This week I am talking to Chip Steiner, Product Manager Healthcare Practice for Kore.ai. Chip has been bringing speech enablement to healthcare for a long time and in the very early days managed to get the Director of The Fantastic Voyage, Richard Fleischer to lecture to his team.
We talk about the history of speech enablement and dive into some of the tools and techniques that are making this technology increasingly close to working the way we see it presented in Hollywood. One of those areas includes a consensus-style model of AI that uses 3 different engines to reach the best conclusion on what is meant in a conversation which reminds me of the precog, Agatha Lively from Minority Report
Listen in to hear our discussion where we see progress in healthcare and what the future opportunity is for using this technology
Listen live at 4:00 AM, 12:00 Noon, or 8:00 PM ET, Monday through Friday for the next week at HealthcareNOW Radio. After that, you can listen on demand (See podcast information below.) Join the conversation on Twitter at #TheIncrementalist.
Listen along on HealthcareNowRadio or on SoundCloud
Raw Transcript
Nick van Terheyden
Today I’m delighted to welcome chip Steiner. He is the product manager for healthcare at core AI Chip. Thanks for joining me today.
Chip Steiner
Oh, great to be here today. And thanks for the invitation.
Nick van Terheyden
So if you would help the audience understand your background and how you reached this point in your career, if you would.
Chip Steiner
Certainly Thank you, Nick. Well, I began healthcare in the operating room, I was tasked with trying to design a technology that would help surgeons navigate the human body. But that was over 30 years ago. Today, I focus on the use of artificial intelligence to advance conversational AI.
Nick van Terheyden
So when you say that, and you say 30 years ago, I’m, I’m thinking of the movie where they shrunk people down and navigating. This wasn’t Hollywood, this was for real, right?
Chip Steiner
This was most definitely for real. And I have to say, as an aside, I have met Richard Fleischer, the director of that film, I invited him to my healthcare institution, back back in the day, and he took us through a journey that many in our audience were mesmerized with, it was fascinating to learn how Hollywood and modern healthcare have merged, have morphed together. And I can tell you, there’s a lot of value in looking across disciplines as you try to advance any field of study.
Nick van Terheyden
Yeah, I always look to Hollywood, and in particular science fiction, I’ve got to say, as, you know, a predictor of the future, I think it was Isaac Asimov that famously said that, you know, difference between, you know, magic and, you know, reality is just a question of time in many instances, and I think that’s been true. And, you know, one of my favorite, you know, I’m a sci fi geek, let’s be clear, and I’m a Star Trek fan. And Star Trek, I think I’d still believe this is true. I’ve certainly said it enough time. So it must be true, was the original presenter of speech as an enabling technology? In fact, I even know the episode, it was all to do with pie and spark asking the computer to calculate pie that eventually managed to reach control back. But we’ve advanced a lot further and you make an interesting point that they’ve marched in lockstep, I feel like Hollywood’s a little bit further out in terms of what they’re showing or talking about, where’s the gap? Where do you see this at the moment?
Chip Steiner
Well, you know, I think the gap is closing every day. But quite frankly, I think it’s a matter of what is what is Hollywood trying to entice the audience with healthcare, from my vantage point, a little more practical, we’ve got patient care, we want to of course, be living and operating and delivering healthcare on the edge, because that’s where innovation and advancements occur. But there’s no question that the ability to look towards the future is really the the goal of advancing life and advancing healthy life, which is a common goal in the healthcare realm, as well as I think, in Hollywood, if you will. There are a few other industries that are behind both of those. But I agree with you, Hollywood does have an advantage in terms of what is presented on camera. Healthcare has maybe the disadvantage, that it’s a conversation between a patient and a healthcare provider. And that’s where the real reality sits.
