{"id":"WILyZbf1rQs","title":"[REDACTED] 2024 | Own Your Data, Own Your Future","description":"Explore how privacy-preserving compute ensures secure AI processing in decentralized environments, protecting sensitive data and empowering user-owned AI ecosystems. Learn best practices for data preparation to maintain accuracy, scalability, and privacy across decentralized data sources.\n\nLukas Bruell - Nillion\nYu Hu - Kaito AI\nJoshua Meng - RSS3\nChi Zhang - Kite Protocol by ZettaBlock\nDr. Nagendra B  Bharatula - G-20 Group (moderator)\n\nTakeaways\nPrivacy in AI is Crucial for User Trust: Enterprises and individuals alike need privacy-preserving technologies to secure sensitive data like healthcare records, financial transactions, and personal activities shared with AI systems.\n\nEnterprise Adoption is Leading the Way: Financial services and healthcare are the frontrunners in adopting privacy-focused AI, driven by clear ROI and the critical need to protect confidential information.\n\nTechnologies Supporting Privacy: Techniques like federated learning, multi-party computation (MPC), homomorphic encryption, and trusted execution environments (TEE) are evolving to support secure data computation and storage.\n\nData Ownership and Monetization: Increasing awareness around data ownership is fostering initiatives where users and enterprises can monetize their data while maintaining privacy, opening new avenues for economic participation.\n\nFuture of Privacy Computing: The integration of scalable and efficient privacy-preserving technologies will be key to unlocking the next generation of AI capabilities, allowing sensitive data to be leveraged safely for better models and applications.\n\nNEAR is the Blockchain for AI\nWebsite: https://near.org/\nNEAR AI: https://near.ai/\nIntents: https://near.org/intents\nChain Abstraction: https://www.proximity.dev/\nSharding: https://nearone.org/\n\nFor the devs\nGitHub: https://github.com/near \nDev Docs: https://docs.near.org/\nDevhub: https://devhub.near.page/ @NEARDevHub\n\nConnect with socials\nTwitter:   / nearprotocol  \nBlog: https://near.org/blog/\nMedium:   / nearprotocol  \n#NEAR #nearprotocol #NEARAI","channel_name":"NEAR Protocol","published_at":"2024-11-15T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/WILyZbf1rQs/hqdefault.jpg","transcript_text":"hi welcome everybody Welcome to the to the uh to the panel today um why don't uh please starting from here introduce yourself first very briefly about me my name is nag found SE of G20 group investment and trading firm based out of Zurich Switzerland background in PhD in machine learning and AI really privileged to be here please over to you hey guys uh I'm Chi I'm the co-founder and the CEO of uh kite Ai and Z block Labs so uh a little bit about myself uh I did my PhD from UC Berkeley focus on machine learning and CAO inference and after that uh started uh uh Automated machine learning platform uh startup with a few friends and then realized data is actually the biggest Boton in industry for real AI adoption then uh join data BRS lead the data engineering Solutions there as a um product management lead and now uh Zena blog or uh kite AI is building a open composable U foundation and that democratize AI uh via enabling access uh scale secure and uh scalable access uh to data and model and uh as a base layer of AI our mission is really to create a a foundational framework um that's uh enabling the universally accessible uh AI landscape to people please hey guys I'm Joshua here so I'm the founder of rss3 uh we've been here since 2021 we been advocating for open web and of course open as one of the most important parts of upper job um basically what we do regarding open a is BAS uh is just open data open models and of course open Frameworks and uh I think one of the Frameworks that we have is called open Agent currently um actually used by a lot of different products to build onchain agents and um and at the same time we do uh gather we process and Abstract data from mostly we three and open protocols and for uh training us agents different stuff so uh yeah that's basically about us excellent please hello everyone my name is Yu I'm the founder of Kito AI uh been around since 2021 before that I was a traded a citadel for my career the reason why I started Kito AI was because I was just kind of doing research on the side on on crypto stuff and then realize that information is extremely fragmented uh most of the stuff is unindexed by a traditional search engine such as Google so I started building a search engine and that is Kito AI today we power somewhere between 400 500 different teams in crypto either on the research and trading side or kind of on the protocal side um we are actually now also kind of U doing a lot of research on the decentralized AI side and also kind of a user um user own data side of uh side of things as well which which is is also part of the reason why if you think about unique positioning we are a AI company but serving crypto audience so naturally we want to bring the best of of Both Worlds to be Sol solving some of the some of the unique challenges happy to share more okay Lucas please hello everyone uh my name is Lucas I'm part of the funding team of nilon and my main focus at nilon at the moment is the external side so Source curate support companies as to start building on the N Network my journey with Nan started obviously the beginning for years ago um I was actually doing a time where I finished up uh my degree at Cambridge University uh I did a couple of other degrees during that time in in the UK and I got fascinated by the potential uh with Nan to fundamentally change the way we secure data and fundamentally change the way we process information um and bring back control of of very sensitive information with the same principles as Bitcoin and crypto started right of self- sovereignty trustless and and decentralization um and yeah for the last three years we have been working really hard in building out what we call Millian blind computation network uh blind computation as the name implies is essentially a decentralized network of nodes who all blind right they don't see what they store nor they see what they process uh and the way we achieved that is through different privacy enhancing Technologies such as multiparty competition and others um to Great fundamentally new design space and infrastructure for the crypto space and Beyond uh on our test net we have seen really exciting implementations already ranging from Healthcare like genomics analysis to self- Sovereign data sharing use cases defi use cases AI use cases and many others thank you so much for that so my first question to the panel is yesterday uh Ilia said near is the home for user owned AI data and maybe please could you reflect some thoughts on why privacy is becoming so important these days and we're hearing again and again about privacy breach of data when you actually try AI system there are attacks on the systems and then the data is being you know taken out Etc why can't I just give my data to you know this big corporations Microsoft Apple meta Etc you know and trust them and then they seem to you know maybe use it so is it is it something should I trust these big companies and we are now in this web3 near conference maybe could you please comment on why privacy is so important why we should own the data to the audience please do you please um when it come to privacy there are two type of privacies uh and a lot of times I think uh uh crypto space definitely talk a lot on the user privacy individual privacy side um but also Enterprise privacy is another type of a privacy that matters a lot right now uh and I think the Privacy become more like a real real thing uh especially after J uh recently I don't know if you guys saw the emerging 10 top 10 emerging Tech that's uh published by word coin uh word uh economic foring the second one is actually privacy enhancing technology that's enable the global uh collaborations at scale and specifically it's about data uh so uh a little bit I think why this really become a thing is think about ji now it's become almost like a blind box where all the data can be put inside and uh help increase the performance and the power of a model uh without been not noticing that the data is going into the model so in that way now uh we are actually the first time on the same side with Enterprise as users because everybody now is afraid that their data is going to be you know stolen in some way in this process of building AI so that's why uh I think uh from the very beginning users always uh uh is concerned about some of their sensitive information