Health Data Management & Ethical AI (Panel 2)

Transcript

Ladies and gentlemen, welcome to the second panel. They are already on stage. Carmen is very efficient as always. It's wonderful. I'm very looking forward now to this one. We already had the first question for uh blockchain and that was actually the perfect introduction. So, we'll speak about decentralized health data management, one of Carmen's core topics. So, please Carmen, take the floor. I'm very looking forward to this one. and a round of applause for them as well.

[applause] Excellent. So, can I ask something please? If you're sitting really in the back, usually at school, what the professors would do with us, if you're really sitting in the last rows, they would ask you to come in front. So, could you please in the last three rows, could you come closer so that you see us better? you can ask more questions and uh we will be very nice and cozy. Thank you. So while we do that, let me yeah give you some some thoughts and clarifications because we already uh you have seen from the previous panel, we already received questions about blockchain and smart uh contracts and technology and AI. So why are we not

talking about technology? we are we're talking about technology and uh the back end let's say of decentralized healthcare in this panel the first panel was more the I would call it the front end of healthcare decentralized healthcare so we looked at the patient perspective and it made sense that the patient should not know whether it's a smart contract or a blockchain or an AI federated learning model they need to know that the data is accurate they can trust it their physicians know what to do with it they are empowered etc. In this panel again we're going to look a little bit at the technical part uh because we do have a very interesting uh set of um colleagues from the panel.

I will introduce them in a second. I will let them introduce themselves but before that it would be good to see who we have in the audience. So if you are from if you see yourself representing the healthcare sector so pharma research organizations hospitals met can you please raise your hand healthcare who is in healthcare very many people [laughter] interesting would you say you're more the web three people so blockchain decentralized let's see how many web three people, technology, hardcore, AI.

Okay, good. And you can raise your hand multiple times. Is anyone else that doesn't recognize themselves in the two categories? Academia, students, other industries. Okay, good. So I think we're talking to uh an audience mostly formed of web3 decentralized tech people which is good because that's exactly what we're going to have the expertise today. Um very briefly about myself because I realized I did not introduce myself until now. Um my background is in digital health and for the last decade I used to I used to work now not so much uh within the pharmaceutical industry. So the business

side have been implementing digital health solutions. We did AI models. Um we looked at at technology but it was mostly the the business side. Um however we do have and now I would like to uh give the the floor to to each of our panelists. We have a very good blend of expertise in this panel in terms of research, scientific research, uh entrepreneurship, uh machine learning, [laughter] AI, blockchain. So Arash could we start with you?

>> Yeah, I think we can do this first. >> So hi everyone. >> Yeah. >> Hi everyone. I'm my name is Arash Rafi. I'm primarily a OBGYn so gynecologist. Uh I was uh mainly in academia for 15 years at Cornell University where I had my own lab. uh my practice was around Gynong and I also led the clinical research department there for two or three years. I did a lot of humanitarian work not with MSF but a lot of travels to Haiti and Africa in particular and uh few years ago I left academia to build an open-source platform for scientific

publication which is called CNFT and from that we spin up with one of our scientists who publish a lot of his work as open source we spin another app called Lenare which is an app for woman All right, I guess I can go next. Uh, so I'm Charles. I'm a machine learning engineer uh at Flower Labs where we uh develop uh an open source framework for training uh AI models in a decentralized way. Um, so I don't have too much of a health background. Uh, I'm more on the technical side or the AI side. Um, but I've been involved in quite a few uh

health related uh projects. Um, and yeah, I guess that's mainly it. >> Um, I'm Choya. I'm actually a psychologist by training. So, I'm not a technician. Uh, but I've been kissed by blockchain virus due to RTO, one of the founders of Crypto Value Association almost a decade ago. So, Roto, you're here and because of you, I'm here as well. Uh, and we have founded um blockchain for healthcare very long ago.

Um, had went to a lot of different iterations and now we're combining AI since a year uh with blockchain and I think this is also what we're going to talk about today. >> Excellent. So, we are going to have three themes. you have seen this theme of three repeating. Um in about 20 25 minutes we will take your questions. So please start um thinking about them. We will look of course at the let's say because we we launched also the report as as we were mentioning in the beginning. We look at let's say the life cycle of of data management and of course when you talk about data there's no data without AI these days. So you can you can you know benchmark it to to

the life cycle of of an AI model as well. So we look at the early phases with data generation. The previous panel covered a little bit the types of data uh that can be collected by individuals in a decentralized way. But I think we have much more experience or or broader um case studies here. We will look into federated learning. So we need to see how many how many in the room know what is a what is federated learning FL. Okay. So we will very good. We will explain that and then at the end we will look into decentralized governance.

