Latest / Crypto Coin Show / This Modular AI Chain is competing with OpenAI! w/ 0G Labs | Blockchain Interviews
Transcript
- 0:00I'm Ashton Addison from the Crypto Coin Show and today on Blockchain Interviews with Michael
- 0:05Heinrich, co founder of zero G Labs. Michael, welcome to the show
- 0:08and thanks for taking the time. Yeah, nice to be here. Thanks
- 0:12for having me. Ashton pretty excited about being on likewise excited to dive into
- 0:18the interception of AI and blockchain and AI is growing so quickly it's hard to
- 0:22keep up. I know your team has been super busy, and I want
- 0:25to dive into the test net, roadmap to the MAInet, everything that's going
- 0:29on with zero G, but I'd love to start off our conversation first of
- 0:32all with just a little bit on yourself in business and blockchain and what got
- 0:36you to starting zero G Labs, and then we can dive into everything.
- 0:42Yeah, it's funny. My first run in with technology really happened when my
- 0:47family, so my dad, my mom, my sister, myself, moved
- 0:51to Silicon Valley and I was kind of put I was around thirteen. I
- 0:56was put into totally new environment and I was a little bit bored at high
- 1:00school, and so I would just kind of show up to my dad's office
- 1:03after school and just you know, utilize the fast Internet and kind of the
- 1:07technology that SAP Labs had, And one day, like a manager just approached
- 1:14me and he was like, hey, I've seen you like a few times.
- 1:17You're interested in making yourself useful. And I'm like, like, what,
- 1:21like what do I do? He's like, how about you learn how
- 1:23to program and you know, help me write some classes. And I'm like,
- 1:26I have no idea how to do that. He's like, no problem,
- 1:30I'll send you to a visual basic food camp and you know, then
- 1:33you help us out. And so four weeks later, I guess I became
- 1:37the youngest programmer there. And because I was so young, they couldn't pay
- 1:41me in cash, so they gave me like free hardware, and I was
- 1:45like totally in. I was totally in heaven had like the latest like think
- 1:48paths and like laptops and all the cool hardware gears at that time. And
- 1:53so that's that's how I came into engineering and building some of my first software
- 1:59programs. And I stuck her throughout high school and college, and then eventually
- 2:04also worked at Microsoft and did some technical product management there and then from there
- 2:08moved over to the business side. I worked for a Dating company consulting Fortune
- 2:14five hundred technology and finance companies. From there, then I moved to the
- 2:20East Coast in Connecticut and worked for Bridgewater Associates on the portfolio construction team.
- 2:27Then I moved back to graduate school at Stanford and built my first web two
- 2:31company. It's a company called Garden, which I scaled to about six or
- 2:37fifty employees at its height, about one hundred million of contracted ARAR and so
- 2:43it became a unicorn and top y combinator company. And then late twenty twenty
- 2:49two, one of my classmates from Stanford, Thomas, called me up and
- 2:53he was like, hey, I invested in this company five years ago called
- 2:55Conflux. Two of the co founders wanted to start something new, more global
- 3:01scale. I interested in chatting with them, and so I said, yeah,
- 3:06of course, let's connect. And six months later of co founder dating,
- 3:10I came to the same conclusion like, Wow, these are the best
- 3:14engineers I've ever worked with. We need to start something. And so those
- 3:16were really the origins of zero g and our journey. It's really started with
- 3:22the core of the team. Yeah, that's amazing. It's such a cool
- 3:27backstory of growing up right into it, and I have something samilar with my
- 3:30father being one of the early innovators in the Internet, and and here we
- 3:34are, and I feel like there's so much innovation still in blockchain. It's
- 3:37the next version of the Internet to be able to be building something that I
- 3:40feel like is going to grow humongously still, especially within now new capabilities of
- 3:46AI. Incorporating that into blockchain, it's an amazing time building. So maybe
- 3:53you can talk about what you saw, you know, you met with those
- 3:57new founders, and then you have that entrepren real journey you've been through this
- 4:00transition of electronics and Internet as it continues to develop, and looking at where
- 4:06blockchain is at right now in AI, what was the problem that you saw
- 4:12that needed to be solved with inventing zero G labs. Yeah, we wanted
- 4:17to do something at the intersection of a what are we passionate to, what
- 4:21is within our domain expertise? And then three what does the space need in
- 4:27order to really be pushed forward? And so after a lot of conversations in
- 4:30the ecosystem with projects like Polygon and Arbitrum and various AI projects, it became
- 4:38pretty clear that data availability was one of those key unlocks, and so what
- 4:45we saw in the space is that the kind of standard was somewhere in the
- 4:49one to ten megabytes per second of data availability capacity. But if you want
- 4:55to truly do let's say, on chain AI Moodel training, we really need
- 4:59to be thinking about tens of gigabytes per seconds, especially for really large models,
- 5:03and so utilizing our backgrounds and I actually haven't introduced Ming or FUN yet.
