Selected Podcast
Artificial Intelligence in Healthcare: What is the Future Beyond ChatGPT?
Dr. Anthony Chang discusses the role that Artificial Intelligence plays in Healthcare.
Featuring:
Dr. Chang completed his Master of Science (MS) degree in Data Science with a sub-specialization in artificial intelligence from Stanford School of Medicine and completed a certification on artificial intelligence from MIT. He earned his MBA in Health Care Administration from the University of Miami School of Business and his MPH in Health Care Policy from the Jonathan Fielding School of Public Health of the University of California, Los Angeles
Anthony Chang, MD, MPH, MBA, MS
Anthony Chang, MD, MBA, MPH, is a pediatric cardiologist and medical director of the Heart Failure Program at Children’s Hospital of Orange County (CHOC). Dr. Chang also serves as Chief Intelligence and Innovation Officer of the Sharon Disney Lund Medical Intelligence, Information, Investigation and Innovation Institute (Mi4) at CHOC. The institute is dedicated to implementing data science and artificial intelligence in medicine and is the first of its kind in a hospital.Dr. Chang completed his Master of Science (MS) degree in Data Science with a sub-specialization in artificial intelligence from Stanford School of Medicine and completed a certification on artificial intelligence from MIT. He earned his MBA in Health Care Administration from the University of Miami School of Business and his MPH in Health Care Policy from the Jonathan Fielding School of Public Health of the University of California, Los Angeles
Transcription:
Melanie Cole, MS (Host): From new diagnostic tools and technology to advances in rare disease research, welcome to Pediatrica, a pediatric research and innovation podcast presented by clinicians and researchers at Children's Health of Orange County. I'm Melanie Cole. And joining me today is Dr. Anthony Chang. He's the Chief Intelligence and Innovation Officer at Children's Health of Orange County, and he's here today to highlight artificial intelligence in healthcare. What is the future beyond ChatGPT?
Dr. Chang, it's a pleasure to have you join us today. What a topic we've got here. I'd like to start by asking you about AI specifically. You know, people not in the media are hearing words like artificial intelligence, deep learning, machine learning. Tell us what that means for the purpose of this discussion.
Dr Anthony Chang: Thank you, Melanie, for having me. So, artificial intelligence, I think recently obviously fueled by the large language model called ChatGPT, and it's many, many variations is basically allowing computers and software to do what takes intelligence by humans to do. So, that's a very simplistic, but I think very spot on definition. That actually initially was put together by Marvin Minsky, one of the grandfathers of machine and artificial intelligence.
Melanie Cole, MS (Host): Yeah, I think that was an excellent description, because it really does speak to what it aims to mimic. So, what's the current status of AI in general? And then, we're going to get into specifics in healthcare.
Dr Anthony Chang: Since the 1990s, when some of us recall the IBM supercomputer beating the world's chess champion, Gary Kasparov, it's really been on the upswing and particularly in the last five to 10 years, thanks to the convergence of several elements. One is the advanced algorithms out of Toronto. And then secondly, the availability of cloud computing and storage. And thirdly, the advances in the gaming industry with fast processing chips, particularly GPUs. So, that, along with now a lot of healthcare data, particularly in imaging and genomic sequencing, have now contributed to this sort of Cambrian explosion of possibilities within medicine.
So obviously, initially, it was deep learning, looking at medical images. And now, in the last six months or so with language models like ChatGPT, so it's all kind of converging together. And you can think of this as, you know, if you call artificial intelligence the orchestra, now you have different sections of the orchestra being assembled to hopefully finally make a big impact in healthcare that we've been anticipating.
Melanie Cole, MS (Host): This is absolutely fascinating. So, I'd like you to speak to the relative successes for AI in healthcare. But as you mentioned imaging, I also think of risk assessment, ECG monitoring. In what ways do you see it innovating in the AI space as it relates to healthcare and these techniques that have been applied in medicine, the things that they've sought to explore that can help improve quality in patient care? Tell us what you've seen as the main successes. Where do you think it's going to do the most help?
Dr Anthony Chang: Yeah. So, I had given a few keynotes recently reflecting on the past decade. So if we look at the modern era of adopting AI in healthcare, I can go back to maybe about 2012 when medical images were being interpreted with deep learning. Deep learning is a kind of machine learning, which is using computers to learn from data and deep learning is a special kind of machine learning in which the methodology involves something called a neural network, which is mimicking how the brain works. So, it's kind of a more sophisticated way of doing machine learning. So, think of deep learning and machine learning as sort of the string section in this orchestra, perhaps deep learning is the violas and machine learning is the violins. And it's just really starting to make beautiful music for the first time, I think, in a decade.
