The Future of Decentralized AI and Data

Transcript

K good morning everyone like to thank ecoin for the opportunity to talk to you I'm going to talk about something really dear to my heart I'm a developer I have been working on various blockchain projects focused on data ownership for the past few years I truly believe that blockchain not only is crucial for data ownership it is an essential infrastructure in the age of AI my my original topic is future of decentralized AI and topic I think I'm going to focus on one thing today um we

probably just have enough time to focus on one thing which is data ownership I want to cover the data ownership talk in four parts first I understanding the issues we are dealing with then describe the types of data and other blockchain solutions for True data ownership lastly I want to take a glimpse into the future with you so let's look at data issues now imagine for a moment that

every piece of data about you your photos messages browsing habits and even your health data is stored in a vault but here's the catch this Vault isn't owned by you it's owned by giants like meta Amazon and Google they decide who get asss they make billions from your data and privacy is an afterthought to change this let's first

look at the types of data that we have first you have personal data the content that you produce this includes private data the personal content you store in the cloud typically or on your local Drive includes photos bank statements notes and so on and then there is a public data data you share it or make available publicly social media public profiles radic and then there's share data data share with specific person or

groups the data may may be owned by you or maybe owned by by your company and also this data includes the first party data such as the surveys that you take that the content come from you the second type of data is a behavioral data your browser history your Google Map search history app usage health tracker data online interactions it could be dating apps it could be any apps and transactional data like your purchase records Financial

transactions service usage logs also there is data of data the metadata that generated with your data like you know the location the device type from your photo this is often referred to as your digital footprint and everyone generate a vast amount of data and the third type of data is the verifiable credential data it is more data about you so what are verifiable credential data they are issued to you by Third parties and verified by Third parties so

the F third party could be the government which issues your ID your passport your your Social Security if you are in the US and your graduation year and your GPA those are all data generated even it's generated by you it's approved or verified by a third party your world history if you you are into Olympics maybe we have Olympic medals and maybe just marathons that you wanton and the fourth type of data is derived data what I meant by this is data generated by processing the previous

data type so the new data created by transforming combining or even benchmarking through machine learning or other ways of processing assisting raw data it does provide new insights to enhance understanding and create a personalized guidance or experience for even for decision- making from interpretation of your health data like you know how healthy you are have you done enough exercise in the last week two your recommendations it could be Netflix recommendation it could be a shopping recommendation any kind of

recommendation engine and even for AI training especially if you are a designer you probably already your work probably is already included in all all those AI training without you even knowing it so how do we take control of all these data and then what we want to look at is not only the ownership portion of it we want to look at also the financial incentives the derived data for example is somebody else aggregating a lot of different sets of data from everyone in the world to give you insights about your own personal health

or your your preferences in shopping or in movie watching it has a a heavy Financial incentive involved so let's look at the solutions the personal data the biggest problem today with personal data is that it could be VIs ible to third parties accidentally or by hacking activities the reason maybe you already aware the cloud service that you use today from Dropbox to Google Google Drive and even Amazon

S3 they are not end to endend encrypted your which means your data is visible on your local device and then even though it's being transported securely to the server from the hosting party your data is only encrypted on their server so there is a risk where your data is being visible throughout in the third party ecosystem and if they screw up which how many of us um heard about um the Equifax right even the unencrypted power password from Facebook right it

happens beyond our belief so the risk of data um is existing for All Storage Cloud storages today and the other part is that if you want to generate a endtoend encryption meaning that you want to encrypt your data locally and only transport the encrypted data so nobody else can have a peak into your data it requires setting up a public key and private key I believe many of you probably already very familiar with cryptography because you are in the blockchain industry so cryptography in order for you to encrypt

this data you need one piece of information which is public key and generating there there are tools like proton drives um there are like open-source projects who are doing endtoend encryption for your cloud storage but it requires you to save your private key or your your uh word like training word TR War strings the problem is that most people would just save that in their Google Docs in their notion as plain tax that is the risk that you are taking for your own data privacy and then the second thing is that what happens if you

didn't save the key you lost the key well it's not as bad as losing like 100 Bitcoins but then you lost access to your own data the own content that you generated so there is a solution which just an example it's not the B all uh solution but if you look at from a technical standpoint any user who sign up today using either email or phone as long as they have an email address or phone you can generate a public key private key pair using

