and I'll take you know half the battery life because I'm not not very often up for not many people are up for 40 hours straight, right?

SPEAKER_00 17:45

No, and let's transition into that because you know we've talked about AI, and and that's what we're gonna go into next here a little bit in greater detail, because I really want to have the audience be able to be the beneficiary of your expertise in terms of artificial intelligence. But edge mode is a feature that we've incorporated that is designed to personalize the devices beyond what the automated processing, where we're using machine learning to adapt the devices throughout the day. Edge mode actually takes advantage of an onboard DNN accelerator to start to look at personalizing to the individual's uh uh challenging listening environments. And talk a little bit about because you know, even though you are deeply a tech guy, you also are a person who just wants to put the devices in and just wear them without engaging with them more than you have to.

SPEAKER_01 18:36

Yeah, and so when I first when I got my first devices, I would use edge mode, and I think I complained about having to tap it twice all the time because I wasn't using it right. Um, and so I was I was and it the the underlying problem was actually that my hearing aids weren't set right for what I needed, and so I was used using edge mode to try and overcompensate for that. Um once we got them set right, I don't think I ever use edge mode. Um, I don't see any value in it really, for me at least.

SPEAKER_00 19:11

If we got the if we've got the acoustic set properly, um then it if it adapts to your world in all of those challenging listening environments, that is really the goal, ubiquitous kind of performance and connectivity and battery life.

SPEAKER_01 19:25

And I have tried using it on a plane, but the problem with the plane is that the sounds adjust all the time. And so it's not really, I don't think edge mode works well for that. So I use like I've adjusted, I have a I use the crowd mode for that, and that you works pretty well on a plane.

SPEAKER_00 19:41

Yeah.

SPEAKER_01 19:41

Um, and so so I don't use edge mode at all. Um, especially knowing that, you know, having the you know, having like I said, having the conversations with you and Achen about, you know, all the the sampling. I think last I talked to him it was every six milliseconds, yeah. Which is which is just amazing that on this teeny tiny little device, you can have the power to to do that and and still have 40, 48 hour battery life, right?

SPEAKER_00 20:05

Yeah, and so you don't have to engage with the devices. And that's really that's one of the objectives in using machine learning and AI and deep neural networks. But let's pivot then from your experience and thank you for that feedback. And I appreciate the ongoing feedback and where we get it right and where we can improve. But um, let's first, before we dive into that, let's talk about your background in AI. When, when, you know, what first got you interested in this?

SPEAKER_01 20:34

Yeah, so I got interested in graduate school. So my PhD is actually in human psychiatric genetics. Um, and I was doing it during the towards the end of the human genome project, and we had tremendous amounts of data, and this was in the late 90s. Uh, tremendous amounts of data, even even by today, it's large data, right? So, you know, for those of you who understand different sizes of data, it was my dissertation was about 100 gigabytes of images and and tab and numbers. And so, and I had never taken a computer programming class. Um, and so I had to teach myself how to code in order to analyze all that data because back then Excel was limited to 65,000 columns and 10,000 rows. If I had to do it in Excel file, it would have been hundreds of Excel files. Yeah. Um, and so I had to learn how to do programming and how to code and started using machine learning uh back then. And machine learning is just a type of statistics. It's actually, when we think about machine learning, it's just statistics that's applied and continuously learned. So statistics is static. Machine learning continuously learns, and and it's the same, all the same math, literally. Right. Until you get into things like deep learning or neural networks, like you mentioned, or into the new, you know, new, the new AI, I call it the old AI pre 2023, and new AI was it's really all still related though, but that's just Kind of tongue in cheek. That's getting into net new math. And that's that's uh so so I got into this. I'm getting off course a little bit, uh got into this in the late 90s, kind of as at out of need uh and been applying it in both startups that I worked at in academic situations at research institutes and at Fortune 500 companies. Uh and now I've taken that uh and really focused on the business strategy of AI and my company, Quantum AI.

SPEAKER_00 22:29

Yeah. And when we first met, you were the first uh chief AI officer at IBM. And um you know, you've talked about a lot uh on YouTube talks and I think even Ted, I think TED Talks too, you've done, and um on sort of responsible AI and how AI, big tech is transforming AI. Can you can you for the uninitiated, can you, and maybe even those who are a little wary of AI, can you give us a boil-down explanation of AI and big tech and then even where you're going with responsible AI? Because we're you know involved in that company as well.

