Decentralized Data Science and Data DAO for Public Good - Linda Lu

um thanks trace awesome hello everyone morning happy saturday thank you so much for taking the time thank you so much for being here i know it's early on saturday morning and thank you so much for waking up early i know it's not easy um so i'll do my best to make it worth your while and so let's get started so my name is linda liu i'm a head of ecosystem at oasis and today i want to share our thoughts and vision on data dials and how and data dials were probably good so let's get started so we understand how important data is data is a key driver for the model economy and the critical piece for ai and machine learning data the global data economy is growing exponentially every day each and every day we just have more and more data generated and the same times we're facing unprecedented challenges when it comes to using sensitive data like we're seeing these kind of news making headlines like corporates organizations buying selling users data without their awareness let alone consent and some research has shown that individual personal data is extremely valuable we have seen evidence by this metric ppu price per unit so these as you can see the figure here it covers a wide spectrum from 75 dollars to 20 000 and it's safe to assume that end users it could be your data is not necessarily benefiting from any of these uh daily trans data transactions so the status quo where corporates and organizations sharing data without users awareness and consent is not acceptable it hinders the site of progress undermines human values and fundamental rights and we need to change we need to change that and how do we fix that we're introducing data dao so we believe that dao is a disruptive framework that addresses many challenges we talked about earlier so what is it about them a data dao is a decentralized autonomous organization so dao of individuals and entities that supports individual data rights rewards data owners for data use handles data with confidentiality and privacy with verifiable transparency in its operations i understand it's quite nuanced concepts it has quite a few key components underneath such as what is individual data rights what do we mean by rewarding and compensating for data use and how do we handle data with computationality and privacy and what do we mean by verifiable transparency so bear with me uh just be a little bit more patient we i promise i'll unpack these components one by one very shortly so what do we do with data dial what's our goal we create this data down and to do what exactly uh first we want data data to enable individual data sovereignty this will be meant by individual data rights individual you own your own data and you get to dictate how you want your data to be used uh second we want data to maximize social welfare and economic efficiency so data utility needs to go beyond individual and organizations it needs to serve broader society even the whole world we need to break data silo to be able to do that third we need to be able to provide incentives this will be meant by rewarding on compensating there needs to be a fair distribution values created from data to all members of the dow so that covers the owners the consumers and any parties involved in that transaction and individuals get rewarded and compensated for data use last but not least we need to establish enforceable governance we need to enforce individual data rights so we can prevent misuse or abuse of data and maintain audits with immutability to ensure that operation is accountable traceable transparent and self-pleasing of course we're not naive we understand the reason that adults haven't been so prevalent is because we still have a lot of challenges and complexity when it comes to data practices such as it's very difficult to balance privacy with utility it's almost as either or in the current system versus both end you have to choose between privacy and intuitive it's very hard to have both second data is non-rival meaning when i share my data out i can almost assume that someone is taking is making copies of my data or taking snapshot of my data you can never get your data back once once it's already shared out and third data dependencies so process data can review information about original data and if you know let's say you have a cohort of individual data and if you happen to know sense data of one individual through some robust analysis is it possible for you to derive sensitive information about the rest of cohort and last but not least within the dependency is this data anonymization which is a very common practice in the current data privacy privacy space but we understand the shortcomings of data data anonymization even anonymous data can be re-identified via correlations so how do the dows come to rescue were the enabled technology required for us to solve all these challenges we just talked about um it required two enabling technologies one is data nft second responsible data technologies um so first let's talk about data nft i'm sure everyone here understands the concept of nft so we'll now repeat that you know i don't want to bore you guys with what's nft but data nft is just like nfd but it's slightly more complex it's a token that provides access to wrong computations on data and making sure the computation is compliant with data policies set by the owners so it covers two parts first is access to computations second is making sure computation is compliant with data policies and second is this responsible data technologies these are tools and platforms that enable running computations on data maintaining data privacy and make sure data is confident confidential and private while in use and with all accesses out to the ball in an immutable log i know it's quite a mouthful but we'll unpack it in a bit very shortly so let's talk about data nfts first what do we mean by how do we create a data nfts so it has three basic components first is tokenized tokenization data dials will issue data back tokens that represent it could be docs it could be files images movies royalties and according to terms set by the data owners so if you own a doc and you want to wrap it as a data nft you will set you your cell will set how you want your data to be used second analyze we need to attach encrypt data as well as metadata to the tokens to keep data private and confidential while in use third is share within the ecosystem these data nft tokens need to be compatible with the ecosystem standard so it can be shared freely so people can send and receive as stakeholders and users wish within the data data here's one example of data nft in the genomics space so i trust quite a few people here have heard of this famous professor professor george church from harvard professor georgeter is a founding father of genome sequencing and he actually minted his 40-second genome as nit last year this kind of kind of pretty exciting so this is a kind of big milestone in the industry this is the very first data nft for genomic data and the sequentia is stored and protected by the oasis network as well as technology so it's kind of very very exciting so let's go back to the second enable technology responsible data technologies so it has four major components first is secure compute so that covers secure secure hardware compute such as intel sgx and secure multi-multi-party compute for homomorphic encryption these technologies allows you to keep data confidential in private while in use by any applications or services second is differential privacy so these solve the challenges we talked about earlier around data dependencies like if you know one individual sensitive data like through some analysis you'll be able to derive sensor information about the rest cohort so differential privacy will be