The Exchanges

Q I read that you got an undergraduate degree in electrical and electronics engineering, and it seems like you could have taken a path towards robotics, right? Especially inspired by what happened to your uncle. Um, was that a path that you were potentially pursuing?

A Yeah, definitely. So, uh, when I was studying electrical engineering, I worked with the prosthetics teams at the University of Washington. And then, you know, I was like, I really wanted to create the brain into these prosthetic limbs. So I went to grad school and I started working with brain computer interfaces more. And that's when I worked with people with ALS and children's cerebral palsy. And I was like, wow, this is, this is just a whole nother level. I mean, as, as terrible as it is, for example, my uncle having to go through this experience and having prosthetics that don't work with him naturally. What if you can't even move your eyes, right? Like that's, that's just a whole different level of need. Uh, and so that's really what pivoted me from something that I thought I was going to dedicate my life to robotics to something much broader.

AI assessment note: “Yeah, definitely. So, uh, when I was studying electrical engineering, I worked with the prosthetics”

Q give me some data about my sleep, it'll tell me my heart rate, blood oxygen. So, I mean, you know, given what we can already gather from just our heart rate, right, which is quite a bit of data, Beyond all the things we talked about, what other things potentially could we learn about our health or, um, you know, our general state from these devices, from these, these brainwaves?

A Yeah, you know, what's really interesting is that brain detecting devices are actually the ultimate wearable. A lot of the devices that you wear right now, for example, accelerometers for movement or for heart rate, they can either be picked up through Brain data, or they originally come from the brain, and those are just secondary sources of signals, right? So for example, you can actually pick up Parkinson's responses using the Apple Watch. The issue is that by the time you pick it up at the hands, um, or through walking metrics, your brain's already been dealing with it for the past 10 years. And so with the brain, you can actually pick up a lot of those things earlier. So there's two parts. One is that brain-based wearables are going to replace All the other wearables that you have. That's step one. And so all that data and all that value that we're seeing with existing wearables are going to be all consolidated into one device. But then two, there's certain things that you can only pick up from the brain. You know, for example, traumatic brain injury information, tracking ALS, right, seizure detection. There's so many other things that you can only do with the brain that, you know, not only are you taking care of your previous wearables, but now you're adding a whole plethora of medical Use cases that have already been tested out in scientific literature, but now are able …

AI assessment note: “traumatic brain injury information, tracking ALS, right, seizure detection.”

Q But how do, how do headphones, how do sensors around your ear capture as accurate information as, as those other sensors?

A Yeah, I guess to answer that we have to break it down into two steps. First is like, how do we capture those signals from headphones? Well, the brain is, is very conductive, right? So, for example, one of the brain signals we look at, uh, is called P 300. We don't really have to get into the details of it, but essentially that comes from an area of your brain called the parietal lobe. It's, it's around the back of your head. Uh, and even though that signal comes from around the back of your head, The signal is such a strong response, it can actually go, it goes all over your brain. Only the farther it goes from the signal source, the smaller that signal becomes, so it becomes harder to read. So we know that these signals go across the head, but you lose the ability to record them easily. And so that's where our AI comes in. It picks up these signals, even though they don't come from the most perfect location, they come from the areas that headphones are at. And from there, we're able to boost those signals to a level that makes them usable for different applications.

AI assessment note: “that's where our AI comes in... able to boost those signals”

Q Nodes are, that are attached to, to people's heads, you know, uh, that track brain waves. But what were you trying to, to solve with, with this research and this concept?

A Yeah, definitely. So, um, essentially the work that I did as my graduate student work is I was working with children who had severe cerebral palsy, and the issue there is that we were not able to essentially give them these tests that they needed to be allowed to get physical rehabilitation, and the reason for that is because they couldn't communicate, at least not in the traditional means like talking or pointing, and so we used brain-computer interfaces to solve that problem, and the blocker there was that You know, you have this eight-year-old kid. He just got 20 minutes worth of setup with goop and gel in his hair to make the technology work. There would be about 10 minutes worth of, like, calibrating the system, and then it would take sometimes between one to five minutes to, like, even get a response from them. And so the work that I did was essentially machine learning classification so that we could interpret their brain activity at a much higher level of accuracy. And what that enabled us to do is reduce The response time from one to five minutes to anywhere between 30 seconds to a minute, which is enormous for an eight-year-old kid, because, you know, having them sit there for five minutes to say, you know, yes or no question is, is, is crazy.