- 5:10Ming spent about eleven years at Microsoft Research. He wrote some of the
- 5:15first AI algorithms for Microsoft's being search Engine. He's a distributed storage and scalability
- 5:24experts. He wrote some of the key papers in the space like Tango and
- 5:27Curfew, and then FUN was originally his intern actually at MICROSOT Research before they
- 5:31started Complex together. But he equally impressive was a MIT Computer Science PhD at
- 5:39two Olympic gold medals and informatics and also University of Toronto professor. And so
- 5:47with that kind of domain expertise around scalability AI distributed storage, we were then
- 5:53able to architect a really high throughput system. And our core insight was we
- 6:00need to basically segment data into a data storage and into a data publishing lane.
- 6:04So that we can create an aspect of horizontal scalability within that system,
- 6:12and we've made a lot of improvements in the actual architecture and implementation level specifics
- 6:18and so net, each one of our notes on our network can generate about
- 6:23ten megabytes per second, and then per consensus layer, if you add about
- 6:27five thousand notes in that consensus layer, we can get to about fifty gigabytes
- 6:30per second and throughput per consensus layer, and that then gives us an advantage
- 6:36of also lowering costs by about one hundred x compared to current systems like Celestia
- 6:42and IGNDA and so on, which then means the whole preamble to get to
- 6:46the final goal of we can surely build things in a decentralized context just as
- 6:53well as in the centralized context, same performance, same costs. The only
- 6:58limitations now are physical limitations like speed of light and so on. For distributed
- 7:02systems, definitely, Yeah, there's always that speed of light issue, which
- 7:08it would be great to get there. For data availability, maybe you can
- 7:12explain a little bit further for those people who they're using the blockchain a little
- 7:15bit, they don't really understand the inner workings of it. What the the
- 7:21inner workings of like data availability. What is needed, say it for AI
- 7:28or even just for a traditional search engine, is it getting the information when
- 7:32you ask chat ept something, It needs to look for that information and bring
- 7:36it back or find it in a certain location. Can you just explain the
- 7:41data ability availability a little bit more. Yeah, and using AI as an
- 7:46example. So if you're interacting with something like open AI, it's all in
- 7:50a centralized server farm. So you make a request, you know, there's
- 7:55a bunch of stuff that happens in the background with the model around that server
- 7:58farm, a bunch of like infrint task, because essentially you have a lot
- 8:01of GPUs kind of strung together in a centralized setting. And that's very different
- 8:05from a blockchain. Blockchain is a completely distributed system. You have often permissionless
- 8:11ways of interacting with that system, so anybody can join, anybody can generally
- 8:15become a verifier and a kind of proof of state system. So then you
- 8:20take something like Ethereum and it has kind of what's known as a monolithic blockchain.
- 8:24It's got multiple functions. So you've got the execution layer, you have
- 8:30the consensus layer, you have the data availability layer, and then finally you've
- 8:35got the settlement layer as well, and so a monolithic chain does all of
- 8:39these functions, and most of the transaction costs go towards the data availability layer,
- 8:46because basically it's verifying that a particular block has actually been published on the
- 8:50network. And so by pulling out that specific layer, we can then hyper
- 8:54optimize that part and make it significantly cheaper to published data and significantly more performant
- 9:01to have four throughput on a particular blockchain by making that modular and pulling that
- 9:07out of something like ethereum. And so in effect, that's what we're doing.
- 9:11We're hyper optimizing. There's one one function and part of a blockchain.
- 9:16Mhm. That's very cool, and so then go ahead, yeah yeah.
- 9:22And so then when you have an AI specific use case, the way it
- 9:26would then work is, let's say on an execution layer you have a specific
- 9:30inference task. That inference task then goes to us as a DA layer.
- 9:35We don't pull all of the right serving infrastructure together, find the right like
- 9:39GPU locations, load the particular model that's stored into memory, and then kind
- 9:46of serve that to the end user. So it's it's quite a different like
- 9:48back end process, but it ends up going to the same kind of user
- 9:52experience. Mm hmm. That's good to know. And if you were to
- 9:56compare the open ai experience that people are using right now, as you mentioned,
- 10:00it's centralized using multiple GPUs, to something that is decentralized or utilizing blockchain
- 10:07in some way. Is that something that's viable today and with zerogy Labs technology
- 10:15or where exactly is that at where we can have those options freely available,
- 10:18where hey, this is just as good as open ai but also utilizing blockchain.
- 10:26Yeah, so today what's possible is you can utilize open source models like
- 10:30a Lama three or a mistrol, which are definitely rivaling open ai in its
- 10:35capability, and there's a consistent kind of we want to one up each other.
