AMA RCAP: Opportunities and Challenges of AI x Web3
DeAI Pioneers’ Workshop
TL,DR
- Vision and Mission of EMC: EMC is dedicated to fostering innovation and collaboration by integrating AI and Web3, aiming to develop breakthrough (d)apps that redefine the intersection of AI and blockchain technology. This is part of the DeAI Hackathon 2024, highlighting EMC’s commitment to advancing development in this field.
- Decentralization of Computing: Herbert Yang discussed the decentralization of computational power, an area that EMC values. He mentioned that from a Web3 perspective, the decentralization of computing power includes the generation, sourcing, and distribution aspects. The EMC team has been revolutionary in this regard, inheriting the original vision of the DFINITY Foundation.
- Enhancement of Web3 Platforms through AI: Alex Goh emphasized that off-chain AI processing can improve scalability while leveraging the security of Web3 technology. EMC focuses on making computational resources easily accessible and usable, providing developers with platforms that facilitate the easy training and deployment of models, utilizing decentralized computing power.
- Decentralized AI Marketplaces: Alex also mentioned that EMC is planning to launch an AI model marketplace, which will be a decentralized marketplace where AI models, pre-trained models, and algorithms can be shared and traded. This marketplace will use the economic features of Web3 to provide a more secure and open trading platform for AI models.
- User Adoption and Application: Alex stressed the real challenge of making the platform user-friendly for developers, ensuring that the technology can be widely adopted and used to solve real-world problems. EMC aims for mass adoption, validating the success of its platform and technology by providing solutions that are widely used.
- Challenges and Future Directions: The speakers addressed reputational issues in AI and Web3, the necessity for clear use cases, and the potential of social networks and other applications to leverage AI for user benefit and data privacy.
- Data Ownership and Large Language Models: Highlighting the lack of focus on data sourcing, Yang argued for user-centric data ownership and on-chain training of large language models.
- Decentralization of Computation Power: Herbert Yang emphasized the need for decentralized computation, leveraging Web3 to distribute and source computational power, moving away from centralized monopolies like AWS.
We will proceed to the details replay content of this AMA:
Speakers:
Herbert Yang (Asia GM, DFINITY), Jochem Herber (Head of Ecosystem, Nuklai), Alex Goh (Founder & Chairman, EMC Foundation), and Xinwei (Head of Research, MT Capital)
Introduction
The DeAI Pioneers’ Workshop is, first of all, part of the lead-up to the DeAI Hackathon 2024, and secondly is essentially a series of sessions held weekly throughout March that are designed to provide comprehensive insights into the boundless possibilities of AI × Web3.
And since I mentioned it, the DeAI Hackathon 2024, launched by EMC (Edge Matrix Computing), is a global hackathon aimed at fostering innovation and encouraging collaboration. Our goal is to team up with entrepreneurs and pioneers in the AI and Web3 fields to develop groundbreaking (d)apps that redefine the intersection of AI and blockchain technology.
Today, we’re honored to have guests from ICP, Nuklai, MT Capital, and the EMC Foundation. Let’s extend a warm welcome to them. But before we dive into our discussions, let’s have our guests introduce themselves briefly. Alex, Herbert, Jochem and Xinwei.
Q1 What do you see as the most promising application areas in the integration of AI technology with Web3? Why?
Herbert Yang:
Somehow it feels like the very heavy lifting, one that probably can be better answered at the end of this very interesting panel. But anyway, so from the ICP standpoint, in the computer protocol, we see quite a few very interesting areas that with the integration of AI and the Web3 I just go straight to the spare you with all the fluffy marketing terms.
