We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Ep 491 Agentic and Physical AI in Medtech How NVIDIA is Changing the Space EDAI.

Ep 491 Agentic and Physical AI in Medtech How NVIDIA is Changing the Space EDAI.

2025/3/27
logo of podcast Everyday AI Podcast – An AI and ChatGPT Podcast

Everyday AI Podcast – An AI and ChatGPT Podcast

AI Deep Dive AI Chapters Transcript
#medical services#artificial intelligence and machine learning#generative ai#medical practice#machine learning theory#biotechnology and neuroscience#agi discussion People
J
Jordan Wilson
一位经验丰富的数字策略专家和《Everyday AI》播客的主持人,专注于帮助普通人通过 AI 提升职业生涯。
P
Prerna Dogra
Topics
@Jordan Wilson : 我认为人们对AI的理解存在误区,尤其是在医疗保健领域,生成式AI的影响被低估了。我们应该关注AI在医疗领域的应用,以及它如何改善医疗保健服务。 我认为,生成式AI在医疗技术领域的应用潜力巨大,它可以帮助我们更快地获得医疗服务,并最终挽救生命。 在讨论代理式AI和物理AI时,我们不应局限于传统的理解,例如坐在电脑前处理报告或仓库里的机器人。在医疗保健领域,这些技术具有更深远的意义。 我们需要关注AI在医疗领域的潜在风险,并采取措施确保其安全可靠。 @Prerna Dogra : 我在英伟达领导产品管理和开发者生态系统团队,我们的使命是将英伟达的计算技术应用于医疗保健行业,解决医疗领域特有的问题。 英伟达在医疗技术领域的应用始于12年前的图像重建技术,如今已广泛应用于医学影像诊断和机器人手术等领域。 生成式AI和代理式AI正在改变医疗保健,例如Epic系统整合NVIDIA技术,改善患者体验。代理式AI通过整合现有AI技术,赋予AI系统推理和解决问题的能力,从而提升医疗服务效率。 英伟达的“三计算系统”(AI计算、模拟计算和实时计算)将加速AI模型的开发和应用,尤其是在机器人手术等领域。 代理式AI系统通常与人类专家协同工作,以确保安全性和有效性,并提高医疗效率。 英伟达推出的Isaac for Healthcare平台,旨在通过结合AI计算、模拟计算和实时计算,解决医疗保健领域的人员短缺问题,并提升医疗服务水平。

Deep Dive

AI医疗革命:英伟达如何改变医疗技术领域

人们对人工智能的理解常常存在误区,尤其在医疗保健领域,生成式AI的巨大影响往往被低估。我坚信,AI,特别是生成式AI,在医疗技术领域的应用潜力巨大,它不仅能帮助我们更快地获得医疗服务,更能最终挽救生命。

以往,谈到代理式AI(Agentic AI),人们可能首先想到的是坐在电脑前处理报告;谈到物理AI(Physical AI),则会联想到仓库里的机器人。然而,在医疗保健领域,这些技术的意义远不止于此。

我和英伟达医疗保健AI产品高级经理Prerna Dogra进行了一次深入的探讨,她清晰地阐述了英伟达如何利用AI技术革新医疗技术领域。

Prerna介绍说,她在英伟达领导产品管理和开发者生态系统团队,致力于将英伟达强大的计算技术应用于医疗保健行业,解决医疗领域特有的复杂问题。英伟达在医疗技术领域的探索始于12年前的图像重建技术,这项技术利用GPU将医学影像传感器信息转化为可供医生诊断的图像。如今,英伟达的技术已广泛应用于医学影像诊断(MRI、CT扫描等)和机器人手术等领域。

生成式AI和代理式AI正在深刻地改变医疗保健服务。例如,Epic系统(一个主要的电子病历系统)已经整合了英伟达的技术,显著改善了患者的体验。代理式AI系统则更进一步,它整合了现有的AI技术,赋予AI系统推理和解决问题的能力,从而极大地提升了医疗服务的效率。

Prerna重点介绍了英伟达的“三计算系统”:AI计算、模拟计算和实时计算。这套系统将极大地加速AI模型的开发和应用,尤其是在机器人手术等对实时性要求极高的领域。AI计算负责训练和创建世界级的AI模型;模拟计算则提供了一个安全的AI“试验场”,允许AI模型在虚拟环境中进行大量的训练和测试,从而提高模型的可靠性和安全性;实时计算则专注于在实际应用中实现实时处理,例如在机器人手术中提供精确的控制和反馈。

