82. Future Sci is DeSci: SVN's Michael Fischer on Memefying Science Breakthroughs

[00:00:00] Ray: What's up everyone? Ray here. This is episode 1 0 9 and the first Health Unchain episode publicly published in 2023, so happy new Year everybody. I have a couple quick announcements I wanted to share with you before we get started with the episode. Albeit decide London on January 15th through the 16th to meet with some of the leaders at the quickly growing decentralized science space.

[00:00:23] Ray: If you're planning to be there, let me know so we can meet in real. Health Unchain has launched a super cast premium membership community where you can watch interviews before the rest of the public gets to listen to these conversations. You can find a link to our super cast website in this show. Notes Health Fund Chain is a media partner for the blockchain in Healthcare Today, symposium in New Orleans, Louisiana in September, 2023.

[00:00:49] Ray: If you're interested in buying tickets or sponsoring the event, please reach out to me so I can coordinate with your. In this episode I speak with founder of DB Dow and leader of the Desai, New York City community, Michael Fisher.

[00:01:02] Ray: Michael is a Stanford University computer science PhD graduate, and author of the book Regulating ai, what everyone needs to Know about Artificial Intelligence and the law. We talk about the value of DESI and the details of using decentralized databases for storing and managing scientific data more permanently.

[00:01:24] Ray: This was such a great conversation and I'm thrilled to be able to share it with you. I hope you all enjoy it. Remember, the Health Unchained Podcast is for informational and entertainment purposes only, and we are not providing any sort of legal, financial, or medical advice. Please do your own research and due diligence before making any important decisions related to these matters.

[00:01:42] Ray: And now let's get to the show.

[00:01:44] Ray: Today's guest is Michael Fisher, who is the founder of DB Dow, and they're doing some pretty cool things in terms of decentralized database building. So I'm gonna hand it off quickly to Michael to introduce himself a little bit more in detail and share some of his background as well.

[00:01:58] Ray: And then we'll dive deep into what he is building, a little bit about the DCI community and how that's growing and all things NFTs ai. So I'm really interested in what you have to share with us today. Michael. Thanks for.

[00:02:08] Michael: Thanks Ray for having me. I'm happy to introduce myself. I started ddo and before that I was a uh, PhD student where at Stanford, where I was studying computer science and natural language processing. And then kinda after the PhD I finished it. I moved out to New York and started DB Dow just to my interest in The blockchain and ai also decentralized science.

[00:02:32] Michael: So, you know, as a PhD student, I had a lot of experience, writing papers and conducting research and coming up with new ideas and collaborating. there was ways to do it better blockchain smart contracts and stuff like that were really an important, step in being able to conduct research in a more open and collaborative and democratized way.

[00:02:53] Michael: prior to the PhD uh, was an undergrad at Stanford also sitting computer science. I've written a book on regulating ai. Was the co-author of that based on the class I was the teaching assistant for at Stanford Law School. now doing blockchain.

[00:03:07] Ray: I know your background is in computer science, so you must have heard of blockchain many years ago or, or Bitcoin or something. What was like your first experience or exposure to blockchain technology? Do you remember?

[00:03:17] Michael: Yeah. A couple years ago I started this uh, website I saw like peer review as being something that blockchain could particularly help out with a couple years ago. And early on, I I created this a to rally or a role, but for archive papers.

[00:03:32] Ray: Hold on. What do you mean by that? A rally or.

[00:03:34] Michael: oh, so it's a, like a creator token. the idea was that there's a lot of papers that are published on archive and the issue though is that there's no peer review, if you sort of are just looking at papers, you don't know what's a good paper and what's a bad paper.

[00:03:48] Michael: there's a couple different ways to sort of solve this problem. One is to use traditional peer review. One is to look at sort of metrics from Twitter or, see which papers are being tweeted about a lot, and use that as a signal to determine which are the interesting papers you should read.

[00:04:02] Michael: I was trying to think about what would be like another alternative way review papers. And so the idea is that every time a paper was published on archive, what I did is I automatically created a script that would like create a token and it would issue like a hundred tokens for that paper.

[00:04:19] Michael: And it would assign half the tokens to the authors of the paper equally. And then it would assign the rest of the papers, which would be sort of on the open market. And the idea was that, . If you liked the paper and you thought it was good and you thought it had potential, you would buy this token because you'd think that in the coming years this paper would prove to be correct.

[00:04:40] Michael: And then you would sort of be incentivized you would have to be the holder of this token and you could then trade that token for something else down the road. And the idea is like, cuz you had insight and it's just like thinking about other incentives. Cause now there's like a marketplace of people that are sort of the adjudicators truth on these papers.

[00:04:58] Michael: of course, the other half of the tokens are for the authors. So if their tokens are worth a lot by some definition then they get more reputation. So I built that out a couple years ago using like really early. web three tooling when it was still very nascent and there wasn't as many L two s.

[00:05:14] Michael: And so of course like it was very expensive to operate cause you had to reissue these like smart contracts every time a paper is done. So in many ways it was much more of like a proof of concept and getting my feet wet with the underlying technology and taking it through the incentives and mechanics.

[00:05:30] Michael: that was for my first experience with Yeah. And you know, when you talk about having papers that are impactful essentially is what you're saying, and there's a, thing in science, it's called the impact score. So every paper that's published can be measured on how much it's used in future studies as well and how important, the insights are from that study.

[00:05:49] Ray: So I'm imagining like tokens that you generated over time, if the impact score of that article goes up, the value of the token kind of goes up with it. And the people that initially believed in the paper and it made sense to them had that true like awareness or insight. So they're rewarded for being able to make that call early on.

[00:06:08] Ray: That's kind of the general idea.

[00:06:10] Michael: Exactly. And we'll talk about this more later, but there's money is like an interesting thing in how it relates to science. And sometimes a lot of scientists or projects don't actually want money, but they want sort of impact certificates or something along the lines.

[00:06:21] Michael: But what it's interesting is that you're using sort of market dynamics and economics to incentivize the finding of truth and the finding of good, papers and not metrics. Like oftentimes people cite other papers to increase their score or try and. Game, these impact things or this is why these Twitter thing doesn't really work either, because some things might just be clickbait or published well or something, or have like a meme attachment, but that doesn't really create the good underlying science.

[00:06:47] Michael: So the idea here is obviously like a good impact score, like what you're talking about will probably be some sort of co linear combination of metrics on social media, this economic incentivized finding of truth, the peer review and the impact score that you're like traditional impact scores.