is stolen or like a linked in the process um and uh uh I think plenty of people can talk about that in in in the group um but also now I just really want to emphasize uh we are in a very good position at the moment because everybody care not only individuals but Enterprise care as well now Joshua your thoughts please I think there are two parts when it comes to data uh privacy especially with with AI models um first is about safety so um there are certain data that's kind of really private to yourself right and I want to make sure that um let's just say an AM AEL doesn't access that doesn't doesn't know something that that it doesn't it's not supposed to know about yourself and or or it shouldn't be helping someone else to do you know malicious things to you right because because they you know if if the a model was trained somehow with some kind of sensitive data that that might you know work not so good on your part great um so safety is one thing but I think of course people have been focusing a lot more about uh fairness or or distribution of of of rewards stuff right if you look at chat gbt um as they States on website it it says it's trained with publicly available information on the internet and uh when it comes to public information there are a lot of stuff that's um you know you know in a way it should be uh belongs to you somehow they they might not access that data as as as as as pure public data right we we can't we can't ensure something like that so um when they're making millions of dollars out of your data um we we we got to make sure that this is fair right corre so um I think two parts safety and and fairness is is the sort of like the core key here for data privacy and especially in in AI youo your thoughts yeah maybe I just chime in here so obviously I'm going to touch on the Privacy side of things but if I think about the data ownership part I really think about two parts one is privacy the other one is the ownership part that leads to monetization and actually owning the data and all the economic benefits coming from the data as well so maybe just using our own example so today we power all these clients and a lot of our clients are really care about hey like for example we have a co-pilot feature within Kito people can ask any questions and get an answer from any specific crypto perspectives and people are really concerned that hey like when I ask a question do you store the data does CH GPT for example store the data to put things into context we do serve some of the very sensitive uh entities such as all the listing teams at every single exchange um or the training firms who doesn't want to dis disclose what kind of um ticket and topics that they're currently researching before anyone fronts front runs them so the essentially people are really concerned um you know in terms of where the data is being stored uh who has access to the data whether it's going to be you know going going to be retrained all that kind of stuff currently there's no very very good solution apart from just being basically just having a good will and good faith in terms of all these either cital side obviously um but also the chbt side or whatever the kind of the end points that that people having without compromising obviously the performance side of things you can obviously host you know data and computer and all that kind of stuff but currently I think that's that that that is a pain point the other thing I'm going to share is on the monetization side of things uh there's also a rising awareness in terms of whether I can actually make money of my data obviously you know that there's a whole um you know whole range of discussions after you know chbt came out and then people you know a lot of content creators saying that hey like you're not paying me for my for my dates nowadays I I would say that in the in the model training part is actually pretty difficult because the attribution and set of things but actually nowadays even on the search engine side of things right so when people are giving you links when uh let's say like even for our example when when uh when we index people's data and these data are public can we actually reward these people and rather than kind of paying for example ax um and some of the other you know platforms we can actually reward these users ourselves at the same time a lot of the new protocols that actually you know come out and and and start you rewarding some of the content creators as well for example igen L just dat a season 2 airdrop that actually reward all the content creators um who have been talking about igen layer so I think this is a new matter in terms of all these you know public data that's directly attributable um you can actually as user as content creator you can actually own the data and you can actually get you know monetized on the data and actually own the economic benefits of the data um but but obviously in a privacy preserved way which is very key because a lot of the people don't want to link their uh social identity with a with a onchain wallet address Lucas your thoughts yeah I mean a lot of good points have been made already uh but just to build on that I mean every single time we interact with AI today we mely a lot of very sensitive information right whenever we train those systems and whenever we run those systems and that's of course a big deal for Enterprise um you know whenever you're um sending sensitive information company specific information confidential information to these proprietary a models uh you have to trust these companies that they don't share the data they don't misuse the data they don't exploit the data and that's a trust assumption that um empirically Enterprises are not willing to to take easily um but at the same time we see on the user retail side that in an age of personalized a where we are about to create our personalized agents for pretty much every single task agents that know us better than we know ourselves we have to share and sacrifice a tremendous amount of incredibly sensitive information right our Healthcare records our Hobbies our messages our sleep data our nutritional data that's a level of sensitivity that we have never seen before like this in the history of of humanity I would argue yes we have seen sensitive data being shared maybe the last 15 20 years but this is a degree that goes Way Beyond what we've seen so far and I think to enable such a future to enable the benefits of such a future we need to create new infrastructure infrastructure that fundamentally preserves privacy of users whil at the same time enable them to use the benefits of the all these tools and the last Point around um some of the training aspect as well when you train these models if we look at the amount of data that was used to train for instance jet gbt or llama and the likes and then we look at all the private sensitive data that sits on the edge still with with us as users but also with inter prizes then it becomes very obvious that to train the next generation of models uh we need more probably than just public data and synthetic data derived from the public data which means we need to tap into private data and to tap into private data particularly on the Enterprise side we need privacy enhancing Technologies where Enterprises can make sure that the data always remains private but nonetheless they can utilize that data train models and subsequently use those models to generate the next you know generation of of intelligence yeah thank you Lucas for that so my next question to the panel is which are the industries you think are really in the most urgent need of this privacy is it health care or is it defense is it finance and what are the best practices you think uh you know in your view to process the data securely you have Federated learning approach blind storage that Lucas mentioned and then uh zero knowledge based techniques so maybe start with you maybe want to comment on you know Enterprise or or private which one which industry what are the you know your top picks on how to process the data securely please yeah it's uh actually funny that most of the industries you you mentioned are more or less so we have customers there so accept defense uh but I'm not going to be surprised if one day sure then some of the customers there uh but uh I think one example I definitely think like a healthcare uh Financial to uh both of the indust will be highly sensitive in terms of the the information uh uh what they want to share and actually according to like Lucas mentioned I really love the comment on we need to tap into the private data uh if we you know like I do the math uh gp4 is only train