值得强调的是,代理式AI系统通常与人类专家协同工作,确保安全性和有效性。例如,一些公司开发的数字代理可以进行术后回访,了解患者的恢复情况,并提供个性化的医疗建议,这不仅提高了医疗效率,也提升了患者的满意度。

英伟达最近推出的Isaac for Healthcare平台,更是将这“三计算系统”完美结合,旨在解决医疗保健领域日益严峻的人员短缺问题,并显著提升医疗服务水平。该平台将通过AI辅助诊断、机器人手术等方式,提高医疗效率,改善患者的治疗效果。

总而言之,AI,特别是生成式AI和代理式AI,正在以前所未有的速度改变医疗技术领域。英伟达的创新技术和平台,正在推动医疗保健服务迈向一个更加高效、安全和个性化的未来。 我们有理由对AI在医疗领域的未来充满期待,相信它将为人类健康带来革命性的改变。

Chapters
The episode starts by defining Agentic AI and Physical AI, highlighting their expanding roles beyond typical perceptions in sectors like healthcare. It emphasizes the often-overlooked impact of generative AI in medtech and its benefits for everyone.
  • Agentic AI is more than just generating reports; it's about improving efficiency in healthcare.
  • Physical AI in healthcare goes beyond warehouse robots; it includes life-saving medical technology.
  • Generative AI's impact on medtech is significant and benefits everyone, not just specific industries.

Shownotes Transcript

Translations:
中文

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life.

When we think about agented AI, sometimes I think people think of sitting behind the computer screen and maybe helping you get that quarterly KPI report out quicker. When we talk about physical AI, maybe we just think about humanoids in a warehouse. But both of those things are actually...

so much more than that. And especially when we talk about, I think, highly specialized and very important sectors like health care and medicine and medtech.

You know, I think that generative AI, you know, is maybe always not talked about as much as it should in those fields because it impacts all of us, right? Not everyone here is in marketing, not everyone here deals with legal, but, you know, we all benefit from what generative AI can do in the medtech space. So that's why I'm extremely excited for today's conversation and back at, in

NVIDIA GTC to talk about agentic AI and physical AI in MedTech and how NVIDIA is changing that space. All right, I'm excited for today's conversation. I hope you are too. What's going on, y'all? My name is Jordan Wilson. I'm the host of Everyday AI, and this is your daily live stream podcast and free daily newsletter, helping us all not just keep up with AI, but how we can use all of the latest news and advancements to get ahead to grow our company and career.

That's you. You're on the right clicks. So make sure you do go to our website at youreverydayai.com because if you miss some of the great insights from today's episode, we're going to be recapping it all in the newsletter. So make sure you check that out. All right. So like I said, if you're listening on the podcast, you don't see this. You're on the live stream. You see we're here at GDC and there's been a ton going on. So I'm excited for today's conversation. So please help me. Welcome to the show. Prerana Dobra, the Director of Product Management for Healthcare AI at NVIDIA.

Pranath, thank you so much for joining the Everyday AI show. Thank you, Jordan, for having me. All right. I'm excited for today's conversation, but before we get started, tell everyone a little bit of what you do in your role at NVIDIA. Sure. So at NVIDIA, I head our product management and developer ecosystem team.

And essentially our mission is to translate the amazing technology stack that we built with our computing appliance and our accelerated computing, artificial intelligence into the most impactful industry, the healthcare industry. So we are really solving for like what are the domain specific problems, right? Healthcare is very domain specific. So as an example, in computer vision, things like cats and dogs videos,

to these MRI and CT one metric images. So how do we turn back domain problem into a computer science problem and really bring accelerated computing to that? Similarly, when you're thinking about agents then the big bang of LLMs, conversation layer.