[00:07:06] Michael: And then the idea was like, I think in the future of what you could do is like based on the, what you're trying to study, you could then dial in on each of these different metrics to get to surface the papers that are interesting to you or that uh, are needed for like, if you're just doing like far out there research on some very esoteric topic, maybe you want something different than.

[00:07:26] Michael: If you're doing like drug discovery and there's a lot of risk involved or something. the idea is you could dial in the score and the specific papers that you want based on these different types of

[00:07:37] Ray: Interesting. Just wanna take a step back as well, a little bit. You mentioned you moved from Stanford to New York City and there you're starting to build a community around DESI as well. Desi NYC is a organization you're leading and I know that had a event last week as well. Do you wanna talk a little bit about that?

[00:07:52] Ray: I'm just, you know, I wanna make sure people listening, if they're in New York, they have an opportunity to meet you in person, be part of some of the events you're hosting.

[00:07:59] Michael: I first learned about desig and learned about it an in-person event at um, funding the Commons. And after, I sort of marinated on that idea for a little bit. . And I realized that the reason I found out about was through this in-person event, I was able to talk to other people and there wasn't really anything that was hosting these types of things in New York City.

[00:08:19] Michael: And New York City is in many ways a great place to host VCI things because VCI is about bringing people with disparate backgrounds and different experiences. And New York City is in very much weighs a melting pot of people and ideas as well. So it seemed very natural to start hosting these things.

[00:08:35] Michael: we hosted one and it's basically like two speakers then a lot of time and then time to speak. So we have like a, someone who's focused on science. So with our previous one we had someone who was ologist, someone who studies fungus and they're using this website to find different and identify different mushrooms and name them.

[00:08:52] Michael: And they have this interesting in Way of determining what is what and people voting on it, which seemed interesting. It seemed like it could be a crossover with decentralized science. And then we have one on like decentralized science or web three. it's just a really good way of bringing people together spreading the word around eci.

[00:09:09] Michael: And of course, everything is all recorded and posted online. And if you wanna find out basically about the next events, we'll post them all at db dow d ao xyz slash dai nyc, D e S C I nyc. Maybe we can put a link there or we have a link on our website if you go to dbw xyz and there's a little tab for PCI nyc.

[00:09:31] Michael: And then we have all our events there.

[00:09:33] Ray: Awesome. I'm sure people will, I'll take a look at that as well. And you know what's funny about science? To me it's always been to some degree a. Decentralized thing, science in general, because if you believe we're all scientists, we're all conducting experiments, isolated or not all the time in a certain way.

[00:09:51] Ray: Sometimes you even heard about discoveries being made in different parts of the world around the same time as well. So there's this idea of science is already a decentralized thing. However, in the last 30 to 50 years, it's sort of been the traditional system has centralized the process around publishing science.

[00:10:07] Ray: So what decentralized science is trying to do, what the movement is trying to do in my view, is trying to take back that publishing centralization and make it decentralized. Which is really tough to do because there are gatekeepers in terms of what should be published, what shouldn't, we don't wanna publish, quote unquote science.

[00:10:27] Ray: That's not really well. Thought out design experiments, cuz then you have maybe even the generation of people following those systems or, experiments. And it's just creating bad science on top of bad science. So I do find what you're doing very important overall. I think the movement is incredibly important and it'll take a lot of time before we reach like a potential that we are, thinking about or we're both envisioning here.

[00:10:48] Ray: But maybe what you could do now is sort of describe some of the problems in the traditional current system of science. And you kind of talked about it a little bit already, but things like research funding, patents, ownership royalty distribution. And do you think that's part of what DB DA is doing, or, what you're doing is

[00:11:04] Michael: what's interesting about Desi is it relates to publishing, but one easy way I think of describing uh, sort of the power of DCI is imagine a world in which you are a programmer. Before GitHub or before Source four, if you, one of these sort of open source websites, you had to recreate everything basically yourself.

[00:11:22] Michael: So if you wanted to build something, you'd you'd have to rebuild the library, and there would be no, like cross collaboration, like coordination cost would be extremely high. If I wanted something, might have to email someone and they might have to email me, doctor Source. And it just like increased the time of, doing programming. But with DCI they were really trying to like lower the coordination costs of conducting collaborative research. what that means is instead of having to redo an experiment yourself or redo it, you just opt into a protocol that exists that is on chain.

[00:11:53] Michael: And the idea is you can sort of speed up science by creating these standardized formats or data formats or schema or data pipelines or Jupyter notebooks or all sorts of various things. So I would say like Decise is around using like sort of open standards and web three technology, decentralized.

[00:12:10] Michael: So storage, decentralized compute, decentralized funding decentralized publishing, decentralized incentive mechanisms. To increase the speed at which science is conducted. then in answer to your question about DB Dow DSI runs the gamuts from funding research building teams.

[00:12:27] Michael: if you need to build a team, it would be helpful to just know, know who else is an expert in that area. Collecting data in a standardized format, sharing the data in a standardized format, visualizing the data, analyzing the data, publishing the data. And then, DB Dow relates to all the ethos of dsi because what we're trying to do is we're trying to create an open database standard for conducting science where the community comes around certain data, ches for different types of data points, and data is sort of uploaded into that and data is shared.

[00:12:57] Michael: Between labs and there's unified name space where people can access data and link data between different databases. Data is preserved forever, so a lot of problems we're having with some sciences, the data disappears after 50 years. You know, one knows where it went or you can't track the provenance, which machine it came from.

[00:13:15] Michael: If that machine has been calibrated properly who is the researcher that conducted this experiment? it all gets lost. one of the nice things about the blockchain is that all that data is immutable. So in, hopefully in around 50 years, I P F S and the Ethereum main net will all still be around.

[00:13:31] Michael: you know, you could just go back and re-poll that data and validate that a certain type of experiment has been conducted. So at Dbw we're sort of creating this sort of immutable database infrastructure. For researchers to share data and conduct scientific research

[00:13:48] Ray: Awesome. So I have a bunch of questions related to that. So, you know, many will argue, do you really need a blockchain to store all this data? And might have an answer or something like, well, we're not putting all the data on the blockchain. It's just like a pointer to the actual data and it's just managed that way.