- 10:39Open ai is a fully closed source system, so you can only utilize
- 10:45THEGBT model if you're within their framework and you're utilizing their fine tuning tools and
- 10:50so on. Now, the issue with that is if you imagine running an
- 10:54AI agent that's for example, steering a traffic control system or running a big
- 11:01administrative process in a company like you know, we're going to match candidates with
- 11:05the right jobs when you're in the open AI framework, then the challenge is
- 11:11you don't know what the model was trained on, where the model is sitting,
- 11:16what version of the model you're getting, what the kind of weights and
- 11:20biases were in that particular model, and so you have no control over that.
- 11:24But with kind of an open source model that let's say you can train
- 11:28on your own type of training data, you can know at any point in
- 11:31time, you know what version of the model you're running, what decisions are
- 11:35being made by the model. It's completely publicly verifiable. And that's really the
- 11:39power of combining AI and blockchains together, and so you've got full provenance,
- 11:46full verifiability, total transparency. In fact, you know, you then have
- 11:50an open AI you know, pun intended Yeah, definitely. And do you
- 11:54think that as AI continues to grow there will be more of a discussion around
- 12:01the biased unbiased part of it. You know when in traditional search results,
- 12:07you can search and you can scroll down and see, you know, sort
- 12:09of different results. But with open AI, it's just like here's the answer
- 12:13and that the quotes. Yeah, it may not be accurate, but like
- 12:16it's sort of like this is the end albl truth, But what are the
- 12:18algorithms or the what's the closed source software that's providing you that information? And
- 12:24in sort of who is the person who's saying that this is the truth exactly,
- 12:31and who's deciding what bias to remove from the training data, and who's
- 12:35deciding what the alignment of that particular model is with human interest. That's all
- 12:39happening within one company, and so how as a society can you then have
- 12:45insight into what that one company is doing. In fact, I think I
- 12:48recently just read that the head of alignment for open ai quit because he wasn't
- 12:52comfortable with the practices of open ai. And you know, that's an issue.
- 12:56We're trusting this big entity to do what's right for an entire society.
- 13:01Shouldn't we as a society really dictate what's right for us as a society?
- 13:05And so, how do we really become part of this governance process? How
- 13:09do we make AI a public good? And that's what really we stand for.
- 13:13Yeah, it's a big feat, but AI open ai has grown to
- 13:18be a huge company and there's a lot you know, if you're going to
- 13:20become the next Google search engine and more, then there's got to be a
- 13:24lot into that I'd love to learn more on how your team has been doing
- 13:30this. How long has it been in the works for? In what stage
- 13:33are you guys at exactly right now? Yeah, we've been working, so
- 13:37I went full time in Mayo of last year, so it's almost exactly been
- 13:41a year, which is kind of fun, crazy to see. We released,
- 13:48we came out of stealth about a couple months ago. We made the
- 13:52announcement that we raised the thirty five million precede, and then with that started
- 13:56to do a number of initiatives. So we've all so recently announced our public
- 14:01test net a few weeks ago, and then we're going to announce public test
- 14:07net V two quite soon and then move to main net hopefully sometime in July.
- 14:13Very cool. And with that public test net, is there ways that
- 14:18the end users or developers can interact and test it out. What exactly is
- 14:22possible in the test net? Yeah, so you can run your own storage
- 14:28nodes, you can run your own validator, you can deploy different kind of
- 14:35chains to utilize the data availability property, so you can fully utilize the system
- 14:41and its test net state. Of course, because it's a test net,
- 14:43there's certain things there. We're still upgrading and V two of the test net.
- 14:48We're going to improve some of the kind of storage mechanisms and also the
- 14:52interface with interacting with our test net, but all of the most of the
- 14:58functionality is live on the DA side that we're still kind of testing out and
- 15:03working on as well. So for example, where we'll be moving from a
- 15:09data availability sampling technique to a quorum based technique just for kind of speed and
- 15:13performance gains, and so all of that's going to happen in test and nep
- 15:16YouTube great looking forward to it. And in terms of connecting into the blockchain
- 15:24world, is this something that other blockchains can be cross compatible with or they
- 15:28can interact and use the applications on as well, or is it sort of
- 15:33standalone with zero g No. Absolutely, and for different depending on where you
- 15:39are on the stack. As a blockchain, we can have different use cases.
- 15:43So if you're a layer one, we can support you with doing state
- 15:46snapshots, for example, or you can fully offload your state to us and
- 15:50we can manage that for you. So if you have a blockchain that can't
- 15:54handle a lot of data throughput, then you can just offload it to us
- 15:58and we can deal with it. If you're a layer two, then you
- 16:03can utilize us as both kind of storage layer as well as the data availability
- 16:07layer, and then the same thing for layer three's and app chains as well.