How do these AI models talk about the medical jargon, the medical stuff? We sit at that translation layer of NVIDIA's amazing technology staff or the industry of healthcare and life sciences and my focus is specifically in medtech. And for that, we may create new software capabilities because we found something very unique to solve or sometimes

we adapt our existing staff and really build very intentional good markets with our partners and our customers. But essentially, we need to. Yeah, and it's extremely exciting. But maybe how it explains for our audience a little bit, how does this ultimately play out in their lives, right? Because no one goes to the doctor and sees, you know, an NVIDIA logo on there, right? But I'm guessing, you know, that committee of the major, you know,

names in health are using something along the line, whether it's GEUs from NVIDIA or your software in the health space. But tell us a little bit, who are those partners and customers that are using the MedTech offerings from or through NVIDIA?

Absolutely. So our journey in MedTech really started 12-12 years ago. And the first application that we identified for accelerated computing was this thing called image re-instruction. But what that essentially is, that when you go to get your MRI or your CT scan, it's actually a GPU that can take all of that sensor information and convert it into an image.

And so in diagnostics, generally if you talk to these clinicians, they'll say what you can see, you can hear, right? And all these breakthroughs like with PET scanning and so on, like how with low dosage can you get these really high quality, high fidelity imaging? So that is one area where it's almost like NVIDIA inside all of these devices. Robotic surgery, right? My, my

It became so real, it really came home for me. I always thought of this, right? Like, not all fast food customers are in industries, but we're all consumers of health care. And I thought two years ago, I had to go through a prostate surgery. And that was the Da Vinci system of our media, and it's a minimally invasive surgery. So really, again, right, bringing this ability of keeping care of cancer,

and solving with technology what couldn't be possible in the past.

Super, super touching story. Yeah. Like the word that you and your colleagues are doing, you know, being able to see it help, you know, firsthand, you know, in your father's surgery. That's really, really cool story. So, you know, one thing I do want to talk about is, you know, before we get into, you know, the announcements and what's new and what's changing, can you maybe just bring our audience to like, where are we at today? Because as an example, you know,

I didn't know that, you know, image reconstruction was essentially just powered by GDUs, right? So like, where in the attic you had to give a, you know, the state of the state address for, you know, AI in MedTech. Like, where are we? Are our most, you know, large healthcare organizations, you know, using, you know, Gen AI and AI GDUs, you know, around there?

entire structure? Or is it still like, you know, some companies and some larger organizations are still, you know, kind of quote unquote old school and still figuring out this new Gen AI wave? Yeah. So what I would say is because healthcare is so fragmented, unlike any other industry, we have like one big layer and if they take Gen AI, it just becomes ubiquitous. But I would say that Gen AI and the promise is really not to do less.

The journey really started with a lot of the perception AI piece which have been there for like eight, seven, eight years now and you can see the growing number of FDA algorithms that are approved with deep learning. But this power of Gen AI and Gen Z is now really coming through. So as an example of ethics, I don't know if everybody knows about it, but if you've used your MyChart,

And it is the number one EMR system. It is getting followed by all sorts of these co-pilots and really streamlined innovation experience. We also announced the integration of NVIDIA's technology stack and our models into Epic on their AI phone. That is one direction, right? Once someone like Epic integrates Gen AI, it's going to touch everyone's life. And it's as simple as when I open my MyChart app.

today I look at the summary and it's a bunch of jargon I'm not following. Can I have a patient summary that's in English that I can write? In the future, can I make a call and not be on a waiting line? And even to talk with a digital agent, like, so these capabilities are sort of really bringing that access to care. We're totally seeing that come true.

I think it's a great example. A bridge is again a wonderful example, right? Like when they have really, they have state-of-the-art technology for their automatic speech recognition and text-to-speech and this really ambient intelligence system.

where they are bringing back that conversation between the doctor and the patient, right? Like the last so many years, it's been you're talking to the patient and the doctor is constantly at the screen, helping administrative work. The neurosurgeon is putting in the goals. So that's a great example and a bridge is installed and is having a great effect on it.

Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast.

Companies like Adobe, Microsoft, and NVIDIA have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for chat GPT training for thousands,

So when it comes to a genetic...

AI in MedTech, could you let us know where are we at now? Maybe talk us through some of the benefits or use cases that you're already seeing. Because it seems like last year here at GDC, it's like agentic was more of a whisper, but now it's everywhere. So where are we at and what are some of those use cases already in the MedTech space?