Dows, not Dows, but basically um how do we yeah make sure that kind of at the more mature level of of um of data management there is some decentralization in uh in the governance right and maybe a very quick uh disclaimer before when we when we say decentralized we don't mean everything needs to be decentralized right there's a continuum of centralization to decentralization there elements and we will talk about some of them where keeping a centralized oversight or even infrastructure makes sense. So we're not coming here and saying the future of healthcare should be completely decentralized. Very important. So if we

go to the first round decentralized data generation what new health data can we unlock? I think Eberart you touched a little bit on on the health intelligence but not too much. You just mentioned this blockchain and AI. uh could you tell us a little bit more about what you're doing with the health intelligence and especially what type of data are you helping the users to collect that maybe a centralized um hospital system wouldn't and what are the most promising use cases? >> Well, maybe we have to go a little bit back what the situation is nowadays in healthcare. We talk a lot about silos.

uh data sit in silos and are not accessible and the type of data that sit in the silos pharma hospitals I used to work at the university hospital in Zurich uh this is uh only a kind of a snapshot of data clinical research is also very specialized case drugs are giving under certain conditions you measure the outcome i.e what uh the treatment benefits are uh but that has nothing to do uh with real life. So on the other hand in the first panel they talked about uh sensors uh wearable devices that's totally on the other end but both types of data are equally

unusable for the judgment about the the patients in everyday life. So because we exist uh between those special cases going to the doctor then after 6 weeks we have another consultation. So the doctor doesn't know what happens in between. Uh the patient knows if they care about their variables something about themselves. Uh and actually the patients would be the only ones who are uh able to consolidate the data from the different uh silos. I mean two decades ago they designed the uh the first idea about the national rec health record in

Switzerland. We all know since a couple of days it will take another five years until it eventually will be accessible as a tool to the patient to collect and organize his or her data. And uh I'm very cautious because I used to consult companies in that field so for many years and uh it's a total mess. [sighs] And the type of data we help patients or users I would say not everybody's a patient. Sometimes we're patient, sometimes we're just users. Um, I would like to have an overview uh of what's going on with me and I tell you one

reason why we started with AI is because I went to get my flu shot last November and and you know the uh vaccination record in Switzerland got hacked a couple of years ago. So there's no digital uh vaccination record in Switzerland. So I asked the lady at the travel center in Zurich uh can you give me a digital certificate of the data and she said no but I can give you paper. So what I did I was really uh I of course I knew that's going to happen and then I wrote a blog post about how I sent took a photograph how I sent it to Grock the Twitter AI and asked the uh Brock to give me the fire resource uh

for that event. fire for you who don't know is um data exchange format uh so that you can actually give those data that I just photographed to uh legacy systems and so with that I really brided uh from paper to the digital world all by myself with tools that exist uh and uh then I wrote this blog post that I never trust the government again that they will uh give us uh appropriate tools and I got like 53,000 rep uh uh impressions on that. Then the Bag uh got involved and then uh this was it.

And then we I decided with my team that this is probably a good way of entering uh AI into the whole game because what AI does and what we use it for is just summarizing data from different sources, photographs, PDFs, wearables. Um so that uh you can get a brief overview over your health and you can also share that kind of data with whoever you want to with a doctor with a nurse with a dietician and you can also enter data or tell the AI agent data uh about what's going on in everyday life which you usually don't get.

>> So everything from verbal data to X-rays. I hope they're not leaving because I'm talking too long. [laughter] >> Very interesting. So talking about what type of other data or other correlation between let's say user patient gener patient reported outcomes if we call them like this um Arash in in Lena care when we when we were meeting you mentioned that um you have collected I think data from 3,000 women in 55 countries um different things fertility anxiety sleep etc and you have seen a correlation that maybe has been a little bit missed from the traditional clinical research could Can you tell us a little bit? Um, >> yeah, the story is is really interesting