SPEAKER_01 23:08

So so that's a really good question, especially when we think about most everyone has at least heard of Chat GPT. Uh and these are very, very large models that are trained on essentially the whole of the internet, um, and including some of the worst parts of the internet. So, like a lot of how these models learn how humans communicate is through Reddit. And if you're not familiar with Reddit, it's basically a free form conversation, you know, place for conversation. There's subreddits. Now, the downside to that is it's you get the worst of humanity on there, too. Uh, and they start talking about all sorts of things: misogyny, misogyny, race, racism, bias, hate, misinformation, all of these things that you know you wouldn't want uh uh an AI to learn, but because it was trained on that data as well as other data on the internet, it actually learns this bad behavior. Uh, and so the math is not biased, computer programming is not biased, all the biases come from the data that these models learn from, machine learning, uh, and all that information comes from us as humans. And so I always, you know, it's it's basically a mirror on human humanity. And oftentimes when I end talks or podcasts, I often say, if you don't like what AI is doing, look at yourself.

SPEAKER_00 24:27

Yeah.

SPEAKER_01 24:28

Because everything that you're doing on the internet uh is feeding into that.

SPEAKER_00 24:32

Yeah, and that's what I find compar particularly compelling about the way that you describe this and approach it, because you talk about really responsible AI as being essentially human-centric and and so, as you said, reflecting humanity for all the good and bad that it is. And you also mentioned Chat GPT. I mean, I will say that I've used Chat GPT to streamline abstracts for for papers that I submit, where you know, if I if they have a 250-word uh maximum and I'm at 312 words, I will dump in my prose and say, make it boil that down, get it to 250. And it does a pretty good job of that. But then when I think of human-centric, the gray areas, how do I, how do I remain true to myself while using a tool like Chat GPT to solve problems for me? Is that still remaining human-centric?

SPEAKER_01 25:28

Well, so there's there's there's two parts to this. So one is our consumer apps like ChatGPT or similar technologies. And I use it literally every day. In fact, I teach a master's level course at a university in New York on how to use this. So the topic is AI safety, but throughout the whole course, it's how to use generative AI. And you know, no cheating exists in a class. There's no such thing as cheating. Use every resource at your disposal, including these tools. And I in fact gave a had a conversation with the professors at at CUNY Baruch, which is where I teach, about how they can use it. And they were amazed, right? Building curriculums, doing tests, doing doing all these things, and so kind of encouraging them to let them use their students. So there's these consumer-facing ones. And so when when Dave or Seth use Chat GPT, you know, the human-centric is really in the in the in the GUI in our perspective. When I'm talking about human-centric, or I've actually started talking about as human focus now, I'm really talking about how businesses use it. So how Starkey would leverage it, for instance, right? So when you think about how you're going to use AI, you have to start with the human. So who is the human that's going to be using the AI? And who's the human that's going to be impacted by the AI? In the case of hearing aids, they're the same person. But if you look at like a mortgage, if I go to a quick and loans, which is where we've always gotten our mortgage from, or rocket mortgage, now it's called, and I get a loan, they're going to use AI. So the underwriters and the mortgage brokers are using AI. But ultimately, I'm the one that's impacted by the AI. And so when you're designing an AI system for like credit, you know, for lending, you need to understand the human that's the both humans. So you need to understand who's going to be using it because they're not going to start using it unless you're designing it for them. And then you need to understand the specific impacts of the humans that are going to be that are going to be impacted by or specific ramifications of it or potential harms. And when you do that up front, it enables you to understand things like bias. Yeah. So what are what are the protected classes is what we call so groups of people that we don't want to be biased against, because you'll never be able to eliminate every bias. In the case of lending, it's you know, age, race, gender, ethnicity. Uh, and so you need to keep that up front and and and really understand that. What do I need to do to explain the outcome? So explaining the outcome to you or Achin is definitely different than explaining it to my dad or your mom, right? And so you need to understand what explanation looks like to them. Um, and and in the case of bias, right? Bias, and this is something that people don't think about, bias is actually a social construct. And so bias in the Western world is fundamentally different than bias in Asia or the Middle East or Africa or South America. And the example I always use is, you know, in the US, we worry about racial bias and ethnic bias. If you go to Asia, there is for the most part no racial or ethnic bias. Right. And so different construct of what bias is over there. I was at giving a presentation the other day uh at a for a bunch of executives, and there was a woman from India, and I was talking about this, and she said, you know, most Westerners don't talk about this, but like in India, it's gender bias, it's caste bias, and it's what schools you went to, right? So those are all driving bias in India. And if we design AI without thinking about the human and where that human lives, we may think we're controlling bias, but we're actually not, and we're propagating biases that are important for that region. And if we don't do this, we're going to increase what we call the digital divide. So usually it's north of the equator, south of the equator, but but it's there's other things. And and also the socioeconomic divide, because we're designing these things for people that are middle class or better. We're not really designing these things for socio, you know, economically disadvantaged disadvantaged people. So we need to think about these things while we're designing the AI. That's what I mean by human-focused.

