Presenter: Out of Washington DC, where we look at policy and financial sources, and the DC Institute out of Zurich, where we look to help standardize how you do programming.
Now let's get right into DeFi and DeScii, and what's it all about. So, Vitalik, can you share a little bit your definition of what decentralized science means and what we can use it for.
VItalik: I think decentralized science is a very big label that covers potentially a lot of different things. And depending on how you define it, you can define it in a way that focuses on the blockchain parts. But the decentralization is bigger than blockchains, still. So you could even ask, does Wikipedia count as a DiSci project? And maybe it does. So I think within our community, there have been this collection of people that have been basically trying to ask, with the tools that we're working on, whether it's blockchains or DAOs or ZK or whatever else, are there ways to try to apply some of those technologies and at the same time our community has to really try to push forward different kinds of science that we care about. So, I mean, just like using coins to fund science is, you know, it's a thing that's been, I think, done for about, like basically, I don't know, over ten years now. I've definitely donated coins to science myself. There have been coins and there have been NFTs and similar things like that. Just groups of people organizing to fund things. Then we can talk about using some of the cryptographic tools that we have to try to collect data, which is very important for a lot of medical things in ways that protect our privacy. So I talked about some of that yesterday. And I know that there are projects already working on that. We can talk about DAO a bit and methods lots of people have been experimenting with as a way of organizing people, initially organizing people to invest together, but I think now over time also organizing people to do things together. And there are projects like VitaDao, and there's an increasing number of these, and basically trying to organize around supporting particular types of science, organizing around supportive trials, organizing around the specific treatments that they care about. So a lot of interesting work there and a lot of work on that is public-grants funding projects. So we've done things like, you know, people in grants, optimism funding, with all of these public-grants, decentralized public-grants funding technologies that we've developed for use in kind of the agrarian system that are often very valuable to our need for some science as well. There's something like all of these things that I considers DeSci and just anything that just helps empower communities to work together to support science that they care about without everything having to be, you know, a centralized corporation.
Presenter: Less centralized thinking is less single-threading. Funding, being able to open up to more communities, that's part of the things that you shared, right?
Vitalik: Decentralized science, Yes.
Presenter: And let me turn it over to Tuan.
Share your definition, what it's all about.
Tuan: So first of all, thank you Vitalik for coming here and talking about decentralized science. So I think one of my main objectives before I give my own definition of decentralized science, I would say that the keynote Vitalik gave yesterday is awesome. I would love to see the next keynote Vitalik gives in AI and blockchain, then Vitalik will actually give more examples of AI in decentralized science, please.
Yeah.
Will this happen today?
Okay. So, another way to define decentralized science, let's move back one step. So I've been, we've been working in science, like centralized science for many years before we actually started to decentralize science. Okay, and then we can hear a lot of extremists in decentralized science saying that, okay, decentralized science will replace science someday. So I would love to share some of my views on that first, and then we can continue the discussion. Yeah? So I think centralized science is still doing okay. It's like the Apple iOS, right? It is beautifully controlled by the monopoly, but it's still working, okay? And then I can see that decentralized science is like Android, okay? It's a moment Google publishes Android to the world. There's so many different developers who can actually go on top of Android. So that's why we have more Samsung phones and many other phone operating systems, not just IPhone. And I think that's centralized science. And decentralized science can grow one side by side with each other. And we can move to any other example later. Thank you.
Presenters: So having decentralized science and science to work together, to work side by side. You see an ecosystem where they are collaborative.
Tuan: Yes. Yeah, so for example, a lot of, you know, like in science, a lot of research is happening right now in university or at big corporations. But the moment we see that the definition of a decentralized science or decentralization is just basically put users in control. And so right now, a lot of research institutes or universities or corporate are actually actively doing that. More engagement.
Presenter: Talking about engagement, he said that getting the corporates involved, let's talk a little bit about the financial aspect too. You mentioned that sometimes a science project, especially outside of the major mainstream, is hard to get funding for, hard to get data as well. In today's context, what could you look at as a decentralized project? What's the framework we should think is a practical decentralized project that someone could be working on?