01 I think the first one is computation power should be decentralized.
I think that’s obviously when we talk about Web 3, decentralization is a dominant thing that is often, I think, ignored by the traditional folks from back to the world. Certainly in the area of AI, that is now overshadowing pretty much everything else we do. But computation power I think should be decentralized. So that includes the generation, the sourcing and also the distribution of computation power. Before joining DFINITY Foundation and became the general manager for Asia for ICP, I was a head of startup for Amazon Web Services AWS Greater China. So we had a very good idea of the power of in a way, monopolizing computation power in the whole world. And so that is I think the first thing that we can do a lot from the web 3 standpoint. And one of the main reasons why I think EMC team is doing a phenomenal job because they saw the original vision by the DFINITY Foundation when we created ICP back in 2016. And we funded the foundation in Zurich, Switzerland. We wanted to liberate the computation power of the world that was that’s really one of the main intent from the very beginning and that has not changed. But it’s only taken a few years for the industry to catch up to. You know, we’ve been trying to build. So that’s number one. So you know I will leave that to you. I’m sure Alex will can share a lot about.
There is some progress by the EMC team how they have been revolutionising this space.
02 And in the second area probably compared to the first one is kind of small but it’s very relevant for web three that is smart contract
I think there’s AI think there’s a low hanging fruit, basically auditing smart contract using artificial intelligence and that is already a very popular way of using AI in the traditional Web 2 with the popularity of Copilot from Microsoft.
And so there’s a lot of programmers and now I worry, oh am I going to lose my job and all that. Some of those worries may not be may not be justified, but in any case that is. It certainly calls to into question what a modern day programmer needs to do. And by any case, in the Web 3 area, we the entire a lot of the protocols or most of the protocols are running. So if we have a way to bring artificial intelligence, the data collection layer, the big the training of large language models, as well as influence layer to be unchanged, that means potentially we’ll be able to create a copilot for of AI that can do audits for all the smart contracts. Running on whether it’s EVM non compatible to EVM doesn’t matter. Automatically. So that could be quite powerful. And of course it may or may not be able to completely replace human audits. And then we still for example, DIFINITY foundation. We work with a lot of very reputable and experienced security audit firms such as Trailbids and be all seeing to provide audit service to the emerging startups from ICP ecosystem. But I think the AI can do probably a lot in the improving the efficiency and the scale of auditing smart counter in this area.
03 And the third area I think with a lot of promising potential for integrating AI with Web 3 is really data ownership, right.
We again this area has not received enough attention from the Silicon Valley.
Because now all attention is on is on models, large language models, transformer, but nobody talks about where do we get a data from. I’m not of the view that without a good data, it doesn’t matter what kind of models do we have because the model can only take the inputs and then it runs through that from the inference layer to produce an output. But the data that is right now in a very undesirable place that is not in the pocket for you or myself is a. It’s on the resides on the platforms of a few big platforms such as AWS, Google, Meta, Apple, Microsoft, Alibaba and Tencent. And that doesn’t feel right. I think eventually we need to find a way to put a data ownership which is central to the whole webstery spirit, into the hands of the end users, from consumers, retail users to enterprise users, even government agencies. And that’s how we can I can really harness the power of those data coupled with the power of NVIDIA chips and all the large language models.
04 And the last one is just a training of the large language model. I think should be on chained as well.
That sounds quite difficult, but in DIFINITY Foundation we we’re experimenting ways to do that. On Chain Fully on chained, our founder and Chief Scientist Dominique Williams just released a video a few days ago that has now received more than 1,000,000 views on Twitter. It shows it shows how Dom using on the testnet created by the Difinity team to do an image detection. Using one of the large language models through the Canister smart contract running on ICP and it’s pretty accurate. It was able to detect correctly a Porsche sports car, a Tiger, and also Copy Mart. So this is already happening on ICP because ICP is a fairly unique blockchain that can do computation as well as providing the data storage at scale.
So in any way, I think in summary these four areas is what we see as really promising verticals that has a lot of potential for growth and also for new generation of AI minded applications at least from the ICT ecosystem. Thank you, Andrew.
