The Real Bottleneck for Healthcare AI: Models, Data, or Trust?

Okay, it looks like Tom from Z-Wallet is running late, so let's get started. Hello, A1 and welcome to today's Joanne Care Space. Today's topic is the real bottleneck for healthcare AI, models, data, or trust. It is a simple question, but the answer will shape how quickly healthcare AI moves from impressive demonstrations into useful, responsible adoption. Healthcare AI is already entering real research and care workflows. One recent example is Verily's announcement of new investments from Nvidia as it expands infrastructure designed to make healthcare data AI ready and supports responsible AI deployment. But this also highlights a bigger question. Do we still need significantly better models or are fragmented data, difficult collaboration, privacy concerns, and limited public trust, not the more important barriers? Today we are here from perspectives from projects working across healthcare AI data infrastructure and the designs. We'll discuss where AI is already creating value, why healthcare data remains difficult to use, how projects can earn trust, and where decentralized technologies may offer practical solutions. So let's begin with a quick introduction from each of our guesses. First, let's welcome Peter from Joanne Care. Peter, could you please introduce yourself and tell us more about Joanne Care? Yeah, thank you very much, Candice. Delighted. I'm Peter Saxenma, I'm actually the CEO of Joanne Care. My career has focused on connecting science, technology and industry. I'm a member of Akatek, Germany's National Academy of Science and Engineering. And over the years, as you will see if you research me on the internet, I've worked of course international research, business leadership and advisory roles, and transformation and complex innovation systems. One thing I've learned is that promising scientific idea needs more than good science. It needs the right funding, expertise, collaboration, and a pathway to real world impact. At Joanne Care, we're trying to build an open incubation network for early stage biomedical innovation, and make research projects with scientific and medical experts, contributors, communities, and Web 3 infrastructure. So I really look forward to be able to engage in a discussion with all of you today. Thank you, Candice. Thank you, Peter. Hi, I'm Joanne. Next, let's welcome Victoria from PFPI. Victoria, can you please introduce yourself and introduce PFPI? Thank you so much for having us. I'm Victoria, the community manager at PFPI, which is, we call ourselves peripheral intelligence network. The short question of what we do right now is, if you create something with AI model and RPU, say a data set, you clean and labeled, but that thing has like no real identity, so there's no owner and no fingerprints, no paper trails, so it just could get used, copied, or resold, and you have no claim to it, and no way to prove that you made it in the first place. So PFPI gives AI-era intelligence on chain identity. So we call it an intelligent fingerprint, so basically a vector based process. This assets, this is homemade it, and here is its whole history from creation, so every time someone buys it or it buttes on top of it. So it does get paid when their work gets used, and everyone else gets receipt they can actually check. So I'm bringing that lens into this conversation, copied with being a medical biochemist, and because healthcare AI has the exact same one, just dressed up differently from my perspective. So it's not really a model problem anymore. I think it's an ownership and proof problem who made this and who can be trusted with it, and who gets credit and gets paid for the data that trained it. That's such bread we're pulling on today. So yeah, I'm pretty excited to hear from you, because thank you. Okay, thank you, Victory, and let's welcome Hong from Sea Wallet. Tom, please introduce yourself, and tell us what Sea Wallet is building. Hello, Tom. Tom, can you hear me? Okay, seems like Tom has some connection issues. Okay, so let's get started. The first question is from your perspective. How is AI changing healthcare today? Hi, can you hear me? Tom, Tom, I can hear you. Oh, good. Oh, finally, finally. Okay, that's fine. Yeah, thank you so much for inviting me, and sorry for the delay in the space. You know, like space always having problems. So it's me, Tom, a core contributor at Sea Wallet. So for those who are unfamiliar with Sea Wallet, so we are all in one crypto wallet, and Web 3 financial platform focused on making digital assets actually usable in every applications and businesses, waste support payments, transfers, swaps, earning lending cards, and business focused tools like bulk payment and HR payments across multiple chains, and thousands of assets. So while we are in order, healthcare AI company, but still we say that an interesting intersection of Web 3 payments, data, and user trust. And I think those lessons are highly relevant to healthcare AI, actually when we start talking about data ownership, privacy, incentives, and trusted infrastructure. Yeah, that's it from my side. Thank you so much. Thank you, Tom. Okay, let's get started. Get into question one. So the first question is out in AI, changing healthcare today, and we choose case is most likely to reach wider adoption within the next one or two years. So let me invite Peter. Peter, what do you think the question? Can you hear me? Yes, I can hear you. Very good. Now my answer will surprise you a little bit because I think I will not automatically and not magically change healthcare. We need to ask much more specific questions. The reason for this is that we have a corpus problem. Medsin has no open in today to find a really big 50 year old corpus of massive data, mainly text and visuals. And we actually constructed the current models on their basis, but there is no such thing for healthcare. Healthcare and medicine have large quantities of text, code, images, public digital material. But all of this is very much dispersed scattered across health records, imaging archives, laboratory, platform, device logs, clinical notes, guideline PDFs, reimbursement rules, consent records, hospital workflows, supply chain data and tacit experience. By the way, being a very important point. So I find there are two industries that are currently looking to changes and you very rightly can this in your introductory remarks mentioned the investment of Nvidia in healthcare. There is a similar much more pronounced investment. It's a company called Prometias. It was instigated by Jeff Bezos. It had an initial capital from the beginning, if I remember correctly, of 35 billion, repeat billion dollars. And it tries to establish models specifically for the manufacturing sector. I mentioned this because there are so much to learn from this Prometias test as I call it. Prometias pursues the artificial general engineer where document of last year they tried to get to a situation where concrete solutions are achieved not like previously by a hundred engineers in 10 years. But by 10 engineers in one year, if you have that kind of basic Prometias approach, then you actually have to ask much more concrete questions as to what you actually want to improve, what you actually want to compress in terms of time, cost, manpower, error, and all of this by an order of magnitude. I find this in my own. So I'm approaching this whole thing from much more CO perspective, where I know that in the next iterations, especially because we're so successful in advancing in these knowledge fields, we will not have enough money to pay for all of it. Peter, I can hear your voice is your answer over. Okay, Victoria King. Yeah, I think I think like he have some problem on. Okay, so Tom, what do you think the topic where? Okay, so.