[00:14:02] Ray: But you can clarify that in a moment. I guess my question really is why do we need a blockchain and why can't this be done by a centralized entity,

[00:14:10] Michael: one thing that I think is interesting around how we conduct one is, the first is around incentivization. I have this concept of like negatively priced data. So what that means is every piece of data, like right now, I wanna store a piece of data on s3. I have to pay a small amount of money.

[00:14:26] Michael: But what, what really should happen is I should get paid a small amount of money so I should only pay if that data has no utility, if that data has negative utility, . if that data has any amount of utility, let's assume that storage costs zero. Like at the moment, any amount of utility that, that data creates should be paid back to the person that created it, or the incent or the overall, the overlying hyper structure that created it. if you need to pay these sort of small amounts to different people, in many ways, the easiest way to do that is through the blockchain and through using, one of the things that blockchains are really good at is paying people small amounts of money in a cheap, easy way across different borders and everything.

[00:15:06] Michael: Like, and when I say money here, I don't mean money, like dollars per se. It could also be impact certificates. It's just like some scarce digital asset that needs to be transcended. So the first reason you need a database, On web threes, you wanna be able to pay people for the data that you create.

[00:15:21] Michael: The easiest way to do that, you don't wanna link everyone's credit card to a database. That's too time consuming. So the easiest way to build that, given that we wanna incentivize people is on the blockchain. The second reason is around data permanence, especially for conducting science. You don't want this data error disappear.

[00:15:36] Michael: The idea that a hard drive crashes is very web two or web one. if people choose to have their data live forever for, and they think this is important for scientific applications, it should. So you want data permanence. You want data composability. A lot of, like now a lot of research labs conduct different experiments.

[00:15:52] Michael: the reason you can't link those two databases together really easily because they're on different name spaces, they're on different servers. But really you should just be able to link a hash of that one piece of data and link it to another hash of another piece of data and have. Cross database linking.

[00:16:08] Michael: So if I could say like I wanna reference some experiment or some row in another database, I should just be able to click on that, row and link it, and then put that link on the blockchain and have cross database linkage. You wanna be able to also sort of opt into the greater ecosystem of token omics that exist in web three.

[00:16:28] Michael: you wanna be able to, in that sense, tokenize data. So every row should be able to be bought and sold in the same way in N F T can be bought and sold. So if someone has a very popular row in a scientific database and they wanna get out of science, they wanna buy a house or whatever, they should be able to take that nft with that genome or with that.

[00:16:48] Michael: Image in it. And instead of selling a four, eight on open C, you sell this image of this new thing that you have created, that it's generating a lot of revenue or ip. And so you wanna be able to tokenize all the data that exists and then use that in these like larger defi protocols.

[00:17:04] Michael: Cuz any, right now, a DPI protocol just takes in an eer, an NFT, or an E R C 20 and then applies all this different leverage and staking, but there's no reason that that can't be done with data Instead of just future promises of an underlying token you wanna be able to encrypt the data and store it in a way where you own the keys.

[00:17:22] Michael: So maybe in a lot of examples, I might upload an x-ray or my genome or a health record. I wanna be able to protect that. not your tokens, not your keys. And you have same thing, you want to be able to uh, do it with exact same principle with data. So you'll be able to encrypt it yourself and only grant people access to that.

[00:17:40] Michael: And it's a little bit harder to do on web two traditional. You want be able to tie it to your wallet. So your data should be tied back, you know, identity is changing and you wanna be able to tie your data to your specific meta mask wallet in either anonymously, pseudo anonymously, or with your real identity.

[00:17:57] Michael: And then lastly, like everything is going in web three, we just need a web three alternative to traditional Mongo Postgres Firebase type databases where it interop operates with multi SIGs wallets and sort of the larger ecosystem.

[00:18:13] Ray: and that's what DB dad was trying to do, is trying to create that web three version of Mongo db. and it's quite a task. I think we would both agree that's a, big mission but a very noble one, I guess I would say in terms of trying to allow for users to build something that's decentralized. I a whole like database system that is decentralized.

[00:18:32] Ray: Maybe we can go into what do you think would be better to talk about the governance of the DAO and the limits of decentralization, or should we start to describe the stack of DB DAO first and then we can go into the

[00:18:43] Michael: Let's do the governance of the dao. I think that's like pretty interesting. In terms of how you create these ecosystems of data. These data

[00:18:49] Ray: mm-hmm.

[00:18:50] Michael: so basically like we have this concept of a database and we wanna create a lot of really high quality data in this database. You don't want a lot of spam or messy data because that doesn't create value for. why are they called db? Do? So each database is managed by ao, and ADAO stands for a decentralized, autonomous organization, which is basically, there's a lot of different ways to describe ao, but I think one way is just like a group of people who are, aligned towards some common goal. so a group of people align towards a common goal, come together, and they create mission for a database.

[00:19:24] Michael: let's say this mission is around finding plant leaves that are around the world. So this plant leaped out, then creates a schema for the database. So they say, our schema for our database is going to include an image.

[00:19:36] Michael: It's going to include when the specimen was collected and the location. then they're gonna create that database and then they're gonna come up with a structure of which data is going to be accepted or rejected from this database. this is like a big problem that we see on web two today is like what types of content should and shouldn't be allowed on Twitter.

[00:19:54] Michael: And this is a very hot topic, and right now it's basically defined by the terms of service. And maybe on Twitter it is defined by like what Elon Musk wants or doesn't want, or with the rules that he creates in the terms of service. then deputies many different types of people whether they be AI or humans, to search the website and remove content that is not in line with the terms of service.

[00:20:12] Michael: But in the same way, this Dow creates a terms of service and then they create a multisig on top of this database. And the Multisig basically says it's a governance structure where it says, like, one possible example would be like in our leaf dow, there are three members. If two of the three vote to delete a piece of data from the database, then it'll be deleted. what's cool about this is it creates a very like structured and organized way for disputes to be resolved around what types of content shooters shouldn't live within the database. And of course you can go back and audit all the different types of content that are submitted cause it's all on chain.

[00:20:48] Michael: So if you think that this dow that is managing the database is corrupt because someone has submitted 10 different leafs and they all fit within the guidelines and this dow just keeps rejecting them. you can whistle, blow to the larger audience on another platform and saying, look, this Dow is corrupt and is no longer accepting things that should be included in the database.

[00:21:05] Michael: Now we have another interesting mechanism in place, which basically disincentivizes spam. like I said previously, if you get content into this database and then the database monetizes. your leaf is used a lot and ads are displayed against it. The database earns a bunch of money and then, well, you earn a portion of that revenue as someone that's contributed to the database.