Xinwei
Yeah, I think a lot of application areas like for example decentralized AI agent for intelligence service like the employment of AI agent on decentralized network to provide like Internet intelligence service such like automated trading, knowledge assistance and security auditing like this approach leveraged the decentralized nature of web 3 studio like enhance transparency and user control while utilizing capabilities to deliver a sophisticated personalized service. Either lower computational and bandwidth requirement for AI inference makes this like a practical entry point into the AI crypto space, Yeah. This fusion promised to like revolutionize various sectors by offering like decentralized, intelligent solutions that are more accessible and user centric. And also data as a citation and decentralized data market like utilizing blockchain you facilitates citation of data address like key, challenging data, accessibility, privacy and the utilization by creating the decentralized platform for data sharing and trading like this area can unlock vast potential of untapped data making it available for like training and fine tuning AI models. This not only improves the quality and efficiency of AI application across different fields, but also like demonstrate access to valuable data, like the emphasis on privacy and incentivization of data sharing through models like Label to Earn, reflect the significant shift towards more equitable and participatory data ecosystems. And also decentralized computing for AI training like although currently challenged by high computational and bandwidth demands of training large AI models, decentralized computing remains a promising area for innovation, like it represent a fundamental shift in how computational resources are sourced and utilized, and like potentially overcoming centralized centralization bottlenecks.
As like technological advancement to continue to address these challenges, the central computing could offer more scalable, efficient and accessible way to training their models, like them democratizing access to AI development and for staying in a new generation of AI applications. And also the last area I think is like model and agent assassination via NFTS.
The concept of converting AI models or agent into a NFT introduce another way to earn, trade and invest in AI technology. While like currently facing limitation in term of depth of integration and innovation like this area still hold the promise of merging a is transformative potential. With the unique ownership and transfer capability of blockchain, this could lead to a vibrate ecosystem where I am models and agent are not just tools but assets. With intrinsic value encouraging more like creative and economic engagement with AI technologies. Yeah. These are the like 4 areas, I think the most valuable, yeah.
Q2 Privacy and data ownership have been significant concerns in AI algorithm development. How do you believe Web3 technology improves privacy protection and data ownership for AI algorithms?
Jochem:
Yeah, thank you. So, Herbert, you already mentioned something really interesting where you mentioned like, OK, we need to think about where the data actually comes from.
Machine where it adds to that. He’s looking for decentralized data sharing platforms and that’s actually where Nuclei comes in. So that’s where you know the ICP/EMC and Nuklai could work together really, really nicely. To answer this question, I think that it has actually two different dimensions. So you’re talking about data ownership and privacy and I think that they’re related but slightly different.
I would like to change the narrative a little bit towards data control, because data control includes obviously data ownership, but it’s much bigger than that.
So to give you an example, you know you should be in control of your data. You should be able to decide what happens with it. You should be able to combine your data with other people’s data so that you can actually use the data to train AI. You should be able to monetize it. You should be able to decide who gets access to your data.
And that is actually, in my opinion, really the foundation of what we do in Web3.
So to me it’s all about data control. And effectively, that’s where blockchain comes in, right? So that’s what the blockchain is good for. Blockchain is great for data tracking, data monetization, data access control, that sort of stuff. And whenever I talk with businesses, this story also resonates really, really well, because when Xinwei mentioned that, yeah, it’s actually still quite difficult for companies to enter this data economy, then I fully second that because it’s still data that economy is still heavily centralized.
What we do at Nuklai is we have a maintenance subnet kind of structure similar to what from avalanche, but then blockchain and fully built for data sharing only. So for corporates that the resumes are really, really well, because imagine that for example, you’re part of a supply chain and the supply chain can have 50, 100 different kind of companies that all need to share data. But I don’t want that data to be public. So for us we use blockchain to build a permission network where data can still be shared safely within the data consortium. This principle drips down all the way to small and medium enterprises and individuals, where eventually individuals can work together on data sets, contribute to those data sets and get paid fairly for their contribution. So there again, it’s very much about data control. And so therefore the story is read told us is yeah, that resonates very, very well with me. So to me it’s about data control.
The previous topic is slightly different.
In the sense that their wonderful new ways to deal with training of AI opposed to the traditional way. So traditionally you would train an AI on top of a data set. But in these new ways, for example via Federated Learning, instead of sending your data to an AI, you do the opposite. So you send the AI to a data set.
AI is trained and then the AI is being sent back to the originator privacy and data ownership. Web three and AI for that matter is a perfect combination. So yeah, this is my answer to this question. I love this question, very close to my heart.