on one paby data but JP Morgan alone have 150 py data uh in house but how can we really help them to you know like unlock the the data power that's another thing um so it goes to some of the best practice we observed uh in the market and what uh some of the uh tactics that we techniques we are adapting in house uh when it come to privacy there's a two type of privacy uh we're talking about the storage privacy the data privacy and the the the compute privacy so for the compute privacy here uh right now actually T is becoming more ready in production so uh a few actually a few weeks ago we did a another Benchmark testing on running a llama 3 on top of a uh h100 uh actually just one h100 with a t the performance is only 10 to 30% worse than without T so actually it's getting very you know like getting much better than before uh that's the compute privacy side and uh the data privacy side uh I think uh uh uh Nan is definitely using some of the uh very good tools like uh uh MPC uh and also like a data privac encryption such as like uh at rest for data storage and in motion for uh streaming data processing um those are all good and F is uh not production ready phase yet because we did the test but we felt like it's getting closer uh and the last thing is uh uh Google Deep Mind actually has done a lot of work uh such as like D locco and D Pico U that just really slice the data and compute into a lot of smaller components and that's actually what really enable uh what we call decentralized training and decentralized uh uh fine-tuning uh that will become more like a a reality uh in the future um but a lot of the are really uh something we adopt something we're still seeing how it goes and for example when it come to the medical customers we're working with uh uh ironically they don't even like of any of them they just straight say no data go out it stay all locally with us so what we have to do is in the end of the day uh help them to because we are building this Universal framework so in the end of the day they are uh hosting their data in house um but uh we uh help a foundation model uh provide to deploy the model into a t and Trend and fine-tune the model on on on on top of the C cretive uh or like the confidential Medical Data those people have and the benefits here is now the medical data provider is not afraid that the data is not going out and also like the but the foundation model owner now is not afraid their model weight is going to be stolen in this process because everything is encrypted in or like a protected in the te environment so uh it's a both side hat and in the end of the day the final model is being used by some of the 500 uh Fortune 500 companies yeah yeah sure please over to you um since pretty hard to say like like which industry or which particular project or companies does the the best right yeah um a lot of things are happening in in healthcare in in in finance a lot of different things happening in crypto as well um I I'll just share a little bit about what we work on um I mean like more related to to to you know publishing content this kind of stuff I think uh if you look at uh traditionally in web one web 2 era the this kind of privacy is in a way protected by a centralized agency right if you want to say okay I'm GNA send out a a Facebook post or a Twitter post that's only available to particular kind of viewers or audience then that company's going to handle it for you right they they they promise you that they're going to handle you well but of course their their people can you know somehow see it and also if they do you know things not uh the right way they should be then there's going to be you know bad things happening so I think one of the things that we have been uh advocating for is is open information right open have an open information but at the same time when we advocate for open we also advocate for strong privacy stuff or like strong privacy posting that's that's basically uh enforced by encryption right if you if you actually post something that's that's encrypted if if you know that when uh you you post something on a on an open framework that would be in in a way accessed by other people freely um and there's no centralized agency protecting you or you know then you're going to be more careful right in a way then um you somehow have to make sure that data is encrypted yeah right even though in a public space you want to make sure that you know that specific people or like that specific group of audience that's we would be the only ones that would access please your thoughts yeah so I think um just maybe I just share some of the experien and the conversations that we've been having um personally if if you look at the Spectrum of all the Enterprises versus all the consumers uh I would think that all the Enterprises going to be the first adopters just cuz typically Enterprises care a lot more about privacy than than consumers and this this could be a take that some of you guys may disagree but this is kind of just based on my convers conversations and observations as well and then within the Enterprise part I my personal prediction is the the the entities that are closest to money IE finance will be the first adopters because there's a very very clear Roi in terms of hey if I if I adopt this then then that's going to be X prod productivity boost it's going to be X return if I don't adopt this because the privacy concerns all that kind of stuff I need to hire five more people or 10 more people essentially doing the same minor job versus some of the some some of some some of our competitors so I think that's a very very clear uh value proposition as is all kind of the same conversation that we've been having with our with our clients everyone essentially wants us to be tapping into their private data you know all all the all the stack data all the notion data all the you know all the basically all the you know private data that's um stored on on on their third party softwares to be able to have the same um you know data injection that that can that they can essentially leverage but privacy is currently the biggest concern is a zero versus one and decision um in in their mind so which industry lookas you Nan is focusing on is it Finance defense Healthcare or in Industry can you reflect on the on the conversation please with the yeah I mean I don't think it's just the one or the other I think um and obviously all the industries that are dealing with sensitive information are generally very vary when it comes to potential leakage of information um I very much agree with a lot of points you just brought up there I mean obviously Healthcare is is one of the strongest ones because we're dealing with you know DNA information or even like if it's not DNA information very sensitive Healthcare information and using AI that gives you tips and um you know feed bag of how you live your life that's you know obviously a level of sensitivity that you don't necessarily want to share with anyone and um healthcare does as a consequence is a big one Financial Services equally um like having agents where um they can actually know you really well know you better than you know yourself what's your risk appetite and and everything else and then subsequently be able to execute traits that Taps into level of sensitivity to evaluate your risk appetite and the likes which is you know very important to protect right so final Services Healthcare are obvious choices but the reality is even even agents that are uh more user focused rather than Enterprise focused like relationship advisers right and the lik I do think I'm more optimistic that users start caring uh about privacy in that aspect because again the last 15 years uh one may argue that the one or the other might not have cared that much about it but I think now when we enter the territory of relationship advisors um financial advisers and the likes this is a level of sensitivity if your data that we have as I said before never seen before and I think proves a consequence will become uh incredibly important last point on the technology side the different Technologies there's the cryptography based Technologies like NPC and volume of encryption they arguably are the most secure ones because it's cryptography based um there isn't the the single point of failure uh there but it's true that sometimes the performance uh doesn't necessarily yet uh achieve what needs to be achieved uh and at the same time then we have trusted execution environments which I mentioned which are on the harbur side what gets me very excited is having the model of joice to orchestrate between different privacy and hand Technologies