Vitalik: I think there's a couple of different patterns that make sense, right? So one of them is just, you know, this kind of taking a lot of therapies, treatments, or ideas for treatments, kind of over the hump from the beginning stages, where often we have pretty good tools for funding people or encouraging people to work on them already, but at the end stages where it can just be taken over and pushed forward by the private market. If they stick between the two areas, often this challenging area has a lot of projects, actually not even just in things like biology. A lot of areas, including Ethereum as well, have a hard time getting over. Trying to organize together and have a community and push projects through that. I think it's really the key. So the question is what kind of organizing do we need? And so one area I did was not always bringing together funding. There is a pattern that exists in Ethereum, too. If you give $100,000 and have a good idea, there's lots of people that will just give it to you. But a more important point, when version one of the it takes $100,000, there's always the version two that we need to bring into the real world, and that often takes $100 million. And where does that come from? That individual, actualists, for example, they can get into 100K, but they can't get into the millions. And having communities come together and pull that kind of funding. We've seen examples of this in the corporate space. Even just some things like Constitution DAO, that got over $40 million. So I got to bring these funds together to support what treatment fund that scales might create. Another one is, another dimension is the number of users. There's so much that you can do with just working with just yourself and 500 people. And there's more that you can do if you actually have a sizable community that can try things and give you more information. And so there's been a lot of excitement around decentralized trials. And that's something that can be done. I guess I mentioned yesterday, which is there's a little variable risk version of this, like just assign people to random, try different styles of changes, and then either, you don't even need to get those manual report information, you can make it a smart watch app, and just pass that information over, and so, yeah, doing more of those kinds of things. Exactly. You're not using that participation of, someone doing some small experiment to get your data. And then you use things like DAOs and other tools like Action and Love, organize stuff together, and talk to some people. If you can even have a mechanism where people can first express interest and only get the best people to express interest then what a switch on that would turn on by some group of people if not, by the time it starts. So it is like that. And if you can do what he is doing, like medium scale things to get medium scale data about some kind of treatment or lifestyle change or technology, then that by itself is also a good thing. You cannot, like you said, that's a bit more of a, you know, as an artist or, you know, mainstream scientific research bodies can keep pushing it from there.
Presenter: So bringing together mainstream, kind of, from the quality of the investment in decentralized science back last year, for example, playing part of it as well as you have to. So take me a little bit into the biological side of this. Talk about biology experiments, as sensitivity in terms of medical records is so high, and yet during COVID, every government around the world asked us for our very sensitive data to come into their countries, right? We've also been on blockchain and what have you. How can decentralized science kind of help us with the privacy aspect, as well as you'd be a useful way to support experiments that Vitalik was just talking about?
Tuan: So that's a very nice question. But before that, let me give Vitalik an example. Okay, because like, give me an example right now. So you are now here in Vietnam. Three months ago, we published the largest genetic study for autism for the Southeast Asian population. Let me explain how we did it. So for five, six years right now, we built what we call the Life Network. I can explain it in more detail later. But we have more than 100,000 users actually participating in this hour. So for autism research, we partner with a central hospital here in Vietnam. We do research with them. They give us 300 samples for the autistic children. And then based on that sample, we actually characterize the criteria who is qualified to participate in the research. And then we send back requests to hundreds of thousands of our users, asking, do you satisfy this criteria? We incentivize them, you join us, you participate, you get the reward. And immediately, 20,000 people get back. And only 5,000 people actually satisfy the criteria. We used that 5,000 samples, we actually did a study and it's the largest genetic study for autism for the Southeast Asian population. Our chief scientist sitting right there who's been working on that for like three years and basically not only autism, in Asia we can do a lot of any other research which was only like mostly done by the US or the UK, by the Western. So right now we do a lot of things for Asians. You can see diabetes, okay, in the US, a little bit obvious, they may be diabetics. If you come to Asia, you see a lot of thin people are diabetics. So right now we're also running one of the largest genetic studies in diabetes. So to answer your question, we build a platform for what we call participation to earn. You can participate in one of the research and you earn the reward. So that was just trying to give Vitalik some examples. So next time he'll talk about AI and blockchains, probably he can mention something to help the Asian population to advance the research. mostly done in the US. So our team is actually based in the US, from Cornell University, from UCSF, from Stanford, and Harvard Medical School. And when we came back here to Singapore and Vietnam, we seized a new opportunity for the Asian population to do advanced research. And decentralized science is a mechanism for us to achieve that objective.
Presenter: What do you think? Is that what you just shared as an example? I mean, some of those framework tests that you put up. I mean, some of those framework tests that you put up. And I think that's what it does. And is there anything that would make this, because we talked about an audience, there's some kind of right feel that this disconnect, in three years, you've been working through all this great experiment. It's been around decentralized health and so many opportunities, and yet the audience here didn't really say, hey, I've been involved in it. How can we bring this culture together? How can people know what it is, get the benefits out of it? What do you think?