Herbert:
Yeah. Thank you, Johanna. I totally agree with what you just said. I think these two this privacy and data. They are related by different concepts, so we do need to, but they’re very. They’re super important both of them just from ICP standpoint and then let’s talk about, let’s talk about data ownership first. This is probably the most essential one because you just to just to reiterate without without data ownership in the right hand. We’re going to walk into a very dangerous time of humanity. I mean, I totally. I mean I’m I for one, I totally believe the jihad is going to happen at some point. I mean what Frank Herbert wrote about in 1965, or at least by his son Brian.
I do think we are is we have crossed that line. Not in the original definition of Ray Kurzweil when he wrote The Singularity back in 2012. That a line, a line has been crossed where now this beast of AI, whether it’s from, whether it’s from chat, GPT or you know what have you, is becoming very, very powerful and absorbing data that is from all of us without, really without the slightest regard of our preferences, individual preferences, how those data should be used. So in Dfinity foundation, a lot of the applications have been created by talented developers in the last three years to make sure your data will be truly yours, one example is this talented team from Asia. They are famous by the name of Maura Maura Team. But Maura team before they created the very popular application Maura, they created another application called AD Star.
Is that note which is a note taking application. You’re like well note taking validation, you know what’s the big deal right? We have, we have room research, we have, we have Obsidian, we have notion we have a gazillion types of note taking applications. So what’s the big deal? This download is very unique in the way that you as a user, you will be given a canister. A canister is our form of smart contract on in the universe of ICP.
And this canister will be truly yours. You are the only one in the entire world that has that can access to this canister, which is essentially a live, a live version of a docker that runs on the web assembly in with a stable memory. So all your data, all your notes, and you can use that to keep your password, all of them will persist in the stable memory in your canister and as long as you keep enough cycles, which is a measure of computation. The computation that is consumed by the smart contract on ICP, as long as you keep that sufficient balance, your data will be will be able to persist in this canister forever. Well, for as long as ICP as a layer one blockchain can still persist and then that’s as good as it gets right now. So what you can do in this kind of note taking application is that.
I’ll give you guys a very specific use case scenario that really speaks to the power of privacy as well as data ownership, for example.
You, Bob and Alice, They want to collaborate. Oh, they want to do something else. They want to exchange messages. But it’s not convenient for Bob and Alice. To publicize this kind of exchange, I’m sure you can, you get, I’m sure we can all think of many scenarios where you know that you’re running, you run into that kind of scenarios from your friends and just from or from CNN news.
In the past, how Bob and Alice can do this?
In the most discrete fashion, even devised by the head of CIA, this is a real story because in the past they use Gmail. They use Gmail to write a draft that was never sent out. And so both you know both Bob and Alice will have access to these Gmail account, they just never send out the draft. In the hope that this will never leave any digital trail. But they were still given away by the IP address. So General Patriots was caught one of the most decorated soldier in the history of American.
Fell from Greece. There was a There was some years ago what Bob and Alice could have done with this done, note is they can give access to the controller from their own Internet identities. So this smart contract can be controlled by only the Internet identities that belongs to Bob and Alex. And they can access this canister which has all the notes they want to share with each other remotely. Altogether it doesn’t matter. And then these activities will not, will not be detected by anyone else.
Because once they get on ICP which can host data and the entire dicta node application runs on the smart contract of ICP. Nobody will be able to find out where they are Not from the IP address, not from the domain name and Bob and Alice will be the only two human beings in the entire world that can have access to this canister contract. I mean not even from not even AWS can create this kind of environment. So that is one that is 1 use case scenario. You know I will let her the listeners of this call to imagine what kind of use case scenarios this can be replicated
And the other one is just from a from a privacy that is really important is a spinner cache. Another application from ICP ecosystem. I will share the link momentarily.