- 16:11And so yeah, we're very modular and you can plug us in how
- 16:15however you need. That's amazing. So would you be able to potentially lower
- 16:19the Ethereum transaction fees for all these people that are paying two high fees,
- 16:23especially when it starts plumping, Well, yes, So that's the whole idea
- 16:29behind the layer twos on Ethereum. It's basically like optimism and arbitrum, all
- 16:34of those kind of low fees. One of the reasons is you're offloading kind
- 16:38of the execution to a different layer, and then you're batching transactions together.
- 16:44And then if they implement a better DA system, then you're starting to talk
- 16:48about fractions of a fraction of a penny in terms of transaction costs. So
- 16:51that allows completely different possibilities, like you can now have fully on chain gaming,
- 16:56fully on chain AI because you almost have gas lids feeds at that point.
- 17:02And so yeah, that's that's kind of the next evolution of blockchains.
- 17:06So as as these layer twos and layer threes and app chains adopt DA solutions
- 17:11like us, then you can create completely different experiences. Mm hmm. I'm
- 17:15really looking forward to that. And you know, I saw some recent updates
- 17:19in the AI world with the newer versions of open ai, and I feel
- 17:23like AI in as an industry is just growing exponentially, and I'm guessing that
- 17:29it's going to be using a lot more data as hopefully, you know,
- 17:33we start generating videos and real time metaverses and all of this kind of stuff.
- 17:41Do you have any predictions on, you know, what we might see
- 17:45in AI in in the next year or even six months, and how much
- 17:51more data availability we might need as it continues to grow. It's it's it's
- 17:57always hard to predict the future, but some of the things I think that
- 18:00are going to happen in the next let's call it five years. Is definitely
- 18:07more specialized agents for particular types of use cases. So in the blockchain world,
- 18:11instead of for example, you know, waking up in the morning and
- 18:15saying, oh, I want to buy this particular coin that's on this particular
- 18:18chain, and I have to go through this decks and this bridge, you
- 18:21can just tell your agent like go buy this for me, and so that's
- 18:25that's one key application there, and then other things in the non blockchain world.
- 18:30So for example, there's going to be full self driving mode from Tesla.
- 18:33I've already kind of tested some of the data out and it's pretty amazing.
- 18:38It still does some ways to go, but we'll start seeing AI agents
- 18:42really interacting with the real world as well. And then we'll see a world
- 18:45where we start having humanoid robots also interacting with our world and they're they're going
- 18:51to have AI models in them, and then we just need to ensure as
- 18:55a society that there's certain features and safety mechanisms built in so that full alignment
- 19:00with human interest is in church. And so I'm starting to see a world
- 19:04like that where AI is part of the fabric of society, whether it's digital
- 19:10or physical. M Yeah, it's it's a pretty crazy future and I feel
- 19:17like it's gonna it's going to grow quickly. So it's good for us to
- 19:22be involved in following along and hopefully capitalizing on the opportunities as they come to
- 19:26grow. Is there any insights into the main net? I said it's coming
- 19:33in the near future, but what has to be done and what else can
- 19:36people do to follow along to make sure that they get involved with it or
- 19:41learn more about it, you know, on day one hopefully. Yeah,
- 19:45absolutely, So the main net we're working super hard. Basically the technology team
- 19:52or like ming and fonts teams are basically working at one hundred and ten percent
- 20:00to get there as quickly as possible, and so we're hoping sometime July,
- 20:03end of July or main phase one, which basically gives you the basic features
- 20:07and a stable environment around storage, the data availability components, and so you
- 20:14can definitely utilize our testa right now to do the things that I mentioned,
- 20:17which are running a storage node, running a validator node, deploying app chains,
- 20:22layer twos to utilize our DA property and so on. So definitely definitely
- 20:26do that. You can follow us on x SO just at zero G,
- 20:33Underscore Labs and my personal one which is at m Heinrich and you can see
- 20:40the latest updates there. And if you're interested in building in the space,
- 20:42we also recently launched an accelerator together with one Piece Labs and those applications are
- 20:48live till mid of June, so definitely apply if you want to build part
- 20:55of the kind of biggest AI meets blockchain community in the world. They're excited
- 21:00about that. That's very exciting. And I can leave the link to the
- 21:03socials for zero G and yourself if anyone wants to connect in the show notes
- 21:07below and to the platform and the test net details and hopefully people will be
- 21:12paying attention. I would love to also have you back on right around that
- 21:15mean that launch if possible. I know you guys are going to be super
- 21:18busy, but it's a good thing because you're helping change the world. So
- 21:22I really appreciate your time today. Michael, all the best with everything zero
- 21:26G Labs, and let's definitely follow up in the near future. Yeah,
- 21:30thank you. That's a big pleasure being here.