[00:21:24] Michael: Now a lot of people wanna then be in the database for that reason, cause they'll get, earn money. Now a bad actor could come along and say, oh, I want to submit a bunch of pictures of dirt. And then I hope that I will then monetize those pictures of dirt. And this is gonna take the people who are in charge of the database dow a lot of time they're gonna have to go through all of this data and it's gonna take them a lot, waste a lot of their time. So what we do is the database administrators, set a deposit where each row that is being contributed to the data set has to go along.

[00:21:56] Michael: So the person also has to submit, Like 50 cents along with that row. And if their row is accepted into the database, they get their 50 cents back. And if their row is rejected, database slashes that. And that just goes in towards funding the database. So if someone is not going to then post a bunch of pictures of dirt, cause they're just gonna submit a hundred items and then lose $50 and they're not gonna gain anything.

[00:22:22] Michael: And the idea is that this of losing the money will just encourage people that are there for the reasons to contribute.

[00:22:29] Ray: Yeah. I will say though, I think it also creates some friction with actually using the application too. anytime you ask a scientist, To pay for anything almost there's some friction there. So would just be cautious or I would be aware at least of how that might impact the user experience.

[00:22:46] Ray: And I mean, I think you'd agree, right?

[00:22:48] Michael: Totally. Yeah, a hundred percent agree. And so if that is the case, you can always set the deposit to zero, in which case they can just use it freely. And then sort of once spam becomes a problem or you decide to open it up to a larger audience, you can start upping that deposit to a higher and higher thing to compensate the curators of the database.

[00:23:07] Ray: Cool. So that's customizable. obviously here. Can we talk more about, you told the audience about Multisig is how this database will be managed in the dao, and you used Nosis, from what I understand, to manage the Multisig. How has that experience been?

[00:23:20] Michael: Yeah. It's great. So the idea is you wanna create many different types of governance structure around who sits on the other side of the types of ways in which content should or shouldn't be accepted into the database. From our perspective as DB Dow, we sort of put that all into a black box where a row goes into this black box and then that row is either accepted or rejected.

[00:23:43] Michael: Now, how that black box makes the decision around what types of content should or be accepted or rejected is up to the person that crafts that black box. And in this case, that black box is a notices multisig. And what a multisig is, is instead of just a transaction being approved by one person, it has to be approved by possibly many people.

[00:24:03] Michael: Or you set up the rules in how it is approved. So in some ways, if it's just one person who's managing a data doubt they'll just be the only person. There'll be a one of one. So one person will sign it and then it'll go through. But there's more complicated, like look at a corporate governance structure where there's the CEO and the CEO deputies.

[00:24:20] Michael: Has many deputies beneath them, and any of those deputies can sort of make a ruling but the CEO can then overrule them if they think that ruling is wrong and remove them from the multisig or just overrule that individual decision. And there are many different types of protocols uh, like boardroom is one that allow for more complicated governance structures within this black box.

[00:24:41] Michael: And I think down the road we're gonna see a lot of innovation in the types of things of how people wanna set up the sort of the, human side of how these dowser, make decisions.

[00:24:50] Ray: Yeah, that's pretty cool. You mentioned boardroom, that's boardroom io. I haven't really used them, but they look pretty interesting for anyone listening. I'm just curious. Awesome. And then you have this concept of, and you've already mentioned this, but each row acts as an N F T.

[00:25:03] Ray: So in E R C 7 21 token can you just describe that a little bit more in detail, why you use it and how that. , you know, facilitates

[00:25:11] Michael: there's two big innovations here. One is that the database is managed by a dao. And then the other sort of innovation is that every row in the database is a N F T. And in this specific instance, there's like two types of NFTs. There's this one called the 7 21, which is sort of a mince them in an e in a sort of first one if you just need one ft.

[00:25:32] Michael: But then there's this other one that's called an E R C 1155 if you wanna create multiple different types of NFTs, but it has a lower gast. So we use this lower gas mechanism. But the reason you want this is now every row in the database is token. And so what happens is that when you submit a row to the database, the database has a fixed schema.

[00:25:52] Michael: So in the case of the leaf examples, like an image, a date, and a location, and then any sort of every time you mint a row, you're minting it to conform to that schema of the database. And then what's happening is, is that schema creates a j s O object that j s o object is then stored onto IPF f s.

[00:26:13] Michael: the reason you want to store this on IPFS is it lives forever. can be retrieved by anybody. And it also creates what's happening is you're putting an infinite amount of data, more or less, and then hashing that down to content id, which is a fixed size.

[00:26:27] Michael: a couple bites. . And so now you can store any size object in your database. so then we take that c i D and then wrap that in an NFT and then add that NFT to a specific database. the reason you want this is now you've tokenized sh all the content that's within a database so that the person who created that data can then receive rewards based on how much their data is used within that particular database.

[00:26:57] Michael: And then also, give that data to another person down the road if they no longer wanted an exchange for money or they can also add that piece of data to another database. So imagine that, we've developed this leaf and it goes into this DCI thing, but then someone else down the road, I also want to monetize this in another way, and I wanna add my leaf to Instagram clone, like a web three Instagram clone.

[00:27:21] Michael: So I have this generic NFT object that I can then add to many other types of databases that exist, that have this same schema, or maybe I have and then I added to maybe two or three different databases. And what's interesting is then this N f t there's only one of them will monetize from multiple different data sources.

[00:27:39] Michael: So you have a reference for this that describes this, one object. , which I think is valuable. Cause oftentimes, you know, you might copy databases between them, but then, you know, if later down the road that this leaf, this particular NFT is invalidated, I don't have to invalidate it across multiple data sets.

[00:27:55] Michael: I only have to invalidate it once. So there's a lot of benefit in just having that one NFT describe one specific object.

[00:28:03] Ray: Got you. Yeah, that makes sense. It's quite complicated, what you're doing. I think the building block of database is basically for web three is sort of what you're trying to, build and really interesting stuff. I understand you also include a stablecoin as part of the data curation and incentives mechanisms.

[00:28:19] Ray: Can you describe why you chose U S D C specifically and then how a stablecoin helps to make this more efficient? Effecti.

[00:28:26] Michael: Yeah, we can basically, a stable coin is an E R C 20 That denotes value tied to U S D C, but we can take any by default, kinda the idea is you people don't wanna use these other things, but you can also take in say this a DAO issues their own token and it's a limited supply.