because there's no one siiz fit all all of them have different strengths all of them have different weaknesses and being able to enable developers to build applications that can tap into the right privacy and Technology For The Right Use case that's very powerful uh and I think uh new design space that will unlock a lot of new exciting use cases going forward for every user for every application that have different requirements um yeah yeah yeah just want to add that near and announced yesterday they want to use 1.6 trillion 1.4 trillion parameter data where you can train your own model with 500 million and some of the Technologies to use distributed learning is not even available one of the papers from Google Deep Mind actually coming on your point you know mentions that that some of the Privacy Computing text is not even there so there's a lot we look forward to you know hearing from you and reading your research going forward so with this I conclude let's thank the the panel thank you very much for engaging conversation really appreciate thank you","transcript_synced_at":"2026-11-03T05:19:14.506Z","segments":[{"start_seconds":1,"text":"hi welcome everybody Welcome to the to the uh to the panel today um why don't","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1s"},{"start_seconds":7,"text":"uh please starting from here introduce yourself first very briefly about me my name is nag found SE of G20 group","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=7s"},{"start_seconds":15,"text":"investment and trading firm based out of Zurich Switzerland background in PhD in machine learning and AI really","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=15s"},{"start_seconds":22,"text":"privileged to be here please over to you hey guys uh I'm Chi I'm the co-founder and the CEO of uh kite Ai and Z block","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=22s"},{"start_seconds":30,"text":"Labs so uh a little bit about myself uh I did my PhD from UC Berkeley focus on machine learning and CAO inference and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=30s"},{"start_seconds":37,"text":"after that uh started uh uh Automated machine learning platform uh startup with a few friends and then realized","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=37s"},{"start_seconds":46,"text":"data is actually the biggest Boton in industry for real AI adoption then uh join data BRS lead the data engineering","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=46s"},{"start_seconds":54,"text":"Solutions there as a um product management lead and now uh Zena blog or uh kite AI is building a open composable","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=54s"},{"start_seconds":63,"text":"U foundation and that democratize AI uh via enabling access uh scale secure and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=63s"},{"start_seconds":71,"text":"uh scalable access uh to data and model and uh as a base layer of AI our mission","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=71s"},{"start_seconds":78,"text":"is really to create a a foundational framework um that's uh enabling the universally accessible uh AI landscape","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=78s"},{"start_seconds":87,"text":"to people please hey guys I'm Joshua here so I'm the founder of rss3 uh we've been here since 2021 we been advocating for","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=87s"},{"start_seconds":95,"text":"open web and of course open as one of the most important parts of upper job um basically what we do regarding open a is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=95s"},{"start_seconds":103,"text":"BAS uh is just open data open models and of course open Frameworks and uh I think one of the Frameworks that we have is called open Agent currently um actually","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=103s"},{"start_seconds":111,"text":"used by a lot of different products to build onchain agents and um and at the same time we do uh gather we process and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=111s"},{"start_seconds":118,"text":"Abstract data from mostly we three and open protocols and for uh training us agents different stuff so uh yeah that's","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=118s"},{"start_seconds":125,"text":"basically about us excellent please hello everyone my name is Yu I'm the founder of Kito AI uh been around since","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=125s"},{"start_seconds":133,"text":"2021 before that I was a traded a citadel for my career the reason why I started Kito AI was because I was just","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=133s"},{"start_seconds":141,"text":"kind of doing research on the side on on crypto stuff and then realize that information is extremely fragmented uh most of the stuff is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=141s"},{"start_seconds":148,"text":"unindexed by a traditional search engine such as Google so I started building a search engine and that is Kito AI today","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=148s"},{"start_seconds":156,"text":"we power somewhere between 400 500 different teams in crypto either on the research and trading side or kind of on","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=156s"},{"start_seconds":163,"text":"the protocal side um we are actually now also kind of U doing a lot of research on the decentralized AI side and also","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=163s"},{"start_seconds":173,"text":"kind of a user um user own data side of uh side of things as well which which is is also part of the reason why if you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=173s"},{"start_seconds":181,"text":"think about unique positioning we are a AI company but serving crypto audience so naturally we want to bring the best","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=181s"},{"start_seconds":188,"text":"of of Both Worlds to be Sol solving some of the some of the unique challenges happy to share more okay Lucas please","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=188s"},{"start_seconds":196,"text":"hello everyone uh my name is Lucas I'm part of the funding team of nilon and my main focus at nilon at the moment is the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=196s"},{"start_seconds":204,"text":"external side so Source curate support companies as to start building on the N Network my journey with Nan started obviously","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=204s"},{"start_seconds":212,"text":"the beginning for years ago um I was actually doing a time where I finished up uh my degree at Cambridge University uh I did a couple of other degrees","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=212s"},{"start_seconds":220,"text":"during that time in in the UK and I got fascinated by the potential uh with Nan to fundamentally change the way we","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=220s"},{"start_seconds":228,"text":"secure data and fundamentally change the way we process information um and bring back control of of very sensitive information with the same principles as","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=228s"},{"start_seconds":238,"text":"Bitcoin and crypto started right of self- sovereignty trustless and and decentralization um and yeah for the last three years we have been working","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=238s"},{"start_seconds":246,"text":"really hard in building out what we call Millian blind computation network uh blind computation as the name implies is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=246s"},{"start_seconds":254,"text":"essentially a decentralized network of nodes who all blind right they don't see what they store nor they see what they process uh and the way we achieved that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=254s"},{"start_seconds":261,"text":"is through different privacy enhancing Technologies such as multiparty competition and others um to Great fundamentally new design space and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=261s"},{"start_seconds":270,"text":"infrastructure for the crypto space and Beyond uh on our test net we have seen really exciting implementations already ranging from Healthcare like genomics","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=270s"},{"start_seconds":278,"text":"analysis to self- Sovereign data sharing use cases defi use cases AI use cases and many others thank you so much for","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=278s"},{"start_seconds":285,"text":"that