because it started as a kind of a personal question. My my daughter needed a contraceptive pill. I give it to her and and and she asked me, but how does that impact my fertility? And I'm like, it doesn't. And she's like, but how my what impacts my fertility? And I one of my friend Ginberg is one of the best fertility doctor in France. So I called him and I said, do you have a score? how can I answer my daughter you know like basically that's not my expertise and he says oh I did a score few years ago and he sent me a word file basically with few parameters and then we started doing the research and we created like a 50 question uh website where any woman can just answer questions and and know what

are her chances to get fert you know a baby or what she should work on what she should stop you know what lifestyle changes and we really put it online for our patient because it was just easy And we told them do it before you come in. And so you know exactly you know instead of me talking to you for 20 minutes just do it. And it just like went viral basically it went like suddenly everywhere like Indonesia, China, I don't know how. and and we realize women love to give you I mean women love questioner that I have noticed before but they they're really eager to give you data if you genuinely give them something in return and we don't ask for subscription and we will never ask for any subscription so that's for an answer

to the panel before and we also realize that we're very European centric you know we're here talking blockchain NFT uh I travel the world and I've seen women in Africa they they never will care about these things. And so, um, I think one of the major thing is if we want to get better at medicine, we're way beyond clinical trials. We need millions of women answering questions, millions of men and women. I'm just more interested in women, but millions of them answering questions to find correlation. And for us, for example, we found this very high correlation between anxiety, poor sleep, and lower fertility. And then the biology of it, it also makes sense. is higher cortisol and adrenaline that impacts other hormones. But but so we

then we reach out to this woman and we told them like listen like you're a bunch of you have that like correlation or in in this like ven diagram kind of thingy and this is something we advise and then the woman told us basically uh why don't you can we have more content can we have yoga classes that we can do online because like everybody is not in like San Francisco and cannot go to yoga class nobody some women have never done like uh the most women that are the fabric of society So nurses, teachers, they don't have access to this like I would say like bobo kind of bohemian or like higher class thing and we never think about them. I see them in the clinic every day, you know, and then they have no clue what meditation is, yoga is, you know, they have never seen

a therapist, they never think about it. And so that's how we came to create the app which now has going to have eight to 10 question about everything, sexual life, personality, and a lot of I learned about myself when I do the questioner to test them. I'm like, "Oh, I didn't know that about myself. That's cool." And so I think that the problem is that everything is incentive is a problem. And the incentive right now is to sell you a new drug or get somebody in a university to get his next like professor appointment. That are the two main incentive of every clinical trial, you know, and I've been there. I've done that and so and so it's never patiententric. It's never patiententric and and one of the thing that I always I'm part of many

scientific societies. Where are the patients? Where are the women and the men that we are treating as physicians? Then you know that that dogma that every conference is like pharma sponsored so the people in the the civil society cannot be in contact with them is horrible like you never talk to the patient so you never know what they want actually for their life and you know so so I think the next phase is basically offering and some somebody says something super interesting in the previous panel is that all the EMR system are horrible and it's not transferable today it's very easy to give patient an app. So you said like you know it Switzerland is going to take five years to to create a digital vaccination already took 20 years.

>> Yeah. Yeah. Yeah. I mean give me a tenth of a budget and I build you an app where every patient can have all their data and they can add doctors to their profile exactly like you have a Instagram f picture and you have friends you know and that can be a private space. So so I think we're going into that era. Uh I also think that it's a little bit and I we kind of failed with that with our previous project on that topic. It's people are not educated on blockchain NFD. It makes you know everybody's scared like it was so much thing on the media of like oh you know crypto is used for drug trafficking or whatever. So I think that's another layer that's a little bit later. Right now we just have to get people to have like all their medical data with them.

>> Yeah. Oh, that's uh we know nothing happens very very very quickly in healthcare and yeah but I think what the bhari is also doing just digitizing you know the existing data um finding correlations for for primary and secondary kind of use of data that's that's amazing federated learning is coming next and uh we will come back to also you know how patients or individuals can can um yeah contribute with their data in a privacy preserving manner. in these federated learning um approaches. But first of all, Charles, you have seen the result of the audience. So please explain us how can an AI model train on data sets without moving the data in a central

repository and maybe you can take an example of you know hospital from Geneva Hog that wants to contribute uh uh research with another hospital in US and one in Ghana. >> How do they do this? >> Absolutely. So I think first of all it's important to consider the fact that most data we generate is uh distributed naturally. Uh so it's not all in a centralized server. Um and the classical ML approach so even if we have uh medical centers in different parts of the world um the classical uh machine learning uh paradigm would be to send all of this data to a central uh location. So to a central server where