[00:28:44] Michael: They can incentivize people to do work on their network, do work for them using their own E R C 20. But maybe that, same Dow hasn't issued a token yet and the people just wanna earn U S D C for the work that they contribute for this database. then they can go buy food or repay rent. we're open to many different types.

[00:29:07] Michael: , I mean, that's one of the cool things about Web three is it's all very interoperable. if some data set. Thinks that U S D C is the right way to pay people, you just stick in the smart, the address for that smart contract. But if another Dow say the mushroom, Dow doesn't wanna get involved with money, they just care more about points and reputation.

[00:29:22] Michael: They issue a token, they create mint a million of them, and then they gradually disperse that out and people can then show on their website how much mushroom token they have to gain accreditation within this community.

[00:29:37] Ray: Got it. Yeah. And that's cool that there is that flexibility and customizability within the system itself because you're probably gonna get a lot of interesting requests and questions you probably haven't thought of yet in terms of how people want to use this or build on top of it.

[00:29:51] Michael: that's one of the cool things about the web three is it's all just very plug and play. You can just kinda mush the pieces together.

[00:29:57] Ray: Interoperability is a good word for web three I think. Just because. It is in natively interoperable. the idea of blockchain protocols itself, they're very interoperable or they're getting to become more interoperable as we start building upon them. You also use lit protocol and prisma lit protocol for the columns in the database to be encrypted.

[00:30:18] Ray: And that's an optional thing as well as yeah. You wanna talk a little bit about that just for our technical listeners?

[00:30:23] Michael: totally. So the idea is a lot of this data is posted on chain, in theory you could access it, but you also wanna be able to encrypt certain parts of it. So lit protocol is a way of encrypting based on NFTs. So the idea of lit is that we encrypt a column of the database and When you read it, you won't be able to read that column.

[00:30:42] Michael: So say we have a database of first name, last name, and email addresses with each of those as in a column. Now we wanna encrypt the column side of things then to be able to decrypt it, you need to then purchase an nft, which will give you the private key to be able to decrypt that data in that database.

[00:31:00] Michael: So you can develop a lot of really interesting applications because of this. So imagine you created like a Spotify competitor and you could search the names of the songs, but then to be able to play the song, you have to pay for the NFT that will allow you to decrypt the mp4 that's part of that, for that specific song.

[00:31:21] Michael: Or Netflix, a similar thing. You you could search the index and search the database. . But in order to play the movie you need or for lead gen, you need to pay to get acc You could search for like approximately what you want, but you need to pay to get access to the contact or for medical, you know, you can see the results, but you can't see the underlying molecule that is used.

[00:31:40] Michael: And if you wanna purchase that, you can. So that's for lit. And then for Prisma, we wanna be able to interop operate with existing web two applications. Prisma is something that's used pretty generically online and allows us to uh, have a drop in URL replacement. So if there's an existing web two application that wants sort of like the superpower of web three, they just drop in our URL that points to our database, then we can start issuing and minting NFTs for that specific web application.

[00:32:08] Michael: point is, dbw is very interoperable and is, Uses a lot of existing protocols to give our users more benefits.

[00:32:16] Ray: To make it work.

[00:32:17] Ray: let's talk about the actual user interface and experience. And I know there's different types of stakeholders that would wanna be, be using DB Dow. It includes like just developers. I think it's interesting researchers who actually want to use it to, make their scientific data public or just to manage their data on a decentralized sort of cloud system like you're building.

[00:32:37] Ray: I don't know if I'd call it cloud already, but it's a decentralized database. Right. then you've also listed specific stakeholders, the curator, the scout, and the viewer. So maybe you can just walk us through what the experience is like for these guys.

[00:32:50] Michael: yeah. So as you said, there's basically the curator, the person who comes up with the schema for the database who promotes the database. So if you just have a database on your own, no one's gonna know about it. you need to tell people about it.

[00:33:01] Michael: The curator also does the role of accepting or rejecting data. So for all this sort of work that the person does when the database monetizes, they get 30% of the revenue that is generated by this database. So they're rewarded for the work that they've done. Now, for doing our part of the work, the network takes 10% of the fees as well, of the rewards.

[00:33:28] Michael: And then the additional 60% go towards the people who we call the scouts. And the scouts are the people that contribute data to the database. let's take an example where a database generates a hundred dollars in ad revenue. Now $30 goes to the curators of that database. We take 10, and then the scouts have, let's say there's 60 rows in the database.

[00:33:53] Michael: and $60 goes collectively to the scouts. And because they're 60 rows in the database, each scout earns a dollar a month for the contribution that they have. So if if it's over the course of a year and they have one row or two rows, they're making 12 or $24 a month for

[00:34:12] Michael: that.

[00:34:13] Ray: so let's maybe clarify what we mean by the rows. Alright, so let's use our leaf example. So there's a community of people cataloging leaves in the world from like trees and stuff. I guess you can have a column that says like color, shape tree type, and all these different like attributes are columns, basically that describe the leaf.

[00:34:33] Ray: let's say you have 50 different columns describing a leaf, and what you're saying is whenever a researcher or a, scout identifies a leaf and they put in the information in one of the rows. So leaf A has all these attributes, so they just put it into the database and. Now that's, it's created an N F T representing that leaf sample.

[00:34:54] Ray: walk me through after that, what happens?

[00:34:56] Michael: Yes. every time someone goes to this leaf, say, you found a maple leaf. Now I go to a website and I view your maple leaf. and alongside me viewing that maple leaf is an ad for maple syrup. And that ad deposits a penny into the database. And now let's say a hundred people view your maple leaf.

[00:35:18] Michael: So now a dollar will be stored in this smart contract from the ad revenue that's generated. And then at the end of the month, that ad revenue goes to you because your leaf has sort of generated that income. you fill out the attributes for that specific leaf. And then other people might view it and it might be support by ads or maybe another database might query it, and then it costs a small amount of money per query.

[00:35:44] Michael: the other thing I wanted to mention is sort of how the user interface looks for someone building a database. So when you're building a database, what you're actually doing is sort of, something very similar to a Google form. So we have a drag and drop functionality where you can drag in you wanna post an image, you wanna post a location, you wanna post a color, and you just drag in these different form types.

[00:36:04] Michael: And then you press mint and then that creates a form which you can then send out to your constituents. And then they can fill it out. And every time they fill out this form with this structured schema, they're sort of minting an N F T in the background. And all that data is Being collected in that database, and then you can display that data and generate revenue from that data in any way you want.