It provides privacy transaction for Bitcoin as well as ICP. So because another very famous mixer or privacy platform obviously is no longer no longer functioning for the reasons that we are all aware that’s because it’s it was very the way that the way that particular application was running was very centralized, so when the developers, they run into problems with regulatory agencies, the entire application went to dust. That’s not going to be a challenge on ICP, not because we want to do things that regulatory agencies don’t want to see, but because I think that from this anonymous team on ICP, they value users privacy above all else. So using the Threshold technology from the changing technology from ICP, the team is able to create this platform where you can send Bitcoin transactions to another user or ICP without revealing. Your original wallet address, basically it’s a mixer. It’s a mixer that is totally decentralized that no one as long as there is a sufficient user user traffic, no one will be able to practically pinpoint you. So I think this is a back to AI, I think these two areas are they are very they don’t. They don’t receive enough attention from all the other scientists from working in the AI nowadays.
But they’re very, they’re very important for Web 3. So that’s something that hopefully hopefully at some point the two sort of the Web three folks and the and AI they can see eye to eye on this and then collaborate to make sure to make sure we can really we can bring something useful to the users. That is not that is not only can do, not only can do trading as Xinwei mentioned, but as well as just everyday transactions without worrying that their identity will be, will be, will be in the public.
How can AI be leveraged to enhance the performance and capabilities of Web3 platforms?
ALEX:
Thank you, Andrew. Yeah, I mean Herbert, I hear a lot of deep signs where definitely obviously had lost scientists was working on complicated world’s problem. But what EMC said, well, although it’s barring by DIFINITY on the decentralized compute, we want to focus our work on making a compute available and easy to access for users, right or for developers? Sometimes things could be simpler than what we think despite of doing a rocket science research with bunch of scientists you know. One example that I would like to share is well you could actually do off-chain AI processing instead of on-chained right? Having capable or able to leverage the web tree technologies security of what the technology is bringing. AIcomputation can be performed of chain to improve scalability right? Obviously the scalability is another challenge on the blockchain network despite the powerful ICP network. I think we still need to think of the way and more efficiently training the AI models given that it consume a lot of computing power. So I think the integration of AI application in Web3 sometimes could be easier than you think, right? You know, leveraging the technologies that what Web3 has brought to Earth, we could you know technically trying to use the similar practice that we are using in web two while leveraging the Web3 capability.
Technically like what Jochem has shared, like the Federated reigning is that great areas of how we could leverage the technologies on change and for AI model to be trained across the data set while maintaining the data privacy. So sometimes it’s actually simpler and closer to what we think rather than rocket science, right. So we want our models or products that we are building, they are building really close to other uses. Users can adapt it like you know, just like using it, you know, technically is as good at using AWS services. So in our world, I think our vision on EMC, we try to focus on how do we enable platform for developers to be easier to use, to train the models, to deploy the models while having the capability of leveraging the decentralized compute. And in this world, we try to make things easier even from the computing provider side, you could actually come from Wide range of decentralized providers. It could be individuals who own a GPU workstation and it could be someone that like have a mining machine that used to mine, know others tokens and now they’re they can actually use their GPUs to perform AI tasks while still continue to receive rewards. I think it’s more meaningful than just doing you know hash calculation on Ledger. So that’s I mean exactly example so in our platform we support heterogeneous different models of nodes to make it easier. So I think the benefit of that is actually capable of support a lot of resources to be able to repair ties for AI tasks and I think that’s very meaningful.
For the AI development and balancing the discipline, the amount of the compute power for AM models. So that’s another area I think we would like to share. And the other things that you know, we also see the opportunity of what Web3 and AI could be better integrated is the decentralized AI marketplace. Business in our pipeline and we’re actually planning for AI model marketplace where you can leverage platform to host your decentralized marketplace where AI models, pretend models, algorithm can share, can trade, you can even trade, you know trade your pre-train models using the web free economies right. So it’s actually a more secure and open marketplace and at the same time,Data set is always the important element when we talk about AI in the marketplace, you actually allow the data set to be trade like what the nuclear is actually doing. So I think the potential of AI, using AI to enhance the performance and well I would say that the usage of the current WEB3 infrastructures, their limitless potential where we can explore and I just wanted to share this and open up for other guest speakers to check in.
Could you share some specific examples of successful AI implementations in Web3? What insights do these cases offer for the future development of AI and Web3?