Nick van Terheyden
Yeah, I like the idea that that gap is closing. But, you know, my my challenge to Hollywood is, you know, get your thinking caps on we need some more ideas. Let’s be frank. But let’s talk about the reality because I think that’s important. You know, certainly, based on my experience, you know, speech is the most natural form of communication. It’s the one that everybody has, you know, if they have nothing else if they can’t write or read, they can definitely communicate by speech. So it’s sort of foundational. Yet it’s not the sort of or it hasn’t been the predominant thing. means of interacting with technology, which has essentially overtaken our lives very clearly. But it’s starting to impact and improve things with speech. I see more of it. Where are we with that? And where do you see that? You know, currently?
Chip Steiner
Well, first, I completely agree that speech is the penultimate user interface between human and machine. It has lagged. It was fundamentally hard. We may as evolving humans believe that speech may have been the primary way of conversing with others or with eventually machines. But it’s it’s it’s tech it has has its technological challenges. Where we are today is that it is ever present in many of the advanced healthcare applications, between provider patient, caregiver, and a machine, we use it in our solutions to try to increase operational efficiency, in addition to care care excellence, and I’d be happy to give some examples as we talk further.
Nick van Terheyden
Yeah, so I think that’s true. We’re seeing more of it. But, you know, historically, we’ve seen these sort of challenges. And, you know, it’s a little bit of a sort of fits and starts. And, you know, certainly going back in my career, it was over promised and under delivered, you know, we took the technology, we had their souped up device that was sitting underneath the counter that had multiple cards, as much memory as you could cram in at the time, and it worked. But when you put it out into the field with, you know, basic technology, it didn’t work quite as well. And I think we disappointed and we’ve had to recover. At this point. It’s, I think native I mean, everybody talks about a lax a and, you know, that’s most of us have to do that, without activating things in our own house. And, you know, the other versions of this. Where have you seen the most progress? And what are the opportunities? Do you have some examples?
Chip Steiner
I do. Thanks. Thanks, Nick. And, and I was going to break down for you because at heart, I’m an engineer. And that’s my, my daily perspective, when I tried to solve problems. The group at core, where I’m currently Product Manager of our healthcare practice, we follow a parallel processing approach. We call it our proprietary natural language understanding, but it allows us to simultaneously evaluate the conversation between a human and and up in a in a base interface and look for components of that conversation that are most applicable in a context. So let me let me give you a specific example. Well, we’ve often thought that the, when the space shuttle or any, any rocket goes up, there are computers, calculating launch angles, and an attitude and trying to figure things out, and ultimately reporting to some human management system, what they’re finding. But as I think we’ve also learned, they use multiple machines to simultaneously calculate results. And then they have some management system to evaluate and compare those results. We’re doing the same in conversational AI, we have three simultaneous natural language engines that are looking and trying to divine what the fundamental meeting was, what the, the the contextual knowledge was, and try to report back what the most likely intent was of any specific phrase. And we do that in a way that allows us to rank and then ultimately determine the highest probability of the correct intent. This methodology, Nick allows our solution to have a more effective response and conversation, I would argue that the human brain works in a similar way, not just the left brain, right brain approach, but there’s a multiplicity of cognitive processes that are going on in our brains. As we listen. We’ve got certainly two ears and for most of us, and those are feeding parallel streams of data. But we’ve also got other senses that our body is able to feed to that cortex and ultimately derive what the likely intent was of The phrase from the other individual or individuals we’re talking with, the approach at core is to do the same thing. We have this in a patented process. And we use that to more effectively derive the intent of the conversant. And that’s what we’re doing how that applies to healthcare is that then we can carry on a more natural conversation that’s applying the context of the conversation, you and I might be chatting about motorbikes or skiing, or riding lawnmowers or politics. But in healthcare, we might be talking about the diagnosis or the treatment plan, or your medical history, or the payment restrictions or the approval process, that context is relevant. And our ability to look at the phrases and divine the intent of the conversant in a parallel way, gives us the most effective outcome and solution. And that’s what really sets our approach apart from the rest.