So if you look back on the past decade, Melanie, I think the three successes are a little bit more out in front with other areas to follow very quickly. I would say number one is medical image interpretation using deep learning. So, medical images, including moving images like ultrasound can be interpreted with greater accuracy than even groups of subspecialists. So, that's number one.
Number two is the rapid advances in looking at protein structures just based on genomic sequencing. Something that used to take months and years. I was a molecular biologist before and it would just take a long time. And now, it's a matter of hours that we can figure out what a protein will look like. And obviously, it was big for vaccine design for the recent pandemic.
And then, lastly, I think using primitive AI tools to help with the healthcare administration burden that every health system faces has been really good at cutting cost and increasing accuracy in all the administrative aspects of healthcare. Perhaps the less intellectually stimulating, but nevertheless very impactful in terms of saving money. So those I, think, are the top three with another tier of advances that are very quick to follow like ChatGPT and large language models to help with the many, many communication challenges in the healthcare system.
Melanie Cole, MS (Host): To me, Dr. Chang, the algorithm, I mean, I hear that word, I think of things like TikTok and, you know, that sends you what you have shown that you like, and the public has a certain perception. Obviously, we have to mention the elephant in the room here. How do you overcome the perception that this is possibly bias? There's privacy issues within the healthcare community. What do you think patients would feel about using this kind of technology in their medical care? Will they even notice?
Dr Anthony Chang: I think artificial intelligence and healthcare is already here, perhaps not widely publicized. Even AI tools that really don't require humans to have oversight, which is the kind of AI, it's called autonomous AI, that really intimidates everyone. So, I think as long as the clinicians can start to understand and know AI and be able to trust certain tools, I think we will start to see wider adoption in this coming decade.
I think to balance the discussion on whether or not AI is ready or ethical or scary to people, I think we also have to realize that not using AI in healthcare is perhaps costing lives also. I think there's some mature tools that are already available, for instance, to screen head MRIs or brain MRIs for bleeding or stroke. Right now, an AI tool can expedite the diagnosis, so that the head MRI can go to the front of the line to be looked at by and validated by a radiologist. Now, if you can cut down the time to intervention for a stroke, that's definitely an improvement in the quality of care for a patient. So, those are kind of tools that are already available that are already being adopted by some hospitals. And perhaps in addition to think about the potential downside of AI in healthcare, we should also think about the downside of not deploying AI in healthcare, which is something I always try to remind everyone of.
Melanie Cole, MS (Host): And what a great point you just made because there is a downside to not employing this type of technology while we have it available. And because medical advancements are moving so quickly, we have to move with them. Do you see any challenges disappointments? I mean, we obviously just mentioned public perception, but what about any other barriers to implementing these technologies?
Dr Anthony Chang: Well, if we look at the ChatPGT and large language models as kind of a milestone period in artificial intelligence and healthcare, I think people are afraid that it is not accurate or it's going to give false information to mislead clinicians and patients, but I think we can work around those issues. I think we can curate the answer so that they become more accurate. I think part of the issue with clinicians, sort of low morale, is not having enough time to spend with patients because of the burden of having to look at volumes of electronic records, particularly for new patients, and also the information that's out there now. Medical knowledge is doubling every 30 some days now as opposed to every few years, just a couple of decades ago. So, the amount of information from patients records and the amount of information that's published in the medical literature is making it humanly just not possible to keep up. And I think tools like ChatGPT can start impact in helping the clinicians to practice better and practice higher level of wellbeing, which is, I think, ultimately one of the quintuple aims that we always think about in healthcare and medicine, which is improving quality of care and improving the patient experience and improving the caretaker experience at a time that we need, I think, an equalizer. I think AI is a great equalizer for this period of great challenge in healthcare.
Melanie Cole, MS (Host): We've mentioned it a few times, Dr. Chang, you've talked about ChatGPT. Now not everybody knows what that is, and I'd like you to speak about its role in healthcare, whether we're talking about virtual assistance for patient care, booking appointments, which to be honest, we've been doing automated for quite a while there. You know, you can do it online or you can do it even on the phone. Push this button, this button, when do you want your appointment? But this is a little bit different. I mean, because it can possibly even, what, answer medical questions, provide information. Tell us a little bit about ChatGPT and its role in healthcare.