[00:36:26] Michael: But the point I wanted to make is, it's very easy to create a Schema database. You don't have to like code anything. It's just about dragging and dropping like user interface elements to create a very similar to Google form like experience.

[00:36:37] Ray: Understood. And so these curators, you're envisioning this work being done maybe manually upfront and then in the future, how do you see the curation being done?

[00:36:48] Michael: Yeah. I think down the road we're seeing a lot of stuff with like chat, G P T being more and more important in how systems are built. And I think there's an interesting way of, like right now all the big websites like Facebook and Twitter use a lot of AI to flag certain types of content.

[00:37:05] Michael: But with AI becoming more accessible to people, I think it could be that in addition to having a human curator, . You could also have G P T, which is like this general purpose language model that is in extremely powerful be one of the curators on this multisig. So now the G p T chat can just automatically flagging types of content that are good or bad and marking them, as, for deletion.

[00:37:30] Michael: And then maybe because they're on the multisig. In one hypothetical example, instead of two humans needing to delete a piece of content, the G P T will be the first person who flags it, and then it only needs one more. So this will save a ton of time. And then as it becomes better and better, maybe we can rely on it more and.

[00:37:47] Ray: And I'm sure you can think about how that could be a problem or dangerous in terms of like AI bias, so something that we'll need to monitor very carefully, especially in the beginning.

[00:37:56] Michael: exactly. But I think that this is a very powerful tool because like DB DAO is very much around decentralizing and democratizing content curation and moderation to some extent. And AI certainly has biases, but you don't worry about AI being biased in the same way as humans do.

[00:38:14] Michael: So people might worry that a curator of a database has been compromised by being paid or by being some sort of relationship, like a social relationship that someone has. But I don't think that the threat model for AI is different and no one worries about someone really paying. G P t.

[00:38:30] Michael: maybe that'll happen in the future, but I think each of these has a different set of biases and together, I think you have the ability to craft it in such a way that'll create a stronger content moderation system than either one alone.

[00:38:41] Ray: That's interesting. I appreciate that. Thanks. know, the DC space has gotten a lot of traction in the last couple years. , but there are still challenges. What do you think are some of the biggest challenges in.

[00:38:52] Michael: Yeah. I think we're, we're still figuring out, know, what are the types of problems that will be solved by this? And I think people are starting to get very interested in. So there's a lot of DAOs that are coming together and collecting, self-report data from their Fitbit, from their own experiences in life.

[00:39:08] Michael: I think we're overcoming a lot of the hurdles in terms of getting people comfortable submitting their data, rewarding people through economic or social means., I think it's just about it's still very early for Tsai. Like this concept is only, a year max, probably closer to six months old.

[00:39:25] Michael: we're still even figuring out like, what are even the problems in this space? But there is some good traction starting to occur with DAOs being formed and setting up experiments and conducting research. I mean, this is, the research time is also sort of like along, I mean, PhD is multiple years.

[00:39:42] Michael: Publishing a paper might be multiple years, so I wouldn't expect for there to be results instantaneously the moment the word Desi has been established. So it might take a little bit of time, and then of course the tooling around web three. I think there might be bottlenecks in leadership, people who are willing to lead these DCI projects people you know, have the actual underlying thought of the experiment to conduct and the insights needed to conduct impactful scientific research are, those things are always can be in short supply, but you know, when you open right now, the types of people who do that traditionally are, people who are at universities.

[00:40:15] Michael: But it wouldn't surprise me also if, many types of people had those skills or were willing to learn those skills, but just didn't have the opportunity to be at a research institution. So I think that's an unknown as like maybe there are a lot more people out there who could be scientists than who are actually scientists today.

[00:40:34] Ray: Yeah. And that sounds very similar to what another guest of mine shared with us, Mike Sin from K and he's been collecting his data as well, his personal data, things like your Fitbit and whatnot, and putting that into a, some system.

[00:40:48] Ray: Have you heard of Cureo?

[00:40:50] Michael: Yeah. Yeah, yeah, Yeah. definitely.

[00:40:51] Ray: hey, my listeners can show you that, that episode, episode 1 0 5 as well. So very much we're going down that route. I think a lot of people are becoming more aware of the, data that they're generating and. New ways. And like you said, web three tooling is becoming slowly better.

[00:41:06] Ray: And once it becomes as convenient as web two tooling, I think we'll be able to create much more usable experiences for people. And like any good Web three organization, company, or group of people, community engagement is super important. And I think, there's obvious examples of you trying to create more community engagement.

[00:41:24] Ray: And one example is Desi nyc. So I guess my question to you is, what kind of tips, tricks, thoughts do you have around community engagement that drive, more engagement?

[00:41:35] Michael: Yeah. I mean, working together with other, keeping it fresh and interesting. So we're always trying to work together with new DAOs to keep the content interesting. Working together other teams like yourself to keep people synaps firing in new. Keeping the community small, I think is actually is a good tip is really important in some ways.

[00:41:53] Michael: You don't want to grow the community such that the people who are there are there for, aren't invested in it or aren't or there for maybe the wrong reasons. And so a tip I would say is don't focus on like numbers, but focus on like quality of contributors as being a metric for success. I think once you have established that core or finding it's easier to grow than once you, if you have a very weak connection between different people. , that's one of the reasons we started NYC was uh, it's a way to build stronger community amongst a smaller number of people than it is around growing into a massive loosely connected community.

[00:42:29] Ray: Right. And they'll allow it to scale over time, sort of naturally in a way. But you also wanna prepare for that scaling too, in some ways.

[00:42:36] Ray: and what tools do you use? Like do you have, discord? I know your website is available as well, which I'll put in the show notes.

[00:42:42] Michael: Yeah. We have a discord, we have a DCI nyc disc that we use if you'd like we'll post a link to apply for that. yeah, we have to apply for it to keep the quality high so we don't accept everyone into it. We wanna only accept people who have demonstrated they're interested in DCI and put a, small amount of effort into thinking about DCI first.

[00:43:00] Michael: And I'll have a link to that on the tab. And our website tools, telegram, in-person events those are the best ones so far. I'm sure we'll be thinking of more ones down the.

[00:43:09] Ray: Awesome. So I have a few more, personal questions I wanna share, but is there anything else specific to DB Dow on either the technical or just like the general level that we maybe didn't cover that you wanted to share with the audience?