we have all the compute and we would actually uh do the training there and so uh this is bad for both uh user or in this case patient privacy and also uh just regulations because most of the hospitals cannot just uh share data across uh borders. Um and so ferated learning is actually pretty simple in um conceptually it just shifts this uh or actually reverses this parting and instead of moving the data to the computation we move the computation to the data. So instead of sending data to a central server, uh each hospital is

going to train uh locally on their own data without moving it uh and only send uh the results of the computation to a server. So uh this computation for training a model uh might be complex but we can maybe take a very simple example. uh say that you have all those hospitals and you want to uh compute the average age uh of patients there. So the classical ML paring would be to send uh all of the data about patient ages. So you would just send a list of patients and their ages and instead with federated learning uh

you would for each hospital you would compute uh the average and then only send the average uh in the end and so with all the hospitals sending their average you can get the the exact same result as the case where you have sent all the ages to a central server. And so this is the rough idea of federity learning. And yeah just one thing I wanted to add is that here we are we are in a case where we have few hospitals but this also applies where with cases where we have wearables generating uh data from um thousands or millions of users.

>> Exactly. Because we're going to get to local AI and how individuals can can also uh contribute data without letting the data leave. But before that, Charles, can you share with us? I was researching a bit the flower AI partnership with the with blood counts consortium, which I think is the largest in the world in terms of at least for the the kind of the blood um research on rare diseases in terms of federated learning. Could you tell us a bit about this? >> Absolutely. So this is actually very interesting because so this project uh focuses on uh actually blood tests uh which is I believe the most widely done test uh in the healthcare space uh it's near uh billions of tests done every

year um and so this is worldwide and so this is a perfect use case for federated learning because we have all of this data that's already being generated. And the good thing also is that it's mostly already standardized. And so uh blood counts which is a consortium of different uh research centers and hospitals uh kind of um so they did all the networking work uh to get all those different centers uh aboard and then uh FL we provided the software and the infrastructure to actually uh allow this uh training on all the data that was

already being collected. And so I think this is a very interesting uh example of uh why federated learning uh can be very useful in this space because first of all this is something that could not be done with uh central uh AI. As mentioned before you cannot just uh share patient records u with different organizations. And also another very interesting uh technical uh detail is that um in uh this case of blood tests uh you can have results that are very specific to different demographics. Um so for instance you could have a bias of some populations that have lower iron levels

for instance. And so if you only uh use your local data to train models, you would have models with a huge uh bias. And per learning allows us to generalize better on uh uh because you're essentially improving your model with a bunch of other data that's distributed in a different way. >> Yeah. So I see already the cross kind of crossboard but also this as you say the richness and the representativity of of the data of the data sets but then is also the legal ending kind of the um the ethical thing of are you going to share my you know very very sensitive PII too um so very interesting from a hospital

perspective Eberhart maybe could you comment a little bit you mentioned you you try with the health intelligence you transform the data into fire um standards so it's interoperable yet you are you are not um uh managing the data or you're not uh holding the data on behalf of the patient. Is there something about the infrastructure that you build? And we had the question about smart contracts where such a decentralized uh system could be used by patients to share data for research but in in a privacy preserving um manner.

Well, there a couple of approaches uh to tackle that challenge. Uh first of all, yes, we are very uh data averse. So we don't want to hold any data. We don't want to be the custodian of data because data ownership comes with a lot of regulatory burden. So the way we organize it is right now the models are still too large to run on local devices. Uh so we rely on the usual suspects Amazon and Google cloud.

Uh but and they're very expensive actually if you run your own model there. Um so but the the patient or the user they organize the data in a folder can be local can be on a distributed uh storage system can even be on Google drive if they want. So they give access temporary access uh to their data to the agent and the insights that are generated are returned to to the user and there is no cache the data staying on the server.

So uh but this is the way it works right now and it's a little bit the application comes to the data uh then the other way around. uh there are certain cryptographic uh methods that you can apply. Something is uh serial knowledge proof that uh you can ask the data okay is that the case but you never you get an answer but you never have access to the raw data. Then the other way is fully homorphic encryption that you can do analysis on encrypted data.