different Services I think um more people are familiar with previe um there are account obstructions NPC wallets I use MIP as an example MIP is a hardware non-custodial wallet that actually generates it's a private key priv uh private public key pairer for you and a crypto wallet address so once you have the public key of that user from the email address now you can encrypt the content you can complete end to endend end to end encryption and then you can save the data to even your cloud storage or your decentralized storage so

there are solutions to a lot of problem that we talk about today from usability to adaptability to security um throughout the last few years there's so many projects came up with amazing development and there all solutions available as of today so let's also talk about privacy preservation so basically uh I think yesterday I was here I I heard like three or four projects at least talking about zero knowledge proof you can generate any answer of zero using zero knowledge proof basically just to attest or basically to say whether this is true

or false are you do you own a drive valid driver license are you owe enough to drink I think the drinking age in Japan is 20 and once you set up your Zer knowledge proof you can subject any kind of data to a data marketplace where AI can use it for training but they have to pay you there are a lot of different architecture combination of use cases that's enabled by using this infrastructure so the other piece is the one that most people talk about I think um a lot of you probably already use Brave um trying to solve this problem they trying to democratize your your

browsing history data the advertisement that that paid to to The Operators now is paid to you but why isn't it popular yet so the behavior data from your browser history trackers shopping history To Your Health Data to your other Behavior data should only be stored on your local storage and this local storage can use the infrastructure mentioned before to sync to your decentralized storage with endtoend encryption and then you can generate selective disclosure using Zer knowledge proof and then sub submit it to a data

marketplace now just imagine that you I'm a fibit user right a lot of you may be using Apple health and some of you are using Samsung watch so instead of submitting your health data and be limited to one one data processor instead of just looking at Fitbit I can subject my data to whoop or any Health Data aggregator to tell me the derived data information and generate all the insights you will have freedom of choosing which app have access to your data so the verifiable credential the

the third part of the data is a little bit trickier I'm not sure I have time to go over everything um but here's the flow for issuers isur is very challenging because it may involve a third party entity for example if you graduate from Tokyo University Harvard University so har un University need to issue a verifiable credential to prove that you are a graduate but using the previously um set up workflow you can actually generate the recipient email uh a public key and then use a standard schema of verifiable credential um to

to publish a piece of data and then it can be subject to verify notes to verify that it is indeed Harvard University and then publish all your data to a decentralized storage it will be there forever and and the cering of the data will be much simpler very similar to U what we talked about earlier so what about the future first of all um I think we just heard about ethereum as a um threat to the nation or is it a financial experiment at scale to developers like

myself blockchain like ethereum is an infrastructure think about efficiency a lot of people think about blockchain they think that it's wasting a lot of electricity but the fact that it takes a 100 plus engineer to implement to solve for a double span problem which is inherently built in into ethereum so there a lot of possibility be even beyond the financial infrastructure that you can view such as data ownership and then with blockchain it's a programmable money it's so easy to redistribute the incentives by Shifting the structure the

decentralization is Unstoppable think about the media how many of you still watch TV to get news it's all from the social media self media and this is the trend not only in the finance but it's going to come into the realm of data only ownership as well and lastly I want to talk really quickly um before time run out is um about Ai and uh the implication for AI and Ubi so imagine what we were told that AI is going to free us all we going to receive Universal basic income we don't have to work anymore so what's the meaning of

life think about AI still needs Improvement and the way that it it gets Improvement is through more data more feedback loop to the data and this is number one our contribution by producing data if you are an artist you you're a musician or even you are a coder continue to code and produce content and that will improve ai ai models the other thing is that please do not forget AI I is about enhancing Our Lives we are not a digital being we care about all the biological Pleasures I

enjoy a good sushi dinner I enjoy um all sorts of traveling Leisure and that is the base of some of the recommend recommendation engine that I mention so your sole responsibility in the future is to produce data and the data you contribute will be the meaning of life and that is the true meaning of data ownership that's that's all I have for today and then I would like to open up

the floor for questions any questions goit okay uh how do we do okay sorry I I didn't catch that can you say that bit bit louder what blockers do you see in this being adopted more wide large scale I

really think it's a shift of mindset just like um I I had a couple slides about the challenges one of them I think it was talk about by previous speaker is permissionless versus compliance right uh if you want adoption you need to be more connected to the real world the change is not going to be from zero to one there is going to going to be a process the ability to adapt into that process is eventually going to be the accelerator of how people going to adopt into this new infrastructure new paradigm any other questions

all right let's give a big hand to Calvin thank you Calvin uh next up we'll be having a public goods panel discussion um we need to bring in the chairs that will take a little bit of time