Nick van Terheyden
So I’m afraid we’re back to Hollywood. But what I heard and I want to sort of make sure that people understand this is it sounds a little bit like Minority Report, in terms of an assessment of results based on in your case engines. But you know, in minority reports, it was three individuals that said guilty or not all, you know, but you get the principal, and it sounds like that’s what you’re doing. Is that a fair assessment?
Chip Steiner
That is a fair assessment. And, of course, this, this allows us to differentiate and perhaps if we don’t achieve a, a high enough confidence interval, if you will, we have the ability then to adapt the conversation and ask, ask the conversant a relevant question. That further refines the approach. The further refines the conversation. And that’s, that’s a separate thing that maybe you don’t have in the current al e x a paradigm. But we have when you’re in a conversational AI mode, you can dialogue with someone, and actually, that that’s the same thing you and I would do if we’re trying to understand each other.
Nick van Terheyden
So let me ask I mean, you know, what sounds important here is the fact that you’ve got, you know, varying opinions. You know, that I start with a question Why Why stop at three? Why not have more is that, you know, maybe that’s a processing power issue. But I also want to know, who’s Agatha in this? You know, who’s the dissenter? And, you know, is the son value to the dissenter, in terms of, you know, giving them more authority in the ultimate resolution of what somebody said?
Chip Steiner
It can be and and let me offer, the the current configuration of three is more about three different context trains of thought, rather than any type of processing limitation today, of course, more processing is always valuable. If you can stage it properly. No, we we want to the approach we take wants to make sure that if you’re involved in a, we’ll call it technical conversation between healthcare providers that you apply more influence from what I’ll, I’ll say is an ontology based nomenclature. If you’re having a conversation between a patient, perhaps, and a nurse, you’re going to leave more, you’re going to apply more weight, if you will, to natural context conversations about the status of a family member, the status of their environment. So we are able to tune if you will, that the the influence of these various analyses, and use that to better derive the context. And this is a dynamic process.
Nick van Terheyden
So for those of you just joining, I’m Dr. Nick the incrementalist today. I’m talking to chip Steiner. He’s the product manager in healthcare at core AI. We were just talking about the process for sort of reaching better accuracy, better resolution of what people are saying I think, you know, most folks have some interaction with voice agents, let’s call them you know, as a general term, obviously, in healthcare, there’s some specialization And that experience can vary. There’s, you know, a very sort of staccato, I think I like to describe the frustrating ones as calling into phone systems that are very poorly designed and, you know, essentially force you through a whole process that you might as well press buttons for. And then there’s really good conversational AI that will pick up on topics and jump you. I’ve heard that on some of the phone systems. In healthcare, we’ve got a sort of narrower focus, you know, potentially there’s more capacity to be better at it. Is that true? Are we seeing some of that experience?
Chip Steiner
Well, we are and what I wanted to make sure we also focus on is how this how this tool, this conversational tool is better positioned to improve efficiency of operational delivery in healthcare. So it’s not always about perhaps, a coming up with a diagnosis for a particular condition. But more about can I ensure that the right provider that is most adept at dealing with a condition is interacting with a patient that is experiencing or perhaps suffering from a condition, it’s being able to focus on some of those operational challenges that many of many in the healthcare industry all around the globe, but very pointedly here in North America, are challenged with not enough supply of providers, not not the right alignment between need, and, and delivery of capabilities. And conversational AI allows you to extract those challenges in a way that would otherwise lead to inefficiencies. And that’s really an area that we focus on. Nick,
Nick van Terheyden
so as you think about the landscape today, you’ve got sort of a deep understanding of the healthcare space, you’ve had interactions in different domains. Obviously, we’ve got the generalized experience that most people can relate to. But where do you see this being most successful? Can you give us some examples of you know, how we’ve managed to really deliver on the promise that, you know, I saw in the 60s with Spock, talking to the enterprise computer?