so my first question to the panel is yesterday uh Ilia said near is the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=285s"},{"start_seconds":291,"text":"home for user owned AI data and maybe please could you reflect some thoughts","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=291s"},{"start_seconds":298,"text":"on why privacy is becoming so important these days and we're hearing again and again about privacy breach of data when","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=298s"},{"start_seconds":306,"text":"you actually try AI system there are attacks on the systems and then the data is being you know taken out Etc why can't I just give my data to you know","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=306s"},{"start_seconds":315,"text":"this big corporations Microsoft Apple meta Etc you know and trust them and then they seem to you know maybe use it","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=315s"},{"start_seconds":324,"text":"so is it is it something should I trust these big companies and we are now in this web3 near conference maybe could","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=324s"},{"start_seconds":331,"text":"you please comment on why privacy is so important why we should own the data to the audience please do you please um","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=331s"},{"start_seconds":339,"text":"when it come to privacy there are two type of privacies uh and a lot of times I think uh uh crypto space definitely talk a lot on the user privacy","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=339s"},{"start_seconds":348,"text":"individual privacy side um but also Enterprise privacy is another type of a privacy that matters a lot right now uh and I think the Privacy become more like","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=348s"},{"start_seconds":358,"text":"a real real thing uh especially after J uh recently I don't know if you guys saw","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=358s"},{"start_seconds":365,"text":"the emerging 10 top 10 emerging Tech that's uh published by word coin uh word uh economic foring the second one is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=365s"},{"start_seconds":372,"text":"actually privacy enhancing technology that's enable the global uh collaborations at scale and specifically","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=372s"},{"start_seconds":379,"text":"it's about data uh so uh a little bit I think why this really become a thing is think about ji now it's become almost","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=379s"},{"start_seconds":387,"text":"like a blind box where all the data can be put inside and uh help increase the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=387s"},{"start_seconds":394,"text":"performance and the power of a model uh without been not noticing that the data","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=394s"},{"start_seconds":400,"text":"is going into the model so in that way now uh we are actually the first time on","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=400s"},{"start_seconds":408,"text":"the same side with Enterprise as users because everybody now is afraid that their data is going to be you know","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=408s"},{"start_seconds":415,"text":"stolen in some way in this process of building AI so that's why uh I think uh from the very beginning users always uh","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=415s"},{"start_seconds":423,"text":"uh is concerned about some of their sensitive information is stolen or like a linked in the process um and uh uh I","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=423s"},{"start_seconds":431,"text":"think plenty of people can talk about that in in in the group um but also now I just really want to emphasize uh we are in a very good position at the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=431s"},{"start_seconds":440,"text":"moment because everybody care not only individuals but Enterprise care as well now Joshua your thoughts please I think there are two parts when it comes to","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=440s"},{"start_seconds":447,"text":"data uh privacy especially with with AI models um first is about safety so um there are certain data that's kind of","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=447s"},{"start_seconds":456,"text":"really private to yourself right and I want to make sure that um let's just say an AM AEL doesn't access that doesn't","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=456s"},{"start_seconds":463,"text":"doesn't know something that that it doesn't it's not supposed to know about yourself and or or it shouldn't be helping someone else to do you know","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=463s"},{"start_seconds":472,"text":"malicious things to you right because because they you know if if the a model was trained somehow with some kind of sensitive data that that might you know work not so good on your part great um","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=472s"},{"start_seconds":481,"text":"so safety is one thing but I think of course people have been focusing a lot more about uh fairness or or distribution of of of rewards stuff","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=481s"},{"start_seconds":489,"text":"right if you look at chat gbt um as they States on website it it says it's trained with publicly available information on the internet and uh when","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=489s"},{"start_seconds":498,"text":"it comes to public information there are a lot of stuff that's um you know you know in a way it should be uh belongs to","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=498s"},{"start_seconds":506,"text":"you somehow they they might not access that data as as as as as pure public data right we we can't we can't ensure","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=506s"},{"start_seconds":513,"text":"something like that so um when they're making millions of dollars out of your data um we we we got to make sure that this is fair right corre so um I think","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=513s"},{"start_seconds":522,"text":"two parts safety and and fairness is is the sort of like the core key here for data privacy and especially in in AI youo your thoughts yeah maybe I just","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=522s"},{"start_seconds":530,"text":"chime in here so obviously I'm going to touch on the Privacy side of things but if I think about the data ownership part I really think about two parts one is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=530s"},{"start_seconds":539,"text":"privacy the other one is the ownership part that leads to monetization and actually owning the data and all the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=539s"},{"start_seconds":547,"text":"economic benefits coming from the data as well so maybe just using our own example so today we power all these clients and a lot of our clients are","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=547s"},{"start_seconds":556,"text":"really care about hey like for example we have a co-pilot feature within Kito people can ask any questions and get an","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=556s"},{"start_seconds":562,"text":"answer from any specific crypto perspectives and people are really concerned that hey like when I ask a","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=562s"},{"start_seconds":569,"text":"question do you store the data does CH GPT for example store the data to put things into context we do","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=569s"},{"start_seconds":576,"text":"serve some of the very sensitive uh entities such as all the listing teams at every single exchange um or the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=576s"},{"start_seconds":584,"text":"training firms who doesn't want to dis disclose what kind of um ticket and topics that they're currently researching before anyone fronts front","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=584s"},{"start_seconds":592,"text":"runs them so the essentially people are really concerned um you know in terms of where the data is being stored uh who","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=592s"},{"start_seconds":600,"text":"has access to the data whether it's going to be you know going going to be retrained all that kind of stuff currently