Latest / Crypto Coin Show / Decentralized Computing networks explode : Fluence Mainnet | Blockchain Interviews
Transcript
- 0:00I'm Ashon Addison from the Crypto Coin Show today on Blockchain Interviews with Tom Trowbridge,
- 0:05CEO of Fluence. Tom, Welcome to the show and thanks for taking
- 0:08the time. Great, thank you for having me as and happy to be
- 0:11here, very welcome. Excited to dive into LM's blockchain decentralization not always on
- 0:17chain, which I think is an interesting perspective that many also forget about,
- 0:22and what Fluence is working on, and the lead up to the main net
- 0:24as well. I would love to start off our conversation first of all hearing
- 0:27a little bit about yourself, your background leading up to Fluence, what got
- 0:31you so interested in this, and then we'll dive into everything Fluence certainly so
- 0:37the pore Fluence, I helped found a l one called to their Ashcraft,
- 0:41and I joined Manson Lehman pretty much when it was a white paper and was
- 0:47a key part of building the team, raising the money, getting the first
- 0:51council members and helped kind of get everything going from the white paper to the
- 0:57launch. And that was a you know, intensive multi year kind of baptism
- 1:03by fire into the whole kind of blockchain decentralized you know world and ecosystem,
- 1:10So it was a terrific learning experience it gave me. Also, it also
- 1:14was a great Edu technical education because both doctor Van Barry and Man's Harmon were
- 1:21incredibly credentialed and impressive technologists and frankly scientists, and so I learned a terrific
- 1:30amount about the whole industry through that lens, which was interesting. And then
- 1:34when I left, you know, Hadera, there was really important lessons I
- 1:40have gained from that, and you know, the technical fundamentals being solid is
- 1:45number one, and so that was kind of key point. The other was
- 1:51not over promising and kind of delivering what you said you're going to deliver,
- 1:53especially in this industry, I think is really important and some people kind of
- 1:57take that for granted. I don't think you can here and then the other.
- 2:00And this is something I think I wanted to improve on from the Headeric
- 2:02experience was make sure you are very crypto native and very dialed into that community,
- 2:13because it makes an enormous difference in terms of the time and attention that
- 2:19people pay on your play, you know, pay on your project. And
- 2:22so Fluence had those three characteristics and a lot more, but those were those
- 2:28were kind of some some kind of experiences and lessons I took from that.
- 2:34Wow, very cool, and I do love Derea. I'm actually working with
- 2:37some projects on the Hyderic ecosystem still, so it's really great to hear that
- 2:39you were part of the conception of it. So that's very cool. And
- 2:45now moving into Fluence, how exactly is that different and what exactly is Fluence
- 2:50working on in the industry right now? Well, Fluence we call Fluence cloudless
- 2:55computing network, and so you know, if you take a step back,
- 3:00we think we realize that computing started off in people's homes and people's offices,
- 3:06and then over a twenty year period, people were brought onto the cloud and
- 3:10companies and people have moved to the cloud. You use your Gmail, that's
- 3:14happening in the cloud, your you know, Netflix, stuff is delivered from
- 3:16the cloud. Like everything has moved a lot, not everything, but a
- 3:20lot has moved to the cloud. And that has been a cost savings and
- 3:24probably been a security enhancement given the profiles of hosting most of these applications and
- 3:31servers in house. But that has come at a cost, an economic cost,
- 3:38a centralization cost, a censorship cost, and a resilience cost. And
- 3:43so we now are in a world of the several companies that basically own the
- 3:47cloud, and so we think, and they're also terrifically profitable companies obviously as
- 3:53well, and they provide a valuable service. So that's not you expect nothing
- 3:57else. But then Internet itself, the economics lead to centrally, and so
- 4:02because benefits of scale are rewarded, and so we think we've gone from the
- 4:06in house servers to the cloud. But the next and final iteration is cloudless.