Vitalik: So I think, yeah. Good question. I think we have one part of the answer just is we have to be OK that these things are going to take some amount of time. Right? And I try to remember, like, if I remember zero-knowledge proofs, right? Who here has made zero-knowledge proofs in the past year. Right? But used in applications that involve the sk snark somewhere. Now, who here did that more than five years ago.I mean, the idea that they have sent us is a question of action. So even when you have things that are in the world of ideas, often things take time to really start intersecting. And a lot of people are the kinds of people that come from the mindset of actively thinking about biology, medicine, and health, and the kinds of people who go International to think about software solutions, those are very specific groups of people. And the intersection of those two is an even more significant group of people. But even more specific with people. It's really, really difficult. So I think we just want these kinds of events to work. And just like activities, just try to intentionally bring them and their groups together more and help them understand each other. And so one of those things that needs to happen. I don't think there is an easy solution. I just think there's lots of people talking to each other.
Presenter:And I like that. The intentionality to really share what people learn here today, take it back, try it out as well. So it's about trying to get a framework together and look at the current projects that are in the DeSci as well. Let's take it a little bit further. You mentioned AI and genetics. How can AI, blockchain, sit together, work together, what can we see and think about for the future of decentralized science with AI and blockchain?
Vitalik: Yeah. The most natural one that comes to mind is finding ways to encourage people to provide data that has to do with either specific conditions or specific groups of people like data that otherwise can not be defined, so that is necessary for training models.
Tuan: For AI, I think I talked a lot. But 30 minutes is not enough. I'll give you a couple of examples. I love to give examples. So in AI, right now, because of the amount of data we need for training, right? The amount of data we want for the user. So AI is one of the mechanisms to apply to build what we call the autonomous agent, right? To protect the privacy of the data and to respect the ownership of the data, the AI can work as an autonomous agent, so nobody will actually access the data. ZK-Proof works really well if we can make it work with AI. Another way, when you talk about data, okay, we've been working on a lot of different backgrounds. When I was at Google, I was in a team, we built the largest global database in the world. Right now, it's a database And so the way we look at the data is a way to use the data for training. And the most important thing is clean data. So Andrew Ng, the founder of Google BigMind, or BigMind, said that 80% of the time an AI company actually spends is to clean the data. So that's why we have to build a system so that when we bring the data in, we need a very early way to clean up the data, and the AI can actually work a lot better than humans. Okay, so I can give you many examples to see how AI can actually involve in basically data cleaning, data curation, and actually labeling the data. Okay, so really hope that the next topic we can go in depth about AI in decentralized design.
Presenter: What do you guys think about AI and using blockchain to protect the security of the training data? And so when I want to send sensitive data, genetic data or critical financial records from my data over into an AI. One thing about blockchain is that it's immutable. It's guaranteed delivery. It allows us to make sure that it actually gets there as it's been changed. So perhaps, can blockchain act as a necessary feature, a required feature almost, to make sure that the training data we provide, like you were just saying, is actually getting into the and secure.
Vitalik: Yeah, I mean I think that Potentially they're valuable, though the kind of security that people talk about when they talk about data is like, it's not the same kind as the kind of security blockchains provide, right? Like when we talk about data, we're worried about privacy, we're worried about the data not falling into the wrong hands. When we talk about blockchains, blockchains are more about, like, hey, make sure that Microsoft or whoever else can't edit your spreadsheet. And Microsoft will not be able to edit the spreadsheet, and Microsoft will not be able to see your spreadsheet or your things. And so in this case, I think that some of the stuff around zero-knowledge proofs and combining zero-knowledge proofs with watch-ends to get both kinds of guarantees at the same time. Also, MVC is a big one for training. I think that could be super valuable. So, yeah, basically, I think, you know, there really is this space of privacy for so many, but a lot of these solutions have definitely worked out and there are particular ideas that you can offer that says that those kinds of things are absolutely going to be important.
Tuan: And one of the clear example, I would love to give example. So for example, an AI model, they say that okay, that you're training on a million data records. There's no way to verify it. But if you apply ZKP, and they say that okay, this model, I'm going to put this model, it will train on a one million genetic profile. Using ZKP, you can prove it. So that is one way to preserve the privacy, the data security of the user without actually revealing that data to the user of that AI model. Okay, so that is one example. Another example when Vitalik was saying that, okay, how about the data owner. The data owner holds the encryption key and that is the only key to access the data. Going to a more technical detail, any of the engineers in the audience can check like proxy re-encryption. That is one way for the data owner to approve that Vitalik can access my data. I give the approval for Vitalik without, actually revealing my private key for encryption of the data. And now Vitalik can read it and receive all of the transactions on the blockchain to prove that I give the approval to Vitalik. Vitalik can read the data without getting my private key. They can execute within ZKP or a trusted execution environment.