ALEX:
I just wanted to shift attention toward the Web Tree economy side. I think the other innovation that Web3 actually brought to this world is a decentralized and open economy, but you allow the tokenomics to actually play part and enabling the innovation of a different business model. So in our world, when we visualize the AI integrated Web3. Actually we never forget about leveraging the Tokenomics capability. A lot of the best case that you can actually leave off with that. So I want to give an example right. So in envisioning the futures, if you’re going to build an AI models, I think we have the opportunity to really build it in an innovation business models instead of a traditional buying and seller relationships. For example in a traditional web tool business you always need to buy computer resources from Google cloud or from AWS but in this Web3 world it could be a co-built models where. You’re actually offering this as part of the services to the application providers. So someone who trained the AM models and then deployed their AI that they could actually do a revenue sharing models. Technically it’s like I’m offering this business model or the model that I train in this web3 platform and for the decentralized compute providers to come in and offering the computing resources to power this then end of the day they could actually share the profit. So I think we are already pioneering a completely new business model where every it could be a co-built models where the application providers and the cloud providers actually come together through the offering these services to end users and for every dollar they are making you know it’s kind of throwback to this ecosystem where everybody that contributing to this ecosystem actually take part of the cut including the traffic providers, those who are actually bringing users of referrers that are supposed to do also get their rewards based on this sharing models. Also I think this is interesting topic that you know we could definitely explore in this completely new world AI business model. And the other thing that you know when you’re talking about information in web tree, I also we also have a very interesting case to share. That one of our developers team actually developed LM models, so they built an AI agents and they actually launched it in a telegram as a services. So is the AI agents that free to use at least for the time being allowing people to pull that AI agents into different group chat and then allowing that AI agent to play the part as an assistant. You could ask that the engine. So this is another innovation business model that we’re seeing that AI business model could be everywhere and it’s end of the day it’s bringing benefit to the users and you could access to AI models from everywhere so the intersection is no longer like you need to download an app. So technically even by using Telegram I could easily leverage the AI agents to generate the pictures or even to help me to summarise the context of what I’m trying to say or even just to saying you know trying to do get that agent to summarise the news that they could be collected across the web, so there are a lot of usage cache if you if you’re really integrating AI into the real world application and we should not limit ourselves in the traditional business model. So it’s an interesting topic. So I open it up and I think Johann would like to chip in. He’s raising his hand. Yeah. So I’ll have your hand continue with Real.
Jochem
Yeah. Thank you so much for this, Alex. And I think that what you mentioned about the AI marketplace is super promising and one of the things that we talked about before has to do with opening up this whole data economy for everyone and you’re spot on there? Absolutely. To give you an idea, we’re working on a very early case with farmers. We’re good at farming. And the farmers, they generate a lot of data, but they’re very hesitant to share the data because, well, they’ve been kind of like abused before, so that trust has been abused before so they don’t want to share data anymore.
However, when we’re talking about the combination between EMC, Nuklai and probably ICP, we can actually build an environment where farmers individually can share their deepened data, for example. Put that in a data set. The data set can then be used to train an AI, and whenever someone buys access to the AI, hopefully on the upcoming EMC marketplace, all those farmers that contributed to the training data will be able to get paid directly. So this changes the data landscape and the data economy, in my opinion, changes it significantly. Because blockchain can facilitate the direct payments, that part is solved. And with these technologies, it actually makes sense for smaller companies as smaller data providers to actually pull their data together, work together on the data, and then others can build businesses on top of the data. So I very much second what you say. So thank you so much for bringing that in. It’s a very good point.
In the current AI x Web3 landscape, what do you consider the most pressing issues that need to be addressed? What are your expectations for addressing these issues in the future?
So your help, awesome. Thank you. To be completely honest, I think that both AI and Web3, we have kind of like a little bit of a reputational problem. As I mentioned before, I talk with a lot of Web Two companies and they also mentioned this reputation of problem that we have. To give you an idea, just look at the New York Times lawsuit against Open AI. New York Times claimed that the data was used to train their AI and obviously they want to see the money for that, right.