- 4:11And what that means is a network that is open and isn't constrained to
- 4:18any one provider, where your application lives wherever somebody will host it and uses
- 4:25whatever open source or frankly closed source components that anyone will host. And if
- 4:31the host it's on as an outager goes down, it can go somewhere else.
- 4:35If you decide you want to leave that host, you can go leave
- 4:38that host immediately. You don't need to unwind, and you don't need to
- 4:43worry about being kicked off for a violation of an eighty page term of service
- 4:47that is arbitrarily interpreted differently one day for particular reason. So that's the iteration
- 4:55and influences building is this cloudless computing network and stack and that is what we're
- 5:01been building for years and are very close to launching. Wow, very cool
- 5:06and yeah, yeah, I feel like so many people use the cloud right
- 5:11now and they just they don't really know how it works or where their files
- 5:15are, who's actually controlling them. They might know, yeah, it's somebody
- 5:18else's server computer, but I feel like it's they don't know how good it
- 5:24could be because the infrastructure hasn't really got there yet. You know. There's
- 5:28I feel like eventually there will be the extinct extinction of the cloud as we
- 5:32know it, where now there's just that handful of companies if we're controlling it,
- 5:36and it will be decentralized cloud and probably people will be able to capture
- 5:43some of those rewards that are so centralized right now. Well, I mean,
- 5:46listen, there's there's a whole bunch of elements that you mentioned there that
- 5:49I agree with, and we should spend a minute of time on. I
- 5:53mean, first, people use the best technology. You don't know and care
- 5:57how it works because it just works, right. You and I don't know
- 6:00the details of the iPhone, but we're happy it works the way it does
- 6:01cloud you don't need to know how it all works. But the problem is
- 6:06that the larger companies get, the more intertwined to become with governments. And
- 6:12this goes back to the railroads. I mean, it's not any company's fault.
- 6:15It just happens that that is the case. And so you end up
- 6:17being you end up having kind of a couple of things happen. But you
- 6:20have censorship potential, and that's one thing. But you also the bigger you
- 6:24get, the fewer other options there are. You end up with sort of
- 6:28aligopolistic pricing, which is sort of similar to monopolist pricing, very high margins.
- 6:32You see these companies' massive margins and customers are you know, use it.
- 6:38They're happy, they're but they're they're locked in and if you want to
- 6:42leave, it's a little complicated. And the other point is that they have
- 6:46prevented the open and open architecture that would drive more innovation. And so you
- 6:54even look at application at the application layer, where you had companies like Twitter
- 7:00or Facebook that used to have open integrations and those are now we're cut off
- 7:05and gone. And so we see a world where people build applications that are
- 7:10open and where integrations are easy and cannot be disabled. And so I think
- 7:15there's a whole generation of open applications which will build on an open cloudless stack.
- 7:23That leads to kind of an innovation acceleration. And so on the one
- 7:27hand, the cloudless stack helps because it saves cost, it reduces lock in,
- 7:33it has all these other benefits. It also has verifiability because you're using
- 7:38blockchain. It's not un chain, but because blockchain is part of the stack,
- 7:42you can validate and verify that computation was performed, which you can't do
- 7:46in a cloud stack. A company can certify it and that's useful, but
- 7:49it's very different than actually having a scryptographic proof to something happen. So there's
- 7:55benefits to the kind of infrastructure layer, if you will, and there's benefits
- 8:00to the application layer of where you can open applications that share data and users
- 8:05in a way that can't be disabled or cut off. I am believer that
- 8:09that is going to lead to a terrific amount of innovation that we sort of
- 8:13have a difficult time imagining. Mm hm No, it's very exciting, and
- 8:16thank you for those insights. Tom Now with the new version of like the
- 8:22cloudless, more decentralized computing network. I've seen solutions pop up, but because
- 8:31of the network effect of things like you know, Gmail or Google Drive and
- 8:35these storages, people are so familiar with that. And also they've refined the
- 8:39user interface, you know, for Grandma, and everyone's able to use it.