So nobody really can copy your data. In the case of genomic data this is really important. my co-founder, she does that in a project that might be interesting to you, Ginor in the Netherlands. Um, so this is how we do it right now. We don't touch any data. Um, and we see the future uh in such a way that the AI will be more optimized. So the and the devices will be uh much more powerful. We already see that with iPhone 17 which is prone to run uh Apple intelligence on the device and that's the strategy of them. So we

see that more and more so that the AI will run on the devices. So you don't even have to transfer the data in the future. So uh you're the master and the gatekeeper to your data and you share whatever you want with whom you want. You don't have to care about regulations because it's your data in most countries. And then if we talk about devices and the memory that you build up about yourself and the health might be in the future if you what we see in China is they produce a lot of robots. What are they going to do with the robots? So maybe those are the and they don't have doctors not enough doctors.

So and they treat it very badly. So some of them get killed because the outcome of the treatment is not what the family expected. So make a long short story short, we see that uh in the future we might have those robots running uh health care agents with your history of the treatment for many years and so they can actually be the care for you at home and you don't have to go to the nursing home. So this is very looking very much in the future but given the speed of the Swiss patient record there might be earlier reality than the Swiss gazier.

Yeah. >> Thank you. Okay. So I think uh the person that asked the question about the role of smart contracts um is not here. >> Yeah but I answered that in >> in the break. Excellent. So Arash I think to close the federated learning and I think we will go to to the questions next you have as you said you have done a lot of traditional kind of clinical research but you've also done site NFD with a lot of patientdriven and patient le um initiatives for for research could you share a little bit you know from your old life what are what do you think are the barriers for the healthare system to adopt and implement federated learning but also from your new reality where can patient um a research

>> so from the old life is really limitation. Now mo most of the studies in the let's say 2,000 you know the major studies like 100,000 women like on menopause for example or or like menopause treatment in cancer uh in cardiology they have cores of like a 100,000 to 150,000 you know some genome project now have 200,000 like data set data points in the Middle East or I think Estonia is sequencing everybody but [snorts] you can't follow that number [clears throat] of people there's no hospital that I know in the world and I've been to like Most of them they don't have the ability to follow 100,000 women on a long time period and time scale that would be relevant to you know health as we said you know health is my

patient interactions are 15 minutes but what destroy them or help them is like a daily thing so I realized that limitation and so one of our idea when we created sign is like um any data point so if I do a clinical trial and I'm a diabetes because at that time I was a little bit obsessed by data consistency and robustness. uh my patient get an NFD file that encodes all her data and and then any doctor or any individual any citizen can have a query says I want to know who how many patients under that treatment uh had like uh that like side effect and then the patient would then contribute and the NFT would encode the consent for

that was my first idea and that would be smart you know you receive an email you were in a trial another doctor even unrelated to you maybe in in China has a question do you agree to share you know your data and that we and then I realized all of this is also BS uh patient don't really privacy is very overrated it's was created like it's a legal system I put my own genome on LinkedIn nobody downloaded I put a link to my genome I said like >> that is please >> who cares please do and get back to me because like honestly like I was at some point what is this privacy thing so I'm just going to put my genome that everybody is scared to share and like if

you find something just get back to me. Nobody really downloaded mine. So it's it's okay. I had to do the the work myself. But but uh I take a very simple example. Take a question like um did COVID vaccination impacted women's periods? And a lot of women in the clinic tell us that and they've seen so many other doctors that say no no I don't think so or it's you know it's coincidence. You just can't study this in a in a regular uh study manner. But if you have millions of women's reporting that side effect through any portal, you know, any portal, make it like an app for like women that have weird periods, you know, make it uh then you start getting federated learning,

you know, and then this guy can tap into it, you the Swiss system can tap into it, etc., etc. And I think that the future for me is educate people, empower people, give them power back on health, and they decide what they want to share. you know you want what they want to share like a name and what's your identifier is your name but you know you can have an email for that I mean there's a lot of way to anonymize data if that's important but I really think the whole and it's the problem is not a patient any of you in the room can give me send me a file I've been like people send me files of their like friends all day long I'm not the problem the problem is if like the workplace discriminate against you etc and you can you can protect

people from that but right now it's preventing researchers to even access data you know >> very interesting okay so we have uh some challenging thoughts here you know what because it's 10 minutes before our uh networking apparel we will skip the third um round of discussions about decentralized governance you might have questions about decentralized governance but we do want to open up uh the floor to to you right now for questions Yes. And I'm going to move from my beautiful chair.