And to be honest, I think that they might just be right. And at the same time, the Web3 space is still very much seen as a wild, Wild West. So I always say that it’s a bit ironic that we have a little bit of a trust problem in Web3. Now with all the good things that Alex Xinwei and Habit mentioned before, I’m very much sure that you know using when we use Web3 technologies in the right way, we can actually bring back trust in AI as well, simply because we can track everything that happened. So you can track which data sets were used to train an AI. You can facilitate the direct payments to the data contributors.
And you can even verify the quality of the training data. So I do see that there’s a very bright future for the AI and Web3 landscape, but these are challenges that we need to solve together. Web3 AI is absolutely a match made in heaven, and it’s up to all of us to make sure that what we do, we do it in a responsible way so that we can slowly get rid of that slight reputational problem that we still have in Web 3.
Herbert:
I want to follow up with what Johan just said. I think I very much agree. We do have a reputation problem for both AI and Web 3. But there are many, there are many, there are many layers of this reputation problem. I think it’s a very difficult for the little guys to survive, right? It has become very monopoly based space. Basically it’s a arms race among a few most well funded companies in the world.
01 At the same situation, certainly in America, and increasingly so in China as well, as China is trying to catch up, and where does that leave the independent developers and entrepreneurs? Not a whole lot of space. So that is, I mean that is problem number one.
This is not in my opinion where the industry should be going.
We should be embracing the future where it’s easier to start a company, it’s easier, you know, not necessarily. Everybody has to has to have has a Unicorn become a Unicorn founder to be able to redeem him or herself.
But that is certainly another situation in the last, I don’t know, 10–50 years in the world of startup. So that’s number one.
02 Number two is I think the both AI and Web 3 where we have this awkward challenge of proving to the world that we actually have mass adoption use cases because right now there is none. Yes, there’s a lot of capital that has flow into all the different models etcetera. But you know, one famous, very famous Tier 1 investor, busy investor in China just went on the record and saying that he’s not going to invest in any large language model startups because he sees no end user level applications, Not yet.
And that is very difficult to dispute. So I think that is still, I mean, obviously we face the same challenge in Web Three as well. For the last 15 years, it has been always financial speculation one way or another. And by the end of the day, I mean, I mean, I think we all agree that there’s value for Bitcoin, but how about for others? Can anyone? Prove that as the industry as a whole, we have provided more positive value to the society. They’re not. That is very hard to prove. And people make money and many people lost money.
So I think both industries we are still looking for the use case that is a kill application that can that can make that can create useful. Applications for our neighbors, for our mom and dad. That is hard. That is very hard. So I think that’s a that’s a second one. But I do think there is, I do think there’s a lot of hopeful areas as we’ve mentioned previously, especially in the social in the social networking area because that’s where generative data tends to be generated. So I’m hoping that there’ll be new kind of social networks that cannot they can not only give data ownership to the users, but also be able to run AI modules on-chain. For example, within the ICP we have open chat, open chat that you can do almost like a Telegram on WhatsApp, but it’s entirely running on-chained and we have that can enable users to receive and send emails on ICP blockchain with integration to I think more than other blockchains. So those are the very early stage use cases, but they look promising at least in the right direction. And then there’s another challenge for the in the for both industries, this whole profit versus non profit notion, which is certainly the one of the seeds that triggered this whole. This legal war between the Elon Musk and some and open chat. Because for AI they need to fund, they need to fund themselves, so they need to charge a fee to the users. So my friend who is working in Microsoft, they’re very, they feel on top of the world because they can easily charge extra 10 to $15.00 to the users because of the integration of open of open AI algorithm. But is that where we are going? Is that really where we want to see the industries going? I’m not sure about that because in Web3, we embrace. We got on the path of nonprofit foundation, providing the open source open source code so that everyone can use open source. Code and we want the early users, early adopters to be able to share the profit. I think this area is where probably AI, the way it’s going, and Web 3 is going to have some interesting conflict, or at least dramatically different ways of handled incidences.
03 OK, last but not least is in the challenge is we need to be able to reach consensus if we want to run a GPU based blockchain in the future as DIFINITY Foundation is already planning to do. Our founder Dom mentioned last year that we are in the process of creating this dedicated subnet that consists of GPU machines.