able to solve that problem it ensures computational output won't leak sensitive information about individuals third is federal learning so this is getting common in the ai and machine learning space it ensures data never transported or copied with ai machine learning model so data gets to be trained in distributed manner data gets to stay where it is um last but not least is what we meant by verifiable transparency this is the three ledger um i'm sure everyone is very familiar with this concept this is where blockchain comes in it provides an immutable log to ensure data usage is compliant and traceable trackable and transparent so put it all together we want to help people visualize what the data down look like from architecture perspective so data dial does require a few critical services such as smart contracts with strong blockchain network such as oasis network uh it also requires disputed secure compute disputed confidential uh storage so let's unpack this a little bit with a user flow first let's say yourself is a user of data dial you want to upload data to the dial and the data needs to store encrypted so you will need this distributed key manager which will issue keys to you at the same time you will need to stipulate how you want your data to be used um so this is where we have the policy checker it checks any sort of competition requests either from data down or from data's partners any service partners against the data policy set by the owner such as yourself and once the the policyhacker gives a thumb up or green light it means these computation requests can be entertained then the dao will initiate these credentialed prescriptive competition services such as federal learning or different differential privacy queries on the data and then the smart contract run on a blockchain network will be able to mint the nfts and allow tokens to be freely trans transferred on the network at the same time we implement incentives and governments on chain to make sure it's enforceable so we talk a lot about this data dial so on high level we make sure we want to make sure it has it covers these three basic abilities first it needs the boat needs to be able to tokenize we created nfts from data shared by data owners could be individuals could be entities and make them available in the data marketplace so the data transaction will talk about in earlier slides those would be bad bad data nfts did the consumers will need to buy nft to be able to acquire the underlying data second is compute these will require responsible data technologies such as secure hardware compute differential privacy federal learning to make sure computation is allowed allow enabled data to be private or confidential while in use and last but not least governance we need to provide governance and incentive model to make sure there's a fair value distribution among stakeholders within the data dial so it's not just about value transferred from corporates to go data consumers to corporates did the owners need to be involved as well there needs to be fair value distribution among all stakeholders so let's take a look how the dial looks like in action say you're you're an individual and you register service with this data dao and you upload your data and you set your user policy and on the other side of the dao you have this let's say medical research organization who could use or benefit a lot from individual data more like a cohort or individual data so they will do some search in the data catalog which is maintained by the data dao they will run some search and find relevant data for their research and then they will put together cohorts of individuals they want around do more analysis of and that would require the research organization to buy data nfts of underlying individuals and then we still have the policy checker once the party target gives the green light the data uh the research organization will be able to schedule compute and analysis and then data dao will initiate these confidential compute on the underlying data and then pay individuals for data use so these flow end to end we enable enable individual data sovereignty and we break data silos so data gets to be shared across multiple partners and we enforce security and privacy protection end-to-end throughout the process so we talk a lot about data dial in a very abstract way i want to share a real-world use case by featuring one of our enterprise partners genetica genetica is one of the biggest genomic company in the apac region asia pacific region and they grew their users from 2 500 a couple years ago to 100 000 recently and they also have extensive network of partners such as research corporate hospital partners who could really benefit uh using some individual data to do more analysis out of so how does genetica manage to share those data with with those research partners while honoring individual data rights they build they are building their dials on the os network so how does it work let's say your customer genetica registers their service and you submit your saliva sample and that sample is sent to a lab to run full genome sequencing and the dao will be able to transfer the full ownership of your full genome sequence data back to you via nft and let's say a medical research partner will want to study a cohort on individual genome sequencing data then they will need to buy those data nfts and run computation out of it the whole process end to end we we managed to make sure that data stays private and confidential throughout the process so here's how genetically grows over the past couple years and uh till end of last year they migrated 20 000 genetic profiles to the oasis network and they are very much on track to bring 100 000 genetic profile to this crypto genetic economy via data dowse so yeah so we talk about data dowse and we do believe that out is for public good especially in the healthcare sector as well as other sectors it breaks the silos improved patient data probability enable geo and demographic diversity in clinical trial which is required by regulations and beyond just about sensitive data data is always can also involve nonsense of data so more people contribute data to data dell will allow better ai and machine learning model training and more user contributed data with their data they can create better knowledge graph potentially and we understand how important knowledge graph it is to bring ai machinery to the next level so to quickly wrap up we understand over the next decade it's a golden time to build data dowse you will have significant impact on society as well as the quality of life data dials enable individual data rights enforce privacy protection break data silos and provide fair compensation to individuals as well entities for the betterment of societies and here i always is we recently rolled out 200 million dollars ecosystem fund and one of the categories with city target is data dows so entrepreneurs builders founders teams out there if you share this vision if you want to be a better data economy please apply for this ecosystem fund we also have grants program to support that and i want to end my presentation by sharing one of my favorite codes from william gibson the future is already here it's just not even distributed yet we believe the enable technology for data dow are here those technologies such as nfts as well as responsive technologies are pretty much available for us to use the reason it hasn't been even distributed yet is because we need more people more builders more developers entrepreneurs teams out there if you share this vision if you want to build a better more responsible daily economy please join us please join oasis please uh build the dallas build better economy daily economy together with us thank you so much have a nice weekend guys [Applause]