Chip Steiner
Well, and I’ll and I’ll remind all of this, that the tricorder was one one utopian device in many respects, that allowed a Spock to divine or bones to divine What the what, what the challenge was that one of his crewmates was suffering from, but
Nick van Terheyden
just as an important point on the tricorder, since you bring it up at night, you know, what’s your thought, but that was actually the subject of one of the million dollar prizes. And I ultimately, still struggled. I think we’re not quite there yet, unfortunately. But we got closer with the tricorder prize,
Chip Steiner
we certainly have. And you and I may be a similar alignment and where we hope the future leads us. conversing with a tricorder. It wasn’t necessarily fully fully revealed. But voice and human interaction was also something that that was dealing with. Where we are today is that because we can determine the intent of a patient, the intent of a caregiver, in a healthcare setting, or the intents revealed through a conversation between caregiver and patient, we are better able to apply the limited healthcare resources to the condition of need. And so as an example, oftentimes you hear about patients not being able to get in a timely manner, to speak with their specific with their doctor, be able to get in and get the care they need. And that’s really about a misalignment of healthcare resources that we can overcome. If you just have a conversation about what the patient is, is dealing with or a conversation with a collection of providers to talk about their capabilities and get those aligned properly. Our solution at core, you leverages the tools that conversational AI present to be able to address those supply and demand challenges that exist in healthcare today. That’s where we’re seeing the biggest impact back to your earlier question.
Nick van Terheyden
So Um, as you think about the future, you know, obviously we’ve made significant progress. There’s, you know, tremendous opportunity to apply this. Where do you see this going?
Chip Steiner
Yeah, so conversational AI will continue to be improve, I think we can all we can all imagine that, where where I see the biggest impacts in the near future, are the ability to apply multiple conversations, multiple utterances from a group of Converse since right now we’re somewhat limited to one on one dialogue, just because of the the ability to process all of those conversations, I see bringing in second, third, fourth parties to a group conversation, and being able to segregate and apply the valuable aspects of that. And then add to that video, I would consider conversational AI to include human nonverbal cues to the conversation. And we’re seeing that in some of our collaborations in our organization, where being able to express empathy, and derive the feeling that the emotional context to a conversation between provider and patient between caregiver and family member allows the conversation to be richer, and ultimately, to have a a more impactful outcome, which, which will drive operations as well.
Nick van Terheyden
So, earlier, you when we were talking, you mentioned something that sort of stood out to me, and I want you to sort of expand on this, you said voice is the penultimate interface that we’re dealing with. Clearly, there’s something else in your mind, tell us what you meant.
Chip Steiner
So when I was referring to voice, it was a human using all of their conversant interfaces with a machine, I meant that rather than some direct, digital conversational approach, ie text, ie, even even Morse coded approach to conversation with a machine know the human dynamic that humans on this on this planet leverage on a daily basis to convey their feelings and interpret someone else’s feelings. That is, that is being able to interface that in a direct way to silicon, to a, a process engine that is able to handle all of that, that is truly the penultimate, today we are using the audio signal, we’re going to layer onto that a video signal, we’re going to layer on to that other physiological parameters of a patient and a caregiver to improve that conversation.
Nick van Terheyden
Yeah, so I think interesting, you know, as I think about this, the potential future for sort of integration at a more basic level, you know, the thought process, we’ve certainly seen some progress with that, obviously, critical in terms of those that have been, you know, impacted by disease sometimes, but mostly through trauma, where they, you know, are limited in their capacity. And we’ve seen some progress, but it’s been challenging. I think, famously, there was at least some work on the part of Elon Musk and his organization to look at a brain interface. But you know, maybe that lies in our future. I think, you know, in the same way that we were afraid of speech recognition, maybe we’re afraid of that, but you know, it offers potential, obviously with the appropriate surroundings. Unfortunately, as we do each week, we’ve run out of time, so it just remains for me to thank you for being on the show, Chip and say, live long and prosper.
Chip Steiner
Thank you, Nick. It was a pleasure. I hope to speak to you in the future.