Oh [sighs] yes, I'm too curious about everything. That's why I always have a question. So um there are a couple of different aspects that have been covered. So I'll try to formulate my question in a clever way. I might agree on the statement that sometimes privacy is overrated. And so we tend to make it more complicated than it could be or we create a problem where it's not there. However, at the moment we have two categories in the world. We have the people that might be more liberal from this point of view like me.

I'm okay in sharing this data with less restriction and but we also have other people that want to care about where their data are and then they want to have a local AI that will support the training of the data on site. In the opinion of the panelist, how do we communate those two words? This might actually go in the direction of a form of decentralized approach in sharing those data and making them free or private. How do you see a possible solution maybe 20 years ahead? I don't know. How much do we need to teach to the Swiss government or to the people to which direction do we need to take?

I I can start. I I agree with you. Some people don't want to share, but I don't know if they don't want to share because we've told them it's not good to share, you know, and I also understand privacy is very important for people that are buying drug on Telegram. So, I get that like concept of like, you know, like privacy can protect you from whatever. I I I fully get that. Um uh I I just think that most actors are benevalent. most actors, most people that want to do research. So, so, so for the risk of 01% or whatever, 0001% uh of people that might want to harm you, we we we kind of depriving 99.9% of people. [snorts] And also, I'm not sure

how much privacy is going to hold. Like, they hack everything. They hack the EPIN file. So, I mean like who cannot hack any hospital? There's no hospital system that has a robust uh you know, I mean, you guys are probably better, but I don't think the Swiss hospital, the French hospital can defend themselves against an attack of of China or Russia. There's no way they're going to defend themselves. So, it's a lot of effort for something that's not really worth it. And then, uh, you know, you don't want to share, you don't share. Like, you know, I share my location with 10 friends. It helped me so many times find my phone that I lost in cities. uh you know there's limit also to privacy.

So I'm not saying I'm not against privacy. I'm saying at some point it just like that narrative completely everybody's scared of sharing. >> Yeah. I mean it's interesting because I see what you mean but I also think that it's um we actually have many existing tools that we can put together to ensure some level of privacy. And especially in the beginning, I think it's pretty important to build that trust with the patients. Um, one thing you mentioned and I think it's also quite important is to uh have this be optin at least at the beginning. Um, so people don't discover that their data is actually already

being used uh for things they're not aware about. Um but but it's an interesting perspective to uh I agree that in many cases for instance I mean um in federated learning there are all those uh papers coming out trying uh to kind of break the system those adversarial approaches um that are very sophisticated. They're not realistic in terms of uh what it would mean to do in practice. But still we build tools to mitigate them and and so I think today uh we could already build pretty um uh strong system to ensure the privacy of

users kind of build that trust um and because I think we for all of those uh methods to work so for AI to really have an impact we need quite a big a large adoption and so I think It's a balance between the complexity and the trust that we're able to build with patients. So, yeah, I think it's it's kind of a trade-off that we need to to balance. >> I mean, there probably other questions, but maybe let me touch on that briefly.

Uh, it's important that you have a choice. Can it be private? Uh, I don't care about privacy. Sometimes you don't care. Then couple of years later you care and you have no chance of turning back the time. So in good old times uh when you enter the hospital you gave a you uh universal [snorts] consent of using your data for research forever. Nowadays uh some countries force you to have a casebyase consent uh of accessing the data but how on earth will you will you do that as a hospital? And this is I mean we're at crypto valley so we should talk about blockchain. So first of all

if data is decentralized you don't create honeypotss. Uh so hackers care about big hospital datas and they don't care about the data they just care about the ransom that they get that you can access your data again. So but if they have to hack 5 millions wallets then good luck. Um and so with blockchain you have some traceability and you can actually track when data is being uh created. There is a company in uh the US in Boston circular protocol.