SPEAKER_00 29:39

Yeah, so many good points in there. And then, you know, I do have to comment, you know, you mentioned at the start that when you're wearing our hearing aids and we're saying we're using machine learning, we're starting to use DNN, um, and we're learning, it's learning, but you know, we are a class two, most hearing aids are a class two medical devices. And so we we are required in terms of data privacy and patient confidentiality to I want to set the the set the the stage first that we're not monitoring uh and recording what you're hearing. But you know, one of the issues and one of the concerns when you start talking about responsible AI and and and ensuring the ethics of this is that with the class two medical devices versus an over-the-counter type hearable that isn't encumbered by that same degree of of certainty that the data privacy and and you're not recording things and monitoring things, you know, how do we how do we deal with that as we live with a world where there is both prescriptive devices and where there are an uh uh OTC devices that may not be as restricted in the way that they monitor data?

SPEAKER_01 30:51

Yeah, so I think as a class two device, you need to adhere to HIPAA. Yes, of course, yes. So uh and so whereas the over-the-counter don't because they're not considered med devices directly, you know, same kind of level of med device. Yeah. And most people don't understand how HIPAA applies, but it really only applies to certain devices, providers and insurers, so people who you're giving, you know, who are responsible for your health care in some way. If you give your health your health information to someone who's not, doesn't fall into one of those categories, there is no HIPAA protection. Right. And so they can essentially do whatever whatever they want with with your data. Um, and and so I think the first thing we need to do is educate ourselves. So, how is our data being used? Read terms of service, understand as best you can, use chat GPT or some other tool to tell you what's being you can put something in there and say, hey, Zoom just updated their terms of service. Are there things I should worry about? And and it will tell you. In fact, back in March, Zoom updated their terms of service, and they actually, you were agreeing when you use Zoom to let them record every single conversation, use it to train their data. They basically took ownership of even proprietary information that was being discussed in Zoom calls.

SPEAKER_00 32:06

And yet everyone swipes by the EULAS. Everyone do you read the EULA agreements on everything that you enter?

SPEAKER_01 32:13

So I scan through them, yes. I think there's a couple companies I trust. Like I generally don't do that for Apple. Um, I generally don't do that for Microsoft. I mean, I generally trust that they're doing the right things. Uh, but with I do have a a prompt. So with the way you interact with these generative AI systems like ChatGPT is you interact through that through prompts. And I have a prompt that's built to read EULAS and pick out certain things from your end user license.

SPEAKER_00 32:40

I was just gonna say, yeah, I use the acronym too. So thank you.

SPEAKER_01 32:43

So I have I do use you know these tools to interpret uh end user license agreements for for me. But you know, that's a good point about the devices. People are often saying, well, you know, you're paying all this extra money, you have to go to an audiologist. Um, well, first of all, I mean, most of your listeners know you don't have to go to your audiologist very often. Um, right. I mean, I go see Frank, my audiologist, because I like him half the time. Um and um and you have certain, you know, the in order to be a med device, a class two med device, you have to meet certain guidelines beyond just uh, you know, beyond just the data privacy, which is important in terms of, you know, understanding the different harms, understanding, you know, different failure modes, things like that. And so I, you know, my hunch, and I've never used the pair of OTC over-the-counter hearing devices, is that the the prescription ones are are gonna be more reliable, more robust. Um, and certainly from the data perspective, I, you know, there that you have no choice. I mean, I say I trust you, but you have no choice. Right. We don't have to conform to the basically get fine out of business.