Dr Anthony Chang: Okay. Relatively technical terms, ChatGPT is an example of a large language model, which is based on a technology called a transformer. So, a transformer for the first time is an AI sort of tool that can have the capability of something called attention, which makes it really flexible and nimble to understand how we humans communicate. So rather than just predicting the next word, like you have in autocomplete feature, it is able to complete sentences and paragraphs. It's like autocomplete on massive steroids, so by taking everything on the internet that it can have access to as a corpus or as a body of knowledge and be able to answer questions. So obviously, it is vulnerable if important information is not captured. It's vulnerable if there is misinformation on the internet, the classic garbage and garbage out. So, we have to be careful about how to use those really advanced level of information.
So, I get often the question, which is a great question, "Is ChatGPT information or is it knowledge?" And I think knowledge being something that you can act on based on the information. I think it's what I call superinformation. It's information packaged in a way that is useful and practical and perhaps even something you use to base your decision on. But I think, for it to get to knowledge, we still need the humans in the loop to help make best use of that super information. But they're certainly capable of doing very helpful tasks, like perhaps take a 300-page, medical summary and information from a patient and maybe make it very, very concise and short, so that the clinician can quickly gather the information without having to go through it all.
Melanie Cole, MS (Host): That's so interesting, isn't it? It's an exciting time in medicine and specifically in AI in healthcare. You know, I'm really looking forward to seeing where this goes. I'd like you to wrap it up, Dr. Chang. What would you like other providers to take from this key message about AI and healthcare from this podcast today?
Dr Anthony Chang: Yes, I think it's easy to be a naysayer and not believe in this kind of technology. But I do think for our next generation's sake, and particularly the next generation of clinicians, and you know, to me it's just absolutely mind-boggling that at a time when clinicians can risk their lives in taking care of patients because of the pandemic, that we have more and more people going to healthcare than ever before. So, I think it's our responsibility, those of us who are still in medicine now, to create a very special sanctuary for medicine in the future by reducing the burden and increasing the outcomes and quality of care as well as our wellbeing by deploying the appropriate AI tools that are now available.
So, AI is here ahead of schedule at this level. So, I think it's our legacy to create, to adopt AI in a meaningful way, so that we can trust this as a resource for decades to come. And it's definitely an equalizer for some of the issues that we face, like health equity, you mentioned before. Perhaps we need our, you know, AI partners to help us to assure health equity for all because humans alone haven't been successful doing it entirely. So, we could use the resource more than ever before.
Melanie Cole, MS (Host): So interesting. Thank you so much, Dr. Chang. Really, you gave us such great information and a lot to think about. Thank you so much for joining us. To learn more about Children's Health of Orange County and to refer your patient, please visit choc.org/mi4 for more information.
Thank you so much for listening to Pediatrica, a pediatric research and innovation podcast presented by clinicians and researchers at Children's Health of Orange County. I'm Melanie Cole.
Melanie Cole, MS (Host): From new diagnostic tools and technology to advances in rare disease research, welcome to Pediatrica, a pediatric research and innovation podcast presented by clinicians and researchers at Children's Health of Orange County. I'm Melanie Cole. And joining me today is Dr. Anthony Chang. He's the Chief Intelligence and Innovation Officer at Children's Health of Orange County, and he's here today to highlight artificial intelligence in healthcare. What is the future beyond ChatGPT?
Dr. Chang, it's a pleasure to have you join us today. What a topic we've got here. I'd like to start by asking you about AI specifically. You know, people not in the media are hearing words like artificial intelligence, deep learning, machine learning. Tell us what that means for the purpose of this discussion.
Dr Anthony Chang: Thank you, Melanie, for having me. So, artificial intelligence, I think recently obviously fueled by the large language model called ChatGPT, and it's many, many variations is basically allowing computers and software to do what takes intelligence by humans to do. So, that's a very simplistic, but I think very spot on definition. That actually initially was put together by Marvin Minsky, one of the grandfathers of machine and artificial intelligence.
Melanie Cole, MS (Host): Yeah, I think that was an excellent description, because it really does speak to what it aims to mimic. So, what's the current status of AI in general? And then, we're going to get into specifics in healthcare.