They do exactly that. They just take a snapshot when data is being created and you can actually go back to that that hash that created or that address that created the data to ask them is it okay to use that data for that purpose and this is actually where blockchain can come into place. smart contracts you can use it for insurance cases if you fulfill uh uh condition X you get a payout uh of [snorts] 90% or only 60% so but it's very important that blockchain and decentralization gives you a choice uh and you can be reached in the system >> and I think the last piece of the puzzle

so you have privacy pres preserving technologies we discussed about because we looked at the smart contracts and you know giving individuals a choice of opting in opting out and then I think the the higher level we didn't we didn't discuss too much is decentralized governance so the dowo type of um initiatives where you know individuals can opt in at at population level in a way so at a community level to decide what is important what is why do I want to give my data for cancer research okay are we all in agreement that this is important and then I the individual choice becomes a little bit easier if we slowly [snorts] slowly learn from other industries how to do decentralized governance and that was again we didn't

have time to discuss but I think it's it's it's super interesting um would you like to close with one or two words and hopes from your side so if you think 2030 comes and healthcare is a little bit more decentralized maybe not fully decentralized what is your hope for um for 2030 30 what can be achieved? >> Um I I can I hope the Switzerland will figure out their file their vaccination. That's a hope for you guys. [laughter] I'm not concerned, >> but like >> uh uh No, I really hope that we can prove that um [snorts] I mean with with what we created on the app, I I hope

that we're going to be able to prove that millions of women can build a community of like smart researcher and they can come up with their own question like questions are always asked by us as doctors but but women have a lot of question that we never think about and so I hope not only to see patients sharing ing data, but like women right now most of influential health like I would thinker are on Instagram and Tik Tok and they're not doctors. They're normal people that like put posts. They they're followed by millions of women.

There's no doctor followed by millions of women in the world. So, so I hope like that we'll see in 2030 women asking a question and having a a million data set or platform that help them answer that question without any health care provider being even involved in that. That would be and that we also destroy the scientific publishing system which is only like pay walls pay walls and completely not accessible to to to the general public. you take a cancer treatment, you never have a lay, you know, lay summary that somebody who might have that cancer can read and understand. So that that that's my hope and I with AI, I think it's going to be very very feasible.

>> Go ahead. >> Right. So on my side I think it was actually mentioned during the previous panel as well uh this idea of human centric uh approach and I think it's really important especially now that we have this all this hype around AI and all those uh crazy projects around it. Um I think it's uh important that we not only focus on this uh human uh centric approach and also try to shift the incentives from uh profits to uh the actual needs of the patients and I think this is also um one thing that uh can be

done with uh decentralized AI. Um and I just wanted to touch on uh on some example cuz also the previous discussion uh reminded me of this. Um um at FL we've been working with um a collective or foundation which is called the VA foundation. I don't know if some of you are familiar with them. Um and essentially what they uh noticed is that there are all those uh companies that gather data uh on users and users have absolutely no um power on this or no ownership. And their goal was to kind of

flip this and um instead of uh those companies taking their data, they could actually um uh download the data that was being being collected on them and do useful things with those and benefit from them. And I think uh this uh resonates with what was discussed as well as uh giving the patients some benefits um from the data that they're willing to um use to participate in different uh initiatives.

>> So great I have the final saying now today never managed to get that at home. So, uh, in 5 years from now, uh, I still hope that I don't have to care so much about my health data because I'm hope I'm still healthy. But nevertheless, uh, I think that I can carry my data. I have a copy of my data at least. Uh, so I can talk to my data and my data can talk to me. Abra, you should take your next flu shot. even if it's on paper, just get over it. And so, uh, that I can talk to my data and the AI using my data can talk to me and guiding me to a healthy

life. >> Everything goes back to us, [laughter] the flu shot. Yes. Thank you very much. We are five minutes over time. I will probably um Stephanie, do you want to do the closing? >> Okay. because in reality we are almost done. We have an apparel. I know some of you have to go back to to Zurich. So um thank you very much for for attending very few things. Uh the report is still available if um you haven't taken a look at it. Um of course look at the follow us on the organizations because we're all not for-p profofit organizations and and uh community is very important. CVC crypto valley conference is next year.

So it's it's a very very interesting um conference web three there will be healthcare related topic but of course not not so much MSF has a beautiful Instagram if you guys use Instagram there are beautiful stories of care with what their doctors are doing in in um in crisis situations on the ground and yes Olga is still here so if you want to grab her to understand what what you can do in the western chapter of CVA uh please do so on that note thank Thank you very much to our amazing panel.

Round of applause. And I've been told I should sit here so that we also get the the recording and you will see the recording soon on the CVA and uh and MSF I think channels. So if you want to share it with someone or to just refresh your memory um do that and please join us. Mario, we are going to have some upper row. Just follow Mario. Mario is here. And thank you again very much for hosting us, Mario and MSF. Thank you everyone for joining. >> Thank [applause] you.