- 8:43I feel like you have to build a product in web three that's very
- 8:48you know, as you said, you don't need to know the inner working
- 8:50is in the back, and it also needs to be very easy and user
- 8:54friendly to adopt. And of course it's one thing for somebody who's hosting an
- 8:58application and they probably have more technical expertise to host their app or website.
- 9:03But when you want to do personal file storage or things that people have no
- 9:07technical background. How can you lower the barriers to entry on a cloudless decentralized
- 9:13computing network to allow for people to adopt it quickly and transition from those cloud
- 9:18networks that have all those issues. But people are sort of stuck because they
- 9:22don't want to change, because they don't want to learn anything new. I
- 9:26think we have to separate two elements, which is the infrastructure element from the
- 9:30application element. And so we are primarily targeting the infrastructure layer, which is
- 9:37in your example, would be the developers of the products that would compete with
- 9:43like say the Google applications you mentioned. All right, so but what we're
- 9:48doing is our objective is to make the building of those applications so simple and
- 9:56to commoditize the back ends to such a degree that instead of fifty to sixty
- 10:03percent of developer resources spent building back ends, twenty percent or ten percent.
- 10:09Is so you can have you know, eighty percent focused on the front end,
- 10:15on the user experience, on the user interface, and so we won't
- 10:20be building and working on the applications you're referring to. But our view is
- 10:28that the simplification for developers that we will enable will facilitate more investment in the
- 10:37front end and will also drive more people into the ecosystem to compete to build
- 10:43better products. So that's that's, that's that's where our role in that is.
- 10:48No, that's that's a great point. And you know, I'm not
- 10:50a developer myself, but I've worked on applications. I've used the Amazon Web
- 10:54Services back end, and it is ridiculous. How you know, you also
- 10:58have to take a four year degree to figure out how many functions they have
- 11:03and all these different functions, and they suck you in that. It's like
- 11:07I spent years trying to close my account where there's one hundred different things and
- 11:11it's very complex. I feel like it could definitely be simplified, and also
- 11:13it could definitely be cheaper if we had more decentralized networks. Yeah, I
- 11:20mean, it will definitely be cheaper. I mean to give you a sense.
- 11:22And here's another kind of misconception it's important to clear up. And people
- 11:26think of decentralized networks, or they think of you know, decentralized storage or
- 11:30compute. Sometimes the default is to think of people using their laptops for their
- 11:35phones. We are running in kind of tier one data centers where there's excess
- 11:41compute capacity. So you know, we toured a data center in Las Vegas
- 11:46about four months ago where Netflix is hosted. You know, a whole number
- 11:50of well known companies are hosted. And so I think that's kind of one
- 11:56misconception that want to clear is that this decenttalization does not sacrifice the institutional quality
- 12:05and resilience that I think is required these days. So that is that's that's
- 12:11kind of an important important point. Yeah, no, that's that's that's great
- 12:16to know, because especially with the quality of the apps and the amount of
- 12:20traffic that's coming in, you're going to need things that are going to be
- 12:24able to compete with the warehouse level of computing power. Well, exactly,
- 12:28sorry, and that was my other thing. I forgot the train of thought,
- 12:31which was also price. And so you just have to look and see
- 12:35what the margins are of the cloud providers to see the incredible opportunity there.