[00:43:22] Michael: I think we went into a lot of the detail. I mean, we have an interesting caching system that we built on top to make it all very performant as well. And final resolution is all done on I P F S and we use smart contracts to assign ownership to different people around the different parts of the data.

[00:43:39] Michael: And then it has a very simple to use Google form style interface to build the databases. Beyond that, I think we covered a lot of it

[00:43:47] Ray: maybe you can share just, Any specific applications now or scientists now that are using it, and maybe just some examples of what kind of databases are being used are being created.

[00:43:56] Michael: Totally. So we're working with Hair Dow to figure out different types of treatments for hair. Uh, We're working with psychedelic Dow to, have people do trip reports. We're working with a weed dow to have people crowdsource write a scientific experiment validate different test results that have been published already.

[00:44:17] Michael: So I think reproducibility is a big problem in science today. maybe you don't conduct original research for some things, but if you can just validate or invalidate certain research that is helping a lot To conduct experience. We're working together with build a non D side thing to help them build their applications.

[00:44:35] Michael: There's many different types of applications beyond the D side that we're building together with. And many more in the works as well.

[00:44:42] Ray: Awesome. That's really good to know. in terms of an outlook for next year, 2023 and beyond, what are you guys planning to do? you know, what are some of the milestones that you guys are trying to accomplish?

[00:44:51] Michael: Yeah, I think we wanna strengthen our ability to create simplify the user interface more make sure we're, solving the right problems add in more complicated schema for people. We want to start making it more interoperable with other things so that you can know, plug in different things.

[00:45:10] Michael: We wanna be able to, like modularly plug. With a data visualization or compute, and we can do that now, but I think there's ways that we can make this interface a little bit more easy to do. I think those are our main things is, making it more modular for like a visualization and compute tooling to come in or making it so that it can natively drop into a Jupyter Notebook and you can just store your data automatically onto DBW

[00:45:35] Ray: on that note, you mentioned you wanna have compute as part of the uh, one of the functions my question is, and you also have an encryption mechanism as well, is there like a confidential type of compute mechanism you're trying to build so that you can potentially, without seeing the data at all, apply some algorithm or functions in order to determine like a result of based on a certain data set?

[00:45:56] Ray: Is that something that you're thinking about or,

[00:45:59] Michael: We've thought about this. there's a couple different ways of doing this. The holy grail is homo graphic encryption, homo computing, homo encryption computing. The other one is, trusted com computing environments. I don't think that's our bread and butter right now.

[00:46:12] Michael: So we will work together with other people to do that. But this is the holy grail you have a database co of. Students and you wanna see how many of them have had a Covid vaccine. but you don't wanna reveal who has had that COVID vaccine, but you wanna know the percentage of students maybe that in the school.

[00:46:29] Michael: So you could issue insurance so that an insurer knows, like the premiums that someone has to pay. so they could query your database and say like, oh, 90% of the students have been vaccinated. You're gonna have very low premiums without revealing sort of any of the underlying data.

[00:46:44] Michael: So that will happen. we will get there eventually. Right now there's a couple technical things that are preventing it, but it is sort of on the longer term roadmap to do these sort of more complex and trusted things that you'd want to do with a web three database.

[00:46:56] Michael: we built a prototype actually that does this using it wasn't fully SQL Queryable, but it used some zk technology Z zero knowledge technology to create a ZK database for simple data science queries. And it was very interesting. we'll probably pursue something like that down the road.

[00:47:14] Michael: And in the shorter term, we'll work together with the more compute layers to use this database to then do compute. You basically wanna take a Docker file in the highest level since to then do compute over the database and then run these docker instances on the various d vdo database.

[00:47:31] Ray: Are there any other organizations that are doing something very similar to what DB Dow is doing

[00:47:36] Michael: we're the ones that are the, the Firebase for web three in terms of creating a schema, making it very interoperable, offering many different other features on top of it. And plugging in,

[00:47:47] Michael: I mean, there's obviously like Postgres and Firebase you know, Mongo or Redis, but I don't think anyone's taking the full on web three approach to databases,

[00:47:57] Ray: Interesting. And if anyone listening disagrees or thinks to have one, let us know in the, YouTube or telegram chat. That's the best way to do that. Do you have any favorite d l T projects you think is doing really important work? And this is like a general question. It could be like, you could say Ethereum if you want.

[00:48:12] Michael: I don't know if we include this, but Z Ks Z Ks are going to be the next blockchain. In many ways, ZK reduces the need for reliance so I mean, the issue with blockchains for the most part is that the reason you use them is so that you can get verified compute or verified, storage.

[00:48:29] Michael: But there's ways of doing that without needing to only rely on one central party to do that. So one of the things that's not so great about Ethereum is it's sort of a giant bottleneck that everyone wants to do, compute on this one slow computer. But uh, CK changes that o other things that.

[00:48:47] Michael: D L T cure is interesting. Fleming protocol is interesting. Anyone who's doing sort of science or creating an application I think are the ones that I sort of, I mean, step in is interesting too because I think they're, you know, one of the first ones to do an application that there's some good and bad things about it, but they're the first ones that like create like a straw alternative that is based on the blockchain and uses incentives based on the blockchain as well.

[00:49:15] Ray: Welcome to the Health On Chain News Corner. Google Research and DeepMind have recently launched Med Pom, an open sourced large language model aligned to the medical domain. It is meant to generate safe and helpful answers in the medical field. It combines health search qa, a new free response data set of medical questions.

[00:49:37] Ray: Med P addresses questions posed by medical professionals and non-professionals by using various data. These data sets come from Med qa, med M C, qa, pub Med, QA Live, QA Medication, QA and M M L U. A new data set of curated, frequently searched medical inquiries called Health Search qa. Was also added to improve the multi-med qa.

[00:50:01] Ray: This model was developed on pom, which stands for Pathways Learning Model, A 540 billion parameter large language model for ai. This launch comes after. Chat. J PT three has seen enormous utilization by all kinds of people and industries. These AI enabled chat applications are getting way better very quickly now that these technologies have reached so many people and the data sets keep growing exponentially.

[00:50:27] Ray: I wonder how, if at all, blockchain technology would be useful to better secure the data that's input and generated from the large language AI models. I'm excited to see the largest search engine, Google continuing to push their healthcare agenda forward. Will this tool be successfully used in practice with real users and patients?