AI assessment note: “I was working with children who had severe cerebral palsy”

Q Alright, so let's talk about these headphones, the Enten headphones that your team has been working on and is getting ready to release later this year. Um, what, what are you able to, to track using, you know, putting these headphones on, on people now? What, what can you actually find out about?

A Yeah, so there's kind of like three areas where we use the technology in right now. Uh, on the first end is, for example, understanding an individual's focus over time when they're fatiguing, when they should be taking a break in order to maintain, uh, It, your, your brain is kind of like your body is to dehydration. You should be, in the case of the body, drinking water throughout the day. Well, you should be taking breaks throughout the day too. Even though you may not feel thirsty or you feel tired, you should be doing that in order to maintain a high level of hygiene for your own work and life balance. And so that's, that's kind of the first area. The second one is in control. So we have the ability to, for example, use brain activity to do Very minimal controls on the hardware as well too. So changing music tracks, play and pausing music. And then on the third end is there's so many incredible biomarkers that can be picked up using this type of technology. Like I said, tracking Alzheimer's or cognitive decline or, you know, other types of biometrics. So essentially just like how the Apple watch started out as a system for tracking your movement, um, Now it can actually pick up like heart arrhythmias. And so all of that medical landscape and biomarkers are also available through these brain computer interfaces.

AI assessment note: “understanding an individual's focus over time when they're fatiguing, when they should be taking”

Q them in like, you know, modern day, some modern day city and just saw people talking to themselves, they wouldn't understand what's going on. In 10 years from now, are we, I mean, are we likely to see a version of that except in silence, like people maybe having conversations With other people through their, you know, their device, their earbuds, or whatever it might be, but just in silence?

A I mean, I wouldn't say 10 years from now, right? I think it's going to take longer, but what I would say is that, you know, within the next 10 years, people are going to be communicating to their technology silently. You know, I think that communicating via voice is still going to be a more efficient method of communication with somebody, at least in the near term. But when it comes to, for example, let's say you're having a conversation with somebody, and you get a notification, right? Being able to push it out of the way, or to reply to it real quick in a way that doesn't break the conversation would be very valuable, right? Let's say, you know, for example, you are talking to somebody about a really great place that you went to go eat to, but you forgot the name, right? So helping it pull up that information in a way that doesn't disrupt, oh, hold on, let me pull out my phone, and, you know, Just wait a second while I figure everything out. So I think that we're going to be communicating with our technology seamlessly and invisibly. And then that enables us to also free up, um, some of our cognitive loads that we can continue to have these more engaged and connected conversations in the way that we traditionally do.

AI assessment note: “I wouldn't say 10 years from now, right? I think it's going to take longer”

Q All right. Help me understand what the technology is that you're talking about. Well, you said it would take, you know, five to 10 minutes for setup, and what was it that you were doing differently?

A Yeah, so the reason the setup takes so long is because, uh, getting brain data is incredibly difficult. There's, there's a lot of noise in the environment, even You're blinking, you know, talking, the electromagnetic noise, even the lights impact the ability to collect this data because the sensors are so sensitive and brain signals are so small. And so we essentially developed an artificial intelligence that was able to use brain data that we previously collected and then also brain data that was coming in and essentially increase the signal to noise in order to essentially be able to do classifications of what a person intended To do in this case, the child was selecting a multiple choice question, uh, at a greater fidelity. And when you can do that at greater fidelity, you don't need to repeat the question numerous times. And, uh, through less repetition, you know, it gave people a better user experience. And so being able to make that brain computer interface work in a seamless way really unblocks a lot of its use cases.

AI assessment note: “we essentially developed an artificial intelligence that was able to use brain data”

Q It essentially is recommending when you need to take a break, right? When you're working, but what is that based on? What, what kind of data is it getting to suggest that you need to stop working?

A Yeah, so we did a large study with a, uh, a professor from, from Harvard, who's now a professor at Worcester Political, Technical Institute, and he had created this incredible method for identifying the individual's focus, and so what we were able to do, and we did close to a thousand individuals worth of data collection on this, is we tracked people just doing their work, and then leveraging this algorithm that we co-worked with this individual at, uh, at Harvard, and, uh, Uh, what we saw is that there was very clear breaks in, in the data where we could see that if we were to recommend them a break at this time point, it enabled them to have three to four hours of higher productivity afterwards, feel more refreshed, reduce their errors in the amount of work that they do, uh, instead of just burning themselves out and feeling really, like, uh, you know, bad about their day, essentially.

AI assessment note: “we tracked people just doing their work, and then leveraging this algorithm”