there's no very very good","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=600s"},{"start_seconds":608,"text":"solution apart from just being basically just having a good will and good faith in terms of all these either cital side","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=608s"},{"start_seconds":615,"text":"obviously um but also the chbt side or whatever the kind of the end points that that people having without compromising obviously the performance side of things","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=615s"},{"start_seconds":624,"text":"you can obviously host you know data and computer and all that kind of stuff but currently I think that's that that that is a pain point the other thing I'm","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=624s"},{"start_seconds":632,"text":"going to share is on the monetization side of things uh there's also a rising awareness in terms of whether I can","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=632s"},{"start_seconds":641,"text":"actually make money of my data obviously you know that there's a whole um you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=641s"},{"start_seconds":648,"text":"know whole range of discussions after you know chbt came out and then people you know a lot of content creators saying that hey like you're not paying","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=648s"},{"start_seconds":656,"text":"me for my for my dates nowadays I I would say that in the in the model training part is actually pretty difficult because the attribution and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=656s"},{"start_seconds":664,"text":"set of things but actually nowadays even on the search engine side of things right so when people are giving you links when uh let's say like even for","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=664s"},{"start_seconds":673,"text":"our example when when uh when we index people's data and these data are public can we actually reward these people and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=673s"},{"start_seconds":682,"text":"rather than kind of paying for example ax um and some of the other you know platforms we can actually reward these","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=682s"},{"start_seconds":689,"text":"users ourselves at the same time a lot of the new protocols that actually you know come out and and and start you rewarding some of the content creators","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=689s"},{"start_seconds":697,"text":"as well for example igen L just dat a season 2 airdrop that actually reward all the content creators um who have been talking about igen layer so I think","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=697s"},{"start_seconds":706,"text":"this is a new matter in terms of all these you know public data that's directly attributable um you can actually as user","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=706s"},{"start_seconds":715,"text":"as content creator you can actually own the data and you can actually get you know monetized on the data and actually own the economic benefits of the data um","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=715s"},{"start_seconds":723,"text":"but but obviously in a privacy preserved way which is very key because a lot of the people don't want to link their uh social identity with a with a onchain","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=723s"},{"start_seconds":731,"text":"wallet address Lucas your thoughts yeah I mean a lot of good points have been made already uh but just to build on","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=731s"},{"start_seconds":738,"text":"that I mean every single time we interact with AI today we mely a lot of very sensitive information right whenever we train those systems and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=738s"},{"start_seconds":746,"text":"whenever we run those systems and that's of course a big deal for Enterprise um you know whenever you're um sending sensitive information company specific","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=746s"},{"start_seconds":754,"text":"information confidential information to these proprietary a models uh you have to trust these companies that they don't share the data they don't misuse the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=754s"},{"start_seconds":762,"text":"data they don't exploit the data and that's a trust assumption that um empirically Enterprises are not willing to to take easily um but at the same","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=762s"},{"start_seconds":770,"text":"time we see on the user retail side that in an age of personalized a where we are about to create our personalized agents","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=770s"},{"start_seconds":777,"text":"for pretty much every single task agents that know us better than we know ourselves we have to share and sacrifice","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=777s"},{"start_seconds":785,"text":"a tremendous amount of incredibly sensitive information right our Healthcare records our Hobbies our","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=785s"},{"start_seconds":792,"text":"messages our sleep data our nutritional data that's a level of sensitivity that we have never seen before like this in","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=792s"},{"start_seconds":799,"text":"the history of of humanity I would argue yes we have seen sensitive data being shared maybe the last 15 20 years but this is a degree that goes Way Beyond","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=799s"},{"start_seconds":807,"text":"what we've seen so far and I think to enable such a future to enable the benefits of such a future we need to create new infrastructure","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=807s"},{"start_seconds":815,"text":"infrastructure that fundamentally preserves privacy of users whil at the same time enable them to use the benefits of the all these tools and the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=815s"},{"start_seconds":822,"text":"last Point around um some of the training aspect as well when you train these models if we look at the amount of data that was used to train for instance","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=822s"},{"start_seconds":830,"text":"jet gbt or llama and the likes and then we look at all the private sensitive data that sits on the edge still with with us as users but also with inter","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=830s"},{"start_seconds":839,"text":"prizes then it becomes very obvious that to train the next generation of models uh we need more probably than just public data and synthetic data derived","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=839s"},{"start_seconds":847,"text":"from the public data which means we need to tap into private data and to tap into private data particularly on the Enterprise side we need privacy","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=847s"},{"start_seconds":855,"text":"enhancing Technologies where Enterprises can make sure that the data always remains private but nonetheless they can utilize that data train models and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=855s"},{"start_seconds":863,"text":"subsequently use those models to generate the next you know generation of of intelligence yeah thank you Lucas for that so my next","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=863s"},{"start_seconds":871,"text":"question to the panel is which are the industries you think are really in the most urgent need of this privacy is it","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=871s"},{"start_seconds":879,"text":"health care or is it defense is it finance and what are the best practices you think uh you know in your view to","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=879s"},{"start_seconds":887,"text":"process the data securely you have Federated learning approach blind storage that Lucas mentioned and then uh","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=887s"},{"start_seconds":894,"text":"zero knowledge based techniques so maybe start with you maybe want to comment on you know Enterprise or or private which","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=894s"},{"start_seconds":902,"text":"one which industry what are the you know your top picks on how to process the data securely please yeah it's uh","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=902s"},{"start_seconds":911,"text":"actually