You hit it on the head, Jordan. Last year, it was so like Gen AI. And then I think it was like, oh, we don't have our foundation models yet and so on and so forth. And that is actually the proof point of what a foundation model means, right? Because all these companies, to take these models and attach them to our contracts and really give AI now agency. That means not just that the...

AI model has the knowledge, but it now has the context awareness and will now like this ability to reason through a problem, right? So if I'm a patient, I'd have to call him to the hospital and figure out, hey, what's my care plan? Or how do I get? Where's the party? Or what's my insurance code and so on? You can have this

the agentic AI system that is then calling and knows like I need to call a specialty model from the insurance place. I need to call the specialty model from a nutrition space. I need to pull this out from the EMR and so on. So what agentic AI has done is taken all this amazing technology of AI that's being researched papers and a model here and a model there, but we already brought it together to deliver something of value.

And so where it is right now, I would say is where we are certainly seeing Blackstone down the bridge, not the great ones, but the great AI. When we're seeing these companies that have set on this journey really taking off exponentially because those systems are performing incredibly well in the real world.

But even more importantly, it opens up the avenue for so many more to now come. Because traditionally, NetApp has been like a very regulated industry, as it should be. But the bar to bring real world impacts from all this technology has been very high. We just needed this Gen AI, Agentic AI, and now this ability of reasoning and chain of thought, which allows to bring more transparency into these systems.

the perfect time to really bring value into healthcare. - The benefits, you know, I think are fairly obvious, right? You know, better healthcare, you know,

you know, communication, hopefully improved outcomes, right? But I think, you know, a lot of the early talk about agentic AI was just based around more of like, okay, here's things that I myself would be doing on my computer, right? Researching this, updating this document, you know, sending it over to an internal or external stakeholder.

Can we talk with the benefit or when the potential benefit is so high with agentic AI in medtech, what about the potential risks? Because I know people are always worried about agentic AI and then when you get in the medtech space. So can you talk a little bit about the work that you and your team or maybe your partners are doing to ensure that agency doesn't go off the rails, so to speak?

Yeah, that is an excellent point. And also to the kind of, this is a regulated industry as it should be. So what we see with our part, even with these systems and agents, you are essentially with an expert in them. So the value actually is in the operational efficiency of these systems that they can bring.

As an example, it helped contextualize like Hippocratic Air has built these digital agents that can make a call post-operative to a patient and find out how you're doing after your surgery and have you been taking your medications well. The patient can ask what kind of dietary plans I should be in. One, these are areas that don't happen today.

So these digital agents and systems can bring care where it's not available at all. But then also because they are working, in case of Hippocratic AI, they have a whole body of nurses across different systems really trying out these digital agents and things like that. That works, that doesn't work. And really giving that time back to the clinicians and the nurses. So in all of these cases,

with the bridge, with Hippocratic AI, with a whole suite of these startups, art-style companies. They're all very much with that human expert in the loop that's actually trying to say, like, all these agencies, well, God, and so on.

From a technology standpoint, NVIDIA does that. We have in our own staff to make sure we have the right software and allow developers to build very safe systems with things like guardrails and so on.

So there's been a lot of talk, you know, so far here at GTC about kind of this three computer system, right? Can you explain that, you know, maybe to our audience that didn't get a chance to listen to Jensen's keynote, which you should all go do, by the way. Can you explain what that is and how it's going to be impacting the work that you all do?

Yes, this is truly an inflection point. And me, I've been an engineer and technologist at the heart. So you can really see this convergence of three very key computers, right? The first one is your AI computer. This is where you are training and creating these world-class AI models.

The second computer that's really come to life is now the simulation computer, right? And if you think about the simulation computer was actually one huge missing piece. So the AI computer had its breakthrough when we cracked this notion of like our supervisor. I don't need data that's constantly labeled.

The simulation computer now almost is the AI playground. All this intelligence that I created, right? If you think about any field of robotics, there is no chance that you can have humanoids and all these robots. Imagine in our industry, like several robots, the cost of doing that thing and all of that in the real world is so high. So now you almost have this playground.

I can be generating synthetic cases. I can be generating synthetic words that are mimicking

The real world is not. That is the second attribute of simulation. You take something that you train, you put that into simulation where it goes into post-training and it's getting really refined. So a model that was trained for task S or foundation model is brought in and now put into simulation to get really contextualized in the new task, which is similar but in a different domain.

the best part of the simulation because it's not running on wall clock time, right? So your AI can practice and practice till it really masters this gear.