Dr Anthony Chang: Since the 1990s, when some of us recall the IBM supercomputer beating the world's chess champion, Gary Kasparov, it's really been on the upswing and particularly in the last five to 10 years, thanks to the convergence of several elements. One is the advanced algorithms out of Toronto. And then secondly, the availability of cloud computing and storage. And thirdly, the advances in the gaming industry with fast processing chips, particularly GPUs. So, that, along with now a lot of healthcare data, particularly in imaging and genomic sequencing, have now contributed to this sort of Cambrian explosion of possibilities within medicine.
So obviously, initially, it was deep learning, looking at medical images. And now, in the last six months or so with language models like ChatGPT, so it's all kind of converging together. And you can think of this as, you know, if you call artificial intelligence the orchestra, now you have different sections of the orchestra being assembled to hopefully finally make a big impact in healthcare that we've been anticipating.
Melanie Cole, MS (Host): This is absolutely fascinating. So, I'd like you to speak to the relative successes for AI in healthcare. But as you mentioned imaging, I also think of risk assessment, ECG monitoring. In what ways do you see it innovating in the AI space as it relates to healthcare and these techniques that have been applied in medicine, the things that they've sought to explore that can help improve quality in patient care? Tell us what you've seen as the main successes. Where do you think it's going to do the most help?
Dr Anthony Chang: Yeah. So, I had given a few keynotes recently reflecting on the past decade. So if we look at the modern era of adopting AI in healthcare, I can go back to maybe about 2012 when medical images were being interpreted with deep learning. Deep learning is a kind of machine learning, which is using computers to learn from data and deep learning is a special kind of machine learning in which the methodology involves something called a neural network, which is mimicking how the brain works. So, it's kind of a more sophisticated way of doing machine learning. So, think of deep learning and machine learning as sort of the string section in this orchestra, perhaps deep learning is the violas and machine learning is the violins. And it's just really starting to make beautiful music for the first time, I think, in a decade.
So if you look back on the past decade, Melanie, I think the three successes are a little bit more out in front with other areas to follow very quickly. I would say number one is medical image interpretation using deep learning. So, medical images, including moving images like ultrasound can be interpreted with greater accuracy than even groups of subspecialists. So, that's number one.
Number two is the rapid advances in looking at protein structures just based on genomic sequencing. Something that used to take months and years. I was a molecular biologist before and it would just take a long time. And now, it's a matter of hours that we can figure out what a protein will look like. And obviously, it was big for vaccine design for the recent pandemic.
And then, lastly, I think using primitive AI tools to help with the healthcare administration burden that every health system faces has been really good at cutting cost and increasing accuracy in all the administrative aspects of healthcare. Perhaps the less intellectually stimulating, but nevertheless very impactful in terms of saving money. So those I, think, are the top three with another tier of advances that are very quick to follow like ChatGPT and large language models to help with the many, many communication challenges in the healthcare system.
Melanie Cole, MS (Host): To me, Dr. Chang, the algorithm, I mean, I hear that word, I think of things like TikTok and, you know, that sends you what you have shown that you like, and the public has a certain perception. Obviously, we have to mention the elephant in the room here. How do you overcome the perception that this is possibly bias? There's privacy issues within the healthcare community. What do you think patients would feel about using this kind of technology in their medical care? Will they even notice?
Dr Anthony Chang: I think artificial intelligence and healthcare is already here, perhaps not widely publicized. Even AI tools that really don't require humans to have oversight, which is the kind of AI, it's called autonomous AI, that really intimidates everyone. So, I think as long as the clinicians can start to understand and know AI and be able to trust certain tools, I think we will start to see wider adoption in this coming decade.
I think to balance the discussion on whether or not AI is ready or ethical or scary to people, I think we also have to realize that not using AI in healthcare is perhaps costing lives also. I think there's some mature tools that are already available, for instance, to screen head MRIs or brain MRIs for bleeding or stroke. Right now, an AI tool can expedite the diagnosis, so that the head MRI can go to the front of the line to be looked at by and validated by a radiologist. Now, if you can cut down the time to intervention for a stroke, that's definitely an improvement in the quality of care for a patient. So, those are kind of tools that are already available that are already being adopted by some hospitals. And perhaps in addition to think about the potential downside of AI in healthcare, we should also think about the downside of not deploying AI in healthcare, which is something I always try to remind everyone of.
Melanie Cole, MS (Host): And what a great point you just made because there is a downside to not employing this type of technology while we have it available. And because medical advancements are moving so quickly, we have to move with them. Do you see any challenges disappointments? I mean, we obviously just mentioned public perception, but what about any other barriers to implementing these technologies?