[00:50:49] Ray: That is a very important question, and I will certainly be looking for that answer as their tech improves and has rolled out to more people. I hope you enjoyed this news corner. Be sure to subscribe to the show and give us a review on your Apple, Spotify, or however you're listening. And now back to the show with Michael Fisher, founder of DB Dow.

[00:51:08] Ray: do you have a book that you've read would you say is most influential to you?

[00:51:13] Michael: Yes, I would say radical markets. And it's basically how to use economic incentives to create great governance outcomes. So one of the problems we see is, if you set up an incentive for people to do certain things, a lot of people will come in and do that. And one of the things that, you know, we're seeing with government is sometimes the incentives are very unclear.

[00:51:32] Michael: So this book sort of outlines how to set up incentives to create uh, outcomes that are sort of good for the general population. And when I read this book, it really changed my thinking on terms of thinking about how to do protocol design and really the benefits of being able to design a good protocol.

[00:51:48] Michael: And the only reason a protocol will be good is if a lot of people use it. So it's sort of like a self-selection thing where like the good protocol designs will attract a lot of people and then will have a good outcome. And, and there's ways of doing this sort of mechanism design to do it. And that was one of the first books I read sort of on the subject it really did change my thinking.

[00:52:06] Michael: And the other one would be like the sovereign individual, which also changed my thinking around the interplay between. Technology and encryption like society how these things all fit together.

[00:52:17] Ray: Awesome. How do you like to stay active and exercise or stay fit or just, you know, physically, there?

[00:52:24] Michael: in New York I'm part of a couple running clubs, so running Central Park sprinting and then longer runs, so trying to get my VO two max up

[00:52:34] Michael: and my heart rate pounding and then also longer runs I'm going through running a half marathon in a few months, so I've been starting to train for that.

[00:52:41] Michael: And beyond that maybe some weights also. And then trying not to eat too much. I think, slightly just eating plants and generally trying to eat healthy like that is a good thing. And then sleep. Don't get too much sleep, but also don't skimp on it beyond what is reasonable. Either

[00:52:59] Michael: you need to keep your brain I work too long and then my brain at some point just gets a little, like, I can notice a drop off and

[00:53:08] Michael: just need to sleep or do something else. Go for.

[00:53:11] Ray: Yeah, I kind of hit a wall there after a certain point. Well, I do have one more question, actually, it's my last question and you can answer one of these two questions or none if you wanna, I suppose. But basically have you had an injury or some sort of medical condition personally or maybe in your family or friends group that has affected the way that you see the healthcare industry and then thought of some solutions for it?

[00:53:33] Ray: That's the first question. The second question you can choose to answer is, have you had a psychedelic orogenic experience with some plant substance or fungi that has changed your perspective or outlook on.

[00:53:47] Michael: I haven't done the second one. I can tell you like the first one is I don't even think we need to get into the complicated stuff. I had to recently, get a doctor's appointment. And it took a long time and very complicated to be able to do it. Just even forgetting like literally the most routine physical checkup the wait time was like a month. and very complicated, bad. I don't even think the existing healthcare doesn't even talk about like, dealing with more complicated issues. We just have even the, the most basic things to fix first. Before we can even get into optimizing beyond routine stuff. I mean, I just needed to get something checked on my foot and I ended up just solving it myself before I could even see a doctor.

[00:54:30] Michael: Cause that didn't work. I called 10 doctors. That listing was very out of date. They didn't have appointments for a long time, and it's not cheap either. Like I got a reasonably good one. And so it's not like I'm trying to unduly save money here and it still, doesn't work.

[00:54:44] Michael: And I think is, that is a huge problem. I mean, that's one of the one things I'm very excited about decentralized science, where in many ways we're fixing a real need for a lot of people. Because they might not be able to get care because even if they do have the opportunity, it might not exist.

[00:54:58] Michael: And if they don't have the opportunity, it double doesn't exist to find healthcare. And then also, you know, it makes me think maybe if these basic things aren't being done well probably the more complicated things aren't being done well either. if you can't even get an appointment, the likelihood that things down the road are like the medical, like that type of stuff is working very well, is, not very likely.

[00:55:18] Michael: So there's probably a lot that's left on the table in terms of new types of treatments or new types of opportunities that could exist if there was like a working system. And I think the existing system is very inefficient. you know, I'm not saying that this will for sure, but I think the bar for the gold standard is lower than we think.

[00:55:40] Michael: and the cost and opportunity is easier than we think and easier to achieve than we think. So I'm very optimistic about it.

[00:55:51] Ray: Yeah, and earlier you mentioned G P T or chat, G P T, and I think that can eventually also be used to help make it a little e easier to get some information, medical information. And I know that's like risky because you don't wanna have a patient, you know, make a decision based on some AI recommendation.

[00:56:07] Ray: But over time some of the things can be supplemented with ai. Some of the conversations I'll just say one other thing is I think there's this very amazing opportunity with peer-to-peer medicine. So right now we think of a doctor giving you medical advice or also like G P t giving you medical advice, but there's this also thing where, if another peer of yours who has a similar type of disease has probably done a lot of research on it, perhaps even not always, but in some cases more so than a doctor and has no more empathy towards your, and more time to be able to devote to you.

[00:56:38] Michael: So if we can build these patient communities using data as a initial nu point for people to gather, there's a giant opportunity where maybe you could have other patients supporting other patients there will be roles for each. But this is an untapped opportunity perhaps.

[00:56:52] Ray: It's a great point and I do think that that's true. And you already have some patient communities that have formed over the last. 10 years just used because of the internet. And I think having a providence of data and a more trusted community would even, make that sort of idea explode even more so in a good way.

[00:57:09] Michael: Patients like me is, I mean, I've seen that, but like one of the problems on patients like me is a lot of the data is really unstructured. And so it becomes this forum, which is great, but also you could imagine something well put together and is able to be used by other people.

[00:57:23] Ray: I think that's something they're probably aware of and trying to figure out as well. Michael, I just wanna thank you again so much for spending your time with me and talking to me about DB DAO and decentralized databases, Desi, and really the future of healthcare and science overall.

[00:57:40] Ray: Yeah. Just thanks again. Do you have anything else you wanna share with the audience? Any last final takeaways?

[00:57:44] Michael: um, no, I think, uh, thank you also Ray and anyone who's interested come by DCI nyc. Basically at the beginning of every month. and learn more about it and learn about interesting projects and I love your show thanks for having me.