And the third computer, as we like to call it, is the brain of the robot. That is really the real-time computer. So we have invested heavily into this platform called the NVIDIA Holoscan, which is all about this real-time process. And so these three computers, the AI computer to train models, the simulation computer at the playground for this AI tool, and there are systems, actually, because it's not just one again and one again stuff.

combination of AI models and agentic AI frameworks to really bring this sort of intelligence that's just not knowledge and perceive and act in meaningful ways in the physical world into that simulation computer. And when you think it's a quiet escape, oh, I think this robot knows how to do a suture just perfectly. We could take that and bring it into a real, what we call, field-to-real transfer of a prototype. And

The fact that you can go through all these three loops in constant digital prototyping more is

is what that really means for every industry and then for healthcare specifically in that sense. Yeah, I think that was a crucial explanation because I'm sure there's people out there that when they hear agentic AI and then they hear humanoids performing surgeries, it's probably helpful to understand that this is a very well thought out, I'm sure there's

trillions of data points that go into this and so many, uh, countless simulations. So thank you for, uh, you know, taking us behind the curtains on that one a little bit. Uh, but you know, so, so one question as we wrap up here, uh, because we, we've talked about a lot, uh, in this conversation so far, but you know, out of everything, you know, that was announced here at GTC, um, what would you say that, you know, at least, you know, on, uh,

under the umbrella of today's episode of A.H. Etting AI and Physical AI in MedTech.

What's maybe the one advancement that you and your team maybe are most excited about? And then what will that recent announcement here mean for your space in the future? Kind of like, hey, are you able to tackle the new problem? Are you able to offer a new type of service for different type of healthcare organizations? What's that one thing that you're most excited about? Yeah. Yeah.

I know it's hard to get to one of those. I know it's hard. I've ruined that. Yeah, it's like a two-hour keynote. How can you pick one thing? I want to pick something that wasn't in the keynote. But it was in Kimberly Powell's question address. She's the VP and GM for our healthcare business unit. And we launched what we call Isaac for Healthcare. It is this domain specialization of the three computers I just walked through.

And personally, I am most excited and passionate about this space because the reality is we are millions of people short of healthcare staff. The reality is we have aging parents, we all want to be consumers of healthcare and so this

So it's not just about innovation. In healthcare, it is a necessity to really bring physical AI to help in ways it was not possible. I was walking to the floor and there's this company called Wondercraft who does like these skeletons for people who are paralyzed. And to see that young woman walk, you know, stand up from her wheelchair and be able to walk,

That is as life-threatening as it can get. There's a company called Synchron. They design these brain-computer interface implants and they have this whole way of potentially patients with ALS. How do you take that signal, turn it into an intent, and allow someone who

who is not able to interact with the world because of the impairments of the so many things, give them the agency to be able to do so. To be able to ask your wife, I love you. To be able to turn on the fan. To be able to tell your caregiver, I can't breathe. So this is going to be a long journey. We are aware of that. But we are here because this is an incredibly hard problem to solve.

And we are committed to take this forward. So like I said, it is a necessity. And so that's where my mind, my heart, and my entire focus is going to be. To really see how to bring these three leaders to life. To see as they take on the bigger challenges, robotics, and AI.

Love to hear it. It's just extremely exciting and inspiring to see what's next because I think sometimes people think, oh, GPUs and this, that, perhaps the headlines, but what that opens up in terms of capabilities and possibilities just for us as humans, for our health, I think is extremely inspiring. So, Prerna, thank you so much for taking time out of your day to join the Everyday A-Action. We really appreciate it.

Absolutely. It was a pleasure. Thank you, Jordan. All right. And there was a lot in there. So, you know, if you missed something or if you want to know more about something that Prerana said, it's going to be in our newsletter. So make sure you go check that out. So if you haven't already, go to youreverydayai.com. Sign up for that free daily newsletter. We're going to be recapping this as well as everything else you need to stay up to date, get ahead with Gen AI. Thanks for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.

We're sunsetting PodQuest on 2025-07-28. Thank you for your support!

Export Podcast Subscriptions