- 12:39And so when you are have racks that are unused in the data center,
- 12:46your cost, your marginal cost is pretty much the electricity you're using and maybe
- 12:50some depreciation as well. Right, so there is a terrific amount of margin
- 12:56potential that goes to the customer when you're able to use these. And the
- 13:01key thing is about influence is that if you went to a data center and
- 13:05said I want these thousand servers and I want them for two years, you'll
- 13:11pay a reasonab amount threatend them. But if you said you can give them
- 13:16to me for a week at a time, you will pay a dramatically lower
- 13:22price for a week to week contract. And the reason why it doesn't matter
- 13:26for fluence is because it's an open network. If they decide to go offline,
- 13:31the applications move automatically to a different provider. And so what that does
- 13:37is it commoditizes that compute because it can go anywhere, which leaves those providers
- 13:43with no pricing power and also allows them to monetize. At the same time,
- 13:46they can monetize in small amounts their excess capacity without signing up for a
- 13:52long term contract at a price that they may find some customer willing to pay
- 13:58more than. So you have two benefits that an open protocol has a sort
- 14:03of compelling rationale for why the pricing is significantly lower than a centralized provider that
- 14:11has to invest in those servers, buy them, hold them for a long
- 14:15period of time, or else rent them from somebody else. No, that's
- 14:18a really great point, Tom, because that's how part of the way that
- 14:22they have in traditionally it sort of sucked you in in that you know,
- 14:26lock people in for longer term contracts to give them a bit of a discount
- 14:30on pricing, and you know, other smaller alternatives would cost more, and
- 14:35then you're sort of stuck. And then once you've build your application, there
- 14:37the migration process of something different, it's like it's Almo's not worth your time.
- 14:41Even if new alternatives come out that are cheaper in the future, you're
- 14:43sort of stuck with your application because you don't want to pay to migrate or
- 14:48learn new technology. Yeah, definitely. So now with the computing network,
- 14:54do I have it right that it's there's blockchain involved, but it's not fully
- 15:00the blockchain sort of there's Web two use cases and Web two technology as well
- 15:03as Web three. Yeah, so the way flu Fluence is a off chain
- 15:11compute network. And so the way to think about that is to take take
- 15:13a sent back and realize that the decentralization way created the blockchain. And you
- 15:18know, mentioned that Napster was the first decentralized application at scale. And this
- 15:24may surprise people, but you know, what do you think Napster is users
- 15:26were at peak, you know, millions, eighty eighty million, eighty million,
- 15:37right, And so here we're you know, clawing for thousands or tens
- 15:41of thousands of users or a million users for certain blockchains. Here Napster,
- 15:46ten years before Bitcoin had eighty million users at peak. So that shows product
- 15:50market fit for decentralization, and so that is you know, obviously was it
- 15:54was had other material issues. But the point is that our solution is off
- 16:03chain because there is ninety nine percent of the web is non determinists to compute,
- 16:10which means it does not it will not work via smart the smart contract.
- 16:14So smart contract compute is fully decentralized for sure on a thereium no argument
- 16:18there, and that has terrific use cases for primarily financial applications, great,
- 16:26but tons of other applications like you're not going to have an uber figure out
- 16:30the right route for you to take via smart contract. It's just not going
- 16:33to work, and so you need to have non determinis to compute. So
- 16:37fluence compute is off chain, like file coin storage is off chain, but
- 16:42has validated, verified, reward and incented on chain. And so we use
- 16:48blockchain where you know, to maximize its strengths. We don't figure in blockchain
- 16:53and force a use case into it. Now. I think that's a good
- 16:57way of going about it because sometimes you know, at the current state of
- 17:03where blockchain is, it's it's more efficient obviously, and there's a reason that
- 17:06you guys are just doing the verifying and part of it on chain, but
- 17:11you know, there's a reason that the Internet has been working so well for
- 17:12the last twenty years without blockchain as well. There's still a lot of good
- 17:15things about it, not just the Internet. I mean Internet for sure,
- 17:18but just computation is not all. It cannot all be done in a deterministic
- 17:25manner, and that is I think the most important piece of it. And
- 17:27so you know, you just kind of implied, you know, what use
- 17:30cases are out there, and so the other important thing about Influence is CPUs,
- 17:34which are central processing units, which are you know, again ninety eight
- 17:38percent or whatever it is of the web. When people talk about AI,
- 17:42they're all everyone talks about GPS, and you'll see a lot of GPU networks
- 