funny that most of the industries you you mentioned are more or less so we have customers there so accept defense uh but I'm not going to","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=911s"},{"start_seconds":919,"text":"be surprised if one day sure then some of the customers there uh but uh I think one example I definitely think like a","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=919s"},{"start_seconds":925,"text":"healthcare uh Financial to uh both of the indust will be highly sensitive in terms of the the information uh uh what they want to share and actually","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=925s"},{"start_seconds":934,"text":"according to like Lucas mentioned I really love the comment on we need to tap into the private data uh if we you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=934s"},{"start_seconds":941,"text":"know like I do the math uh gp4 is only train on one paby data but JP Morgan alone have 150 py data uh in house but","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=941s"},{"start_seconds":950,"text":"how can we really help them to you know like unlock the the data power that's another thing um so it goes to some of the best practice we observed uh in the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=950s"},{"start_seconds":957,"text":"market and what uh some of the uh tactics that we techniques we are adapting in house uh when it come to privacy there's a two type of privacy uh","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=957s"},{"start_seconds":966,"text":"we're talking about the storage privacy the data privacy and the the the compute privacy so for the compute privacy here","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=966s"},{"start_seconds":973,"text":"uh right now actually T is becoming more ready in production so uh a few actually a few weeks ago we did a another","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=973s"},{"start_seconds":980,"text":"Benchmark testing on running a llama 3 on top of a uh h100 uh actually just one h100 with a t the performance is only 10","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=980s"},{"start_seconds":989,"text":"to 30% worse than without T so actually it's getting very you know like getting much better than before uh that's the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=989s"},{"start_seconds":997,"text":"compute privacy side and uh the data privacy side uh I think uh uh uh Nan is definitely using some of the uh very","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=997s"},{"start_seconds":1004,"text":"good tools like uh uh MPC uh and also like a data privac encryption such as like uh at rest for data storage and in","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1004s"},{"start_seconds":1012,"text":"motion for uh streaming data processing um those are all good and F is uh not production ready phase yet because we","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1012s"},{"start_seconds":1021,"text":"did the test but we felt like it's getting closer uh and the last thing is uh uh Google Deep Mind actually has done","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1021s"},{"start_seconds":1028,"text":"a lot of work uh such as like D locco and D Pico U that just really slice the data and compute into a lot of smaller","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1028s"},{"start_seconds":1036,"text":"components and that's actually what really enable uh what we call decentralized training and decentralized uh uh fine-tuning uh that will become","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1036s"},{"start_seconds":1044,"text":"more like a a reality uh in the future um but a lot of the are really uh something we adopt something we're still","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1044s"},{"start_seconds":1052,"text":"seeing how it goes and for example when it come to the medical customers we're working with uh uh ironically they don't","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1052s"},{"start_seconds":1058,"text":"even like of any of them they just straight say no data go out it stay all locally with us so what we have to do is","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1058s"},{"start_seconds":1066,"text":"in the end of the day uh help them to because we are building this Universal framework so in the end of the day they are uh hosting their data in house um","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1066s"},{"start_seconds":1075,"text":"but uh we uh help a foundation model uh provide to deploy the model into a t and Trend","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1075s"},{"start_seconds":1082,"text":"and fine-tune the model on on on on top of the C cretive uh or like the confidential Medical Data those people","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1082s"},{"start_seconds":1089,"text":"have and the benefits here is now the medical data provider is not afraid that the data is not going out and also like","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1089s"},{"start_seconds":1097,"text":"the but the foundation model owner now is not afraid their model weight is going to be stolen in this process because everything is encrypted in or","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1097s"},{"start_seconds":1105,"text":"like a protected in the te environment so uh it's a both side hat and in the end of the day the final model is being used by some of the 500 uh Fortune 500","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1105s"},{"start_seconds":1114,"text":"companies yeah yeah sure please over to you um since pretty hard to say like like which industry or which particular project or companies does the the best","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1114s"},{"start_seconds":1121,"text":"right yeah um a lot of things are happening in in healthcare in in in finance a lot of different things happening in crypto as well um I I'll just share a little bit about what we","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1121s"},{"start_seconds":1129,"text":"work on um I mean like more related to to to you know publishing content this kind of stuff I think uh if you look at uh traditionally in web one web 2 era","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1129s"},{"start_seconds":1138,"text":"the this kind of privacy is in a way protected by a centralized agency right if you want to say okay I'm GNA send out a a Facebook post or a Twitter post","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1138s"},{"start_seconds":1147,"text":"that's only available to particular kind of viewers or audience then that company's going to handle it for you right they they they promise you that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1147s"},{"start_seconds":1155,"text":"they're going to handle you well but of course their their people can you know somehow see it and also if they do you know things not uh the right way they should be then there's going to be you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1155s"},{"start_seconds":1163,"text":"know bad things happening so I think one of the things that we have been uh advocating for is is open information right open have an open information but at the same time when we advocate for","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1163s"},{"start_seconds":1171,"text":"open we also advocate for strong privacy stuff or like strong privacy posting that's that's basically uh enforced by","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1171s"},{"start_seconds":1179,"text":"encryption right if you if you actually post something that's that's encrypted if if you know that when uh you you post something on a on an open framework that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1179s"},{"start_seconds":1185,"text":"would be in in a way accessed by other people freely um and there's no centralized agency protecting you or you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1185s"},{"start_seconds":1193,"text":"know then you're going to be more careful right in a way then um you somehow have to make sure that data is encrypted yeah right even though in a","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1193s"},{"start_seconds":1201,"text":"public space you want to make sure that you know that specific people or like that specific group of audience that's we would be the only ones that would","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1201s"},{"start_seconds":1208,"text":"access please your thoughts yeah so I think um just maybe I just share some of the experien and