Presneter: Yeah. So if you look at it, a trusted execution environment perhaps protects information. One of the critical issues that we have in scientific publishing is that failures aren't polished. Can you share a little bit about this piece of how decentralized science enables science to move even faster if we start to share those failures?
Vitalik: Right.Yeah.I mean, I think, well, even outside of the DeSci space, the concept of pre-registering studies is definitely a thing. And people have definitely started doing it more and more over the last couple of decades. And with cryptographic tools like that, that's definitely something that canbe made even more robust in all kinds of ways. And then I think there's two aspects to this. One aspect is just that you're finding that people are being honest and not doing the thing where they could have done it through a hundred parallel studies and it would be published, the one that gets you a 99% confidence in a high random chance. And pre-registration deals with that. And if you pre-register cryptographically, then you're going to have an on-chain record of how many studies that were attempted. And so you'll be able to see, is this one study, should you do 0.99? Or if there were 100 studies you should do 0.99, right? So that's one piece. And then the other piece I think is that if a study is done and it ends up not being useful for research, well objectively there's still data and analysis in there that may well be very valuable for other work. And finding ways to try that, I encourage that that be available and just making that information more accessible so that it just doesn't get thrown away. It's something that has not been thought about as much but could be very valuable to.
Presenter: So much information trapped in centralized labs that could be made visible. I want to bring us up from the depths that we've gone into and wrap up here around the aspect of how it will affect everyday life. What could we see as a future of maybe visiting a hospital or handling our parents? Can you share a little bit about the future of decentralized science?
Tuan: the definition I always put, in my case, like putting the user in control. Okay, so users have the true ownership of their data. So what I see in like three, five years from now, if you are walking into a hospital in New York City, right, see the person open a mobile app, let's say the live app, and then the mobile app actually talk with the doctor and say that, OK, based on the user's genetic profile and all of the personal data, the user doesn't need to fill out a form. So the app talks with the doctor and comes up with a personal life optimal treatment for that user. I see that also if you walk into a restaurant in Singapore, right? You will see people taking the mobile app, they look at the menu, and then they say, Hey, if you're in the weight loss program, and based on your genetic profile, here is the food in this restaurant you should order, because it's optimal for you. So everything becomes personalization. You own the data, and it gives you, let's say, a part of your life you would not imagine. It's like everybody puts out a mobile phone and gives that person a personal life assistant and that's how it changes in life. and that's how it changes in life.
Presenter: So our mom points up on that, hey, don't eat that. Follows the phone, her phone as well. Oh boy. Okay. Good time. Put it in there.
Vitalik: Yeah, I mean, I think it'll be like, we'll see you in lots of different groups of people in different ways. Like one is that absolutely everyone is going to benefit from just better kinds of therapies being available. And hopefully, I think, the DeSci space can also be more globalized and more open than the traditional science spaces. And so, you know, treatments that are available, hopefully also become available even in the parts of the world that centralize medicine is leaving behind.
Presenter: Last 30 seconds. Let's wrap it up. I'll give you the last word, Vitalik. Tom, your 30 seconds. What do you want the audience to leave with?
Tuan: Actually, the audience here is great, but I would have a request for Vitalik. Okay. And we are scientists. We've been working to centralize science for so long. And whenever we bring the cryptos of blockchain technology to centralize science, they are very conservative. They would love to hear that from the leader, from you. Could you please, to go and talk a bit with, go and talk with, like, CEO of Google, to say that, or to say, tell them that blockchain technology is somewhere like this centralized science. And maybe, but now the future of the direction, we signed to centralize science.
Vitalik: I will have those conversations in the future. I'll keep that in mind. Oh, I think that... I mean, one is just called action, I think, to anyone here. It's going to start to try to look into this space and let you guys identify, like, is there something that you can participate in, like, actively? I don't know if you can even use us in a small lab, right? Is there some doubt that you can be part of, like, some of these trials you could participate in, or you can just, like, help, like, find the particular thing. Like, anything that's small, and ideally, I think, anything that's persistent, so you don't just do it once, and then go away. Then, you know, you can actually use that a lot and learn about the space more, and, you know, if you like, find something or less to go a bit deeper.