So that the plan is to enable developers to run AI native smart contract in the next 18 to 24 months. So in Divinity we are still working on that one challenge, technical challenge we need to overcome is that with GPU machines, which is which handles accelerated computing certainly, but also in parallel fashion? Can they still produce deterministic results or the SISIMD instructions?
In DIFINITY we believe we can do, we believe that is viable. So we have some early experiment to show that is.
That is going to be a viable path, but it takes some pretty significant effort to make sure we eventually we can execute this. But anyway, I think this is this one. I think this is one area that doesn’t doesn’t get mentioned too often. I mean, a lot of folks in the industry just assume that Oh yeah, one day we can, we can, we can do the AI thing in Web 3. But if the consensus cannot be reached with deterministic results.
For investors in AI x Web3 projects, what factors do you prioritize? Which projects do you believe have the most investment potential?
Xinwei:
First is the technology like innovation and the uniqueness.
For example, like projects that offer novel solutions or advancement or in AI Web3 space like particularly those addressing current eliminations in decentralized computing or AI model and the data sanitation and integration of AI with blockchain like are more likely to succeed.
For example, render network decentralized rendering platform and our decentralized cloud computing platform. They represent solutions with substantial technological innovation
And the 2nd is like market need and application like purchase that address a clear and present needing market, have defined the user base or solve a significant problem are likely attracted more. Attention and funding
For example, our Weave’s approach to permanent data storage offering solutions for data integration and accessibility of tapping to the increasing demand for reliable and immutable data storage.
And third one is like team expertise and leadership.
Lead by like teams with a very strong background with in both AI and web three like as well as the proven track record in their views are more compelling to investors. For example the leadership of a calm Arkham with founders having a deep routine the crypto industry and render network with significant contribution to 3D rendering technology.
Indicative of strong team expertise.
Fourth like the community engagement and ecosystem support and like project with robust and active community as well as partnership and support from other.
Entities in the ecosystem tend to have a stronger opposition or example bit tensors approach to incentivizing global machine learning. Nodes demonstrates a innovative community engagement model.
And the last one is scalability and the infrastructure project that demonstrates scalability both technological Lee and operationally are essential for long term success. So scalability solution proposed by Akash Network and the Infrastructure Innovation by Rander Network are the examples of project addressing these needs. Yeah, I think these are the like 5 factors.
ALEX:
Yeah you mentioned this. To project like I count like five time you need to look out. There are lot more others better project like Earth who is constantly building. OK, I think you’re summarizing, you’re summarizing the fact on what makes a good project. But in the real world, the way I look at this.
Even from as the founders of the Web3 and AI project, I think the most important thing is make it usable, make your solutions applicable and for developers to be able to use it to really, really utilize the resources that you brought into the network. I mean, like ICP is trying to do that, but I think we can do a better job than just saying that we have all this solution, but the most burning is focused on usability and get our developers to use it. This is what we’re trying to do and eventually.
The future is mass adoptions, meaning that whoever platform that are capable of providing solution for developers to build a project and at the same time enable the solution, the project that they build to be able to use by vast majority of users, I think will win the game.
Technically what I’m trying to say here is in the current web3, there’s no uses, they’re only investors. And the only time that you will see a web3 success is we don’t claim ourselves as a web three company. We only say that we are a company where we bring solution that revolutionize the world. We actually help solving the world problem and there are enough user in our platform, they’re really using the decentralized compute, building the AI models and solutions that for everyone to be able to use and not centrally monopolized by company like Open AI Microsoft or whoever you name it, they’re trying to control of a human being in the future. I think there’s a much better vision that’s the that than just, you know, simply talking about infrastructures, because I do not know how much truth is that of this infrastructure is being used. I think use case is more important. So until we have a mass adoption. Nobody has the right to claim that I built a great project. No. You need to have users that really use the platform. And guess what? This is what EMC is all about. We are here trying to really bringing the platform for people to be able to use it and eventually launching a production for the greater human beings to be able to get access to the AI.