the conversations that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1208s"},{"start_seconds":1215,"text":"we've been having um personally if if you look at the Spectrum of all the Enterprises versus all the consumers uh","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1215s"},{"start_seconds":1223,"text":"I would think that all the Enterprises going to be the first adopters just cuz typically Enterprises care a lot more about privacy than than consumers and","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1223s"},{"start_seconds":1232,"text":"this this could be a take that some of you guys may disagree but this is kind of just based on my convers conversations and observations as well","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1232s"},{"start_seconds":1240,"text":"and then within the Enterprise part I my personal prediction is the the the entities that are closest to money IE","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1240s"},{"start_seconds":1248,"text":"finance will be the first adopters because there's a very very clear Roi in terms of hey if I if I adopt this then","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1248s"},{"start_seconds":1255,"text":"then that's going to be X prod productivity boost it's going to be X return if I don't adopt this because the privacy concerns","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1255s"},{"start_seconds":1263,"text":"all that kind of stuff I need to hire five more people or 10 more people essentially doing the same minor job versus some of the some some of some","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1263s"},{"start_seconds":1270,"text":"some of our competitors so I think that's a very very clear uh value proposition as is all kind of the same conversation that we've been having with","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1270s"},{"start_seconds":1278,"text":"our with our clients everyone essentially wants us to be tapping into their private data you know all all the all the stack data all the notion data","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1278s"},{"start_seconds":1286,"text":"all the you know all the basically all the you know private data that's um stored on on on their third party","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1286s"},{"start_seconds":1293,"text":"softwares to be able to have the same um you know data injection that that can that they can essentially leverage but","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1293s"},{"start_seconds":1302,"text":"privacy is currently the biggest concern is a zero versus one and decision um in in their mind so which industry lookas","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1302s"},{"start_seconds":1311,"text":"you Nan is focusing on is it Finance defense Healthcare or in Industry can you reflect on the on the conversation","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1311s"},{"start_seconds":1318,"text":"please with the yeah I mean I don't think it's just the one or the other I think um and obviously all the industries that are dealing with","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1318s"},{"start_seconds":1326,"text":"sensitive information are generally very vary when it comes to potential leakage of information um I very much agree with a lot of points you just brought up","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1326s"},{"start_seconds":1334,"text":"there I mean obviously Healthcare is is one of the strongest ones because we're dealing with you know DNA information or even like if it's not DNA information","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1334s"},{"start_seconds":1342,"text":"very sensitive Healthcare information and using AI that gives you tips and um you know feed bag of how you live your life that's you know obviously a level","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1342s"},{"start_seconds":1351,"text":"of sensitivity that you don't necessarily want to share with anyone and um healthcare does as a consequence is a big one Financial Services equally","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1351s"},{"start_seconds":1358,"text":"um like having agents where um they can actually know you really well know you better than you know yourself what's your risk appetite and and everything else and then subsequently be able to","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1358s"},{"start_seconds":1367,"text":"execute traits that Taps into level of sensitivity to evaluate your risk appetite and the likes which is you know very important to protect right so final","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1367s"},{"start_seconds":1375,"text":"Services Healthcare are obvious choices but the reality is even even agents that are uh more user focused rather than","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1375s"},{"start_seconds":1383,"text":"Enterprise focused like relationship advisers right and the lik I do think I'm more optimistic that users start","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1383s"},{"start_seconds":1390,"text":"caring uh about privacy in that aspect because again the last 15 years uh one may argue that the one or the other might not have cared that much about it","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1390s"},{"start_seconds":1398,"text":"but I think now when we enter the territory of relationship advisors um financial advisers and the likes this is a level of sensitivity if your data that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1398s"},{"start_seconds":1406,"text":"we have as I said before never seen before and I think proves a consequence will become uh incredibly important last point on the technology side the","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1406s"},{"start_seconds":1414,"text":"different Technologies there's the cryptography based Technologies like NPC and volume of encryption they arguably are the most secure ones because it's","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1414s"},{"start_seconds":1422,"text":"cryptography based um there isn't the the single point of failure uh there but it's true that sometimes the performance uh doesn't necessarily yet uh achieve","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1422s"},{"start_seconds":1431,"text":"what needs to be achieved uh and at the same time then we have trusted execution environments which I mentioned which are on the harbur side what gets me very","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1431s"},{"start_seconds":1438,"text":"excited is having the model of joice to orchestrate between different privacy and hand Technologies because there's no one siiz fit all all of them have","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1438s"},{"start_seconds":1446,"text":"different strengths all of them have different weaknesses and being able to enable developers to build applications that can tap into the right privacy and Technology For The Right Use case that's","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1446s"},{"start_seconds":1455,"text":"very powerful uh and I think uh new design space that will unlock a lot of new exciting use cases going forward for every user for every application that","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1455s"},{"start_seconds":1463,"text":"have different requirements um yeah yeah yeah just want to add that near and announced yesterday they want to use 1.6","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1463s"},{"start_seconds":1470,"text":"trillion 1.4 trillion parameter data where you can train your own model with 500 million and some of the Technologies","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1470s"},{"start_seconds":1478,"text":"to use distributed learning is not even available one of the papers from Google Deep Mind actually coming on your point","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1478s"},{"start_seconds":1486,"text":"you know mentions that that some of the Privacy Computing text is not even there so there's a lot we look forward to you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1486s"},{"start_seconds":1493,"text":"know hearing from you and reading your research going forward so with this I conclude let's thank the the panel thank you very","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1493s"},{"start_seconds":1500,"text":"much for engaging conversation really appreciate thank you","url":"https://www.youtube.com/watch?v=WILyZbf1rQs&t=1500s"}],"youtube_url":"https://youtube.com/watch?v=WILyZbf1rQs","url":"https://near-pedia.xyz/videos/WILyZbf1rQs"}