Dr Anthony Chang: Well, if we look at the ChatPGT and large language models as kind of a milestone period in artificial intelligence and healthcare, I think people are afraid that it is not accurate or it's going to give false information to mislead clinicians and patients, but I think we can work around those issues. I think we can curate the answer so that they become more accurate. I think part of the issue with clinicians, sort of low morale, is not having enough time to spend with patients because of the burden of having to look at volumes of electronic records, particularly for new patients, and also the information that's out there now. Medical knowledge is doubling every 30 some days now as opposed to every few years, just a couple of decades ago. So, the amount of information from patients records and the amount of information that's published in the medical literature is making it humanly just not possible to keep up. And I think tools like ChatGPT can start impact in helping the clinicians to practice better and practice higher level of wellbeing, which is, I think, ultimately one of the quintuple aims that we always think about in healthcare and medicine, which is improving quality of care and improving the patient experience and improving the caretaker experience at a time that we need, I think, an equalizer. I think AI is a great equalizer for this period of great challenge in healthcare.
Melanie Cole, MS (Host): We've mentioned it a few times, Dr. Chang, you've talked about ChatGPT. Now not everybody knows what that is, and I'd like you to speak about its role in healthcare, whether we're talking about virtual assistance for patient care, booking appointments, which to be honest, we've been doing automated for quite a while there. You know, you can do it online or you can do it even on the phone. Push this button, this button, when do you want your appointment? But this is a little bit different. I mean, because it can possibly even, what, answer medical questions, provide information. Tell us a little bit about ChatGPT and its role in healthcare.
Dr Anthony Chang: Okay. Relatively technical terms, ChatGPT is an example of a large language model, which is based on a technology called a transformer. So, a transformer for the first time is an AI sort of tool that can have the capability of something called attention, which makes it really flexible and nimble to understand how we humans communicate. So rather than just predicting the next word, like you have in autocomplete feature, it is able to complete sentences and paragraphs. It's like autocomplete on massive steroids, so by taking everything on the internet that it can have access to as a corpus or as a body of knowledge and be able to answer questions. So obviously, it is vulnerable if important information is not captured. It's vulnerable if there is misinformation on the internet, the classic garbage and garbage out. So, we have to be careful about how to use those really advanced level of information.
So, I get often the question, which is a great question, "Is ChatGPT information or is it knowledge?" And I think knowledge being something that you can act on based on the information. I think it's what I call superinformation. It's information packaged in a way that is useful and practical and perhaps even something you use to base your decision on. But I think, for it to get to knowledge, we still need the humans in the loop to help make best use of that super information. But they're certainly capable of doing very helpful tasks, like perhaps take a 300-page, medical summary and information from a patient and maybe make it very, very concise and short, so that the clinician can quickly gather the information without having to go through it all.
Melanie Cole, MS (Host): That's so interesting, isn't it? It's an exciting time in medicine and specifically in AI in healthcare. You know, I'm really looking forward to seeing where this goes. I'd like you to wrap it up, Dr. Chang. What would you like other providers to take from this key message about AI and healthcare from this podcast today?
Dr Anthony Chang: Yes, I think it's easy to be a naysayer and not believe in this kind of technology. But I do think for our next generation's sake, and particularly the next generation of clinicians, and you know, to me it's just absolutely mind-boggling that at a time when clinicians can risk their lives in taking care of patients because of the pandemic, that we have more and more people going to healthcare than ever before. So, I think it's our responsibility, those of us who are still in medicine now, to create a very special sanctuary for medicine in the future by reducing the burden and increasing the outcomes and quality of care as well as our wellbeing by deploying the appropriate AI tools that are now available.
So, AI is here ahead of schedule at this level. So, I think it's our legacy to create, to adopt AI in a meaningful way, so that we can trust this as a resource for decades to come. And it's definitely an equalizer for some of the issues that we face, like health equity, you mentioned before. Perhaps we need our, you know, AI partners to help us to assure health equity for all because humans alone haven't been successful doing it entirely. So, we could use the resource more than ever before.
Melanie Cole, MS (Host): So interesting. Thank you so much, Dr. Chang. Really, you gave us such great information and a lot to think about. Thank you so much for joining us. To learn more about Children's Health of Orange County and to refer your patient, please visit choc.org/mi4 for more information.
Thank you so much for listening to Pediatrica, a pediatric research and innovation podcast presented by clinicians and researchers at Children's Health of Orange County. I'm Melanie Cole.