SPEAKER_00 33:52

Yeah, and then also just you know, just closing the loop on the discussion about bias and bias globally, regionally differs in the same way the compliance differs. You know, MDR and and EU requirements are different than HIPAA, even more restrictive in many ways. And so we have to, as a global manufacturer, we have to conform to the data privacy and and regulatory requirements in all of the areas where we do business. But I think the point you raised very importantly is considering with uh training for AI models, and it's it's not the math that's biased, it's the data that's going into it, and also the regional differences is something that we really need to think about moving forward, how it is that we're optimizing for different locations around the world where we're doing business.

SPEAKER_01 34:40

Right. And then on the different geographic data protection, there's 27 different data protection regulations in the US today. Right. Um be 50 in the by the end of the next two years. And so that makes it even more complicated.

SPEAKER_00 34:54

So and so you know, we have to go to the to the highest watermark, and that's one of the issues. And you know, the last question I'll ask on this, and then we we need to wrap up. I see I've already used more time than I promised I would take of you. But um, you know, the the issue of you bring up OTC and prescriptive approaches. You're a technology guy who could easily adjust these devices yourself. Yet you could go to, you know, you choose to go see Frank, your audiologist, and then we do telehealth from time to time, just if there's anything new I want to try out on you. Um, and um uh why is it that you choose to continue to use and go down the prescriptive path rather than an OTC path?

SPEAKER_01 35:37

Um well, a few reasons. One is, you know, I am while I understand the technology probably better than most, I'm not an audiologist. I'm not trained in this. And you know, much like you know, AI can be it can can, you know, AI can pass the bar exam today. I don't want an AI defending me in court. I want a human attorney defending me in court. Much like an AI can pass, you know, the medical exam. I don't want an AI practicing medicine for me. I want to interact with my doctor and I want her and I to make decisions, hopefully based on some AI. But it needs to be a conversation between her and I and my situation and her experience and and you know what's right for me. And so so you know, I'm not I'm getting off tangent, but not really. No, I you know, I because because you and Frank are professionals, I trust you to A answer some questions as to why certain things might be happening, and then B, make the adjustments that that fit that. And and plus also, if I go into my audiologist, they keep the old record. So if I screw something up, they can roll it back.

SPEAKER_00 36:46

Yeah, we can we can undo, undo if we need to. But yeah, but yeah, I I I really appreciate that. And I just uh I I I didn't know what you were gonna say, but I was curious about that because from a technology standpoint you can certainly handle it. But I I've heard you speak before about uh you know AI now, AI next, and AI never. And one of the things you say is and you just repeated it with a physician, you want them to use the tools that AI can provide. If they're a uh uh looking at um uh radiology uh uh files and monitoring them over time, AI can do a great job of looking for small spots, but then you still want the plan uh for what how you're going to intervene to be made by a human. Exactly.

SPEAKER_01 37:27

Um, and then you know, to your point about you know AI plus humans, uh you know, I think it's anytime, and we can maybe wrap up on this if if we're done. Anytime AI impacts the health, the wealth, or the livelihood of a human, a human needs to be involved in that conversation. And so you can't, you shouldn't drive a decision from AI without a human being involved if it impacts one of those three areas.

SPEAKER_00 37:51

See, you just keep dropping these little pearls, and so now I can't end it there because um one of the questions really relates to predictions for the future, either what would what it would be, and there we will wrap up. What are your predictions for the impact of AI in the future? It can be related to hearing aids, or the other part that you just mentioned that triggered this for me is the future of AI and the workforce. You know, when there are jobs that AI can do more accurately, like you just raised with radiology, on that part, that's going to eliminate potentially uh or or change a workforce. Uh is this uh are we looking at something in the future where we all have universal basic income and AI is doing the hard work on some of those repetitive tasks? What what what's your prediction for the future in hearing aids of what you want and for AI and the workforce?

SPEAKER_01 38:44

Yeah, so so I'll start with AI and the workforce. So net AI is gonna create you more jobs, right? So the total number of jobs that'll exist, I think, are gonna go up because of AI. Though a lot a lot of that's gonna be new jobs. The thing we need to be mindful of as a society and especially as employers, is that there are also gonna be a significant amount of jobs that are gonna go away. And it is our obligation as employers to make sure that these people who's the two things that we're very upfront with them if we're bringing in an AI to replace them, and that we are retraining them or upskilling them so that they can participate in the economy of the future. Um, and their children can participate in the economy of the future because you typically you see, especially in in blue-collar jobs, low lower skill jobs, children tend to do the same things that their parents do. And so we need to make sure that these people participate in society, are able to participate. No, no one wants to not have a job, right? For the most part. People want to be productive. Uh, so we're not going to turn into WALL-E as long as um long as employers do their part. And that is an obligation in my mind that they need to do. So that's that's my my short answer of that. We could have an hour-long podcast on that conversation alone. Um, in terms of what, you know, what I think there's two different things, what I would predict for AI for for hearing aids, and what I'd like to see. Um, mine's probably a little bit, you know, kind of bespoke top desire because of all the traveling I do. I'd like my hearing aids to translate, you know, within you know half a second in my hearing aids without needing my phone, uh, even if I have to preload a certain language. So that's my that's my uh my number one uh request.

SPEAKER_00 40:38

Noted.

SPEAKER_01 40:39

Um you've heard it before. Um you know, I think in terms of the technology, you know, they do a really good job now. I think um, you know, I think getting a smaller form, the smaller form factors to have similar battery life would be would be really good. Um I think um, you know, some some things that are nearer term is like, you know, now on Apple devices, you can use them on your, I can use connect my hearing aids to my Mac or my iPad, but I need to do it manually, right? It doesn't just like AirPosit just automatically goes across. And this is maybe more of a Mac thing than it is a Starky thing. But uh so something like that where it's more seamless to your point about not even knowing that I'm using them. And I and I suspect more, I predict more stuff like that, where hearing aids become less and less of something that you think about and just more and more a part of your your life that you don't even, you know, like something else you don't even think about, like your heart beating. Yeah, right. Um, and then the other thing is that would be really cool is, you know, again, most people probably aren't aware of this, but one of the best places to take uh, you know, kind of medical information, health information about a human is in your ear.

SPEAKER_00 41:57

Yes, right?

SPEAKER_01 41:58

Blood pressure, heart rate, respiration rate, things like that. So more of that stuff in the hearing aids, and I know you're smiling because you guys are working on this stuff. Um, but then also, like my dad has Parkinson's, uh, and during COVID and things like that, a lot of medical, uh, you know, a lot of diseases or conditions can be predicted better from a device from your ear than they can anywhere else, uh, because they hear you. They, you know, a lot of these things, Parkinson's and COVID and Alzheimer's and other diseases, you can tell by changes in voice patterns that someone's coming down with them much sooner than you could with any other. So, so really more getting more since you're already a class two medical device, getting more into the you know, predictive and and you know, pre, you know, getting earlier interventions because the earlier you're getting an intervention for a lot of these things, the better your long-term outcome is going to be.

SPEAKER_00 42:51

Music to my ears. And uh, we're gonna have to have you back sometime to talk about the health and wellness aspect, or a cast, which is coming and just right around the corner, that will enable that more seamless transition out of Bluetooth low energy as you move from your computer to your phone, even you know, to the hearing aids alone, as you say, with uh your wish for translation and can't wait to talk to you more about that and demonstrate some more of that to you as um as we uh deliver it to the market. So thank you so much, Seth. I've thoroughly enjoyed our conversation and appreciate so much you took the time today. To our listeners. Uh thank you for listening to this episode of Starky Soundbites. And if you enjoyed this uh conversation, please rate and review us and like us. Um share it with your network, your family, your friends, anyone who might be interested in uh Seth's background with AI and also the perspective of someone who wears hearing aids. We'd also like to know what's on your mind. If you have questions for our hearing experts or have ideas for future episodes, send us an email at Starky Soundbites at Starkey.com. We'll be featuring your questions and getting some answers. From our experts on future episodes. We'll see and hear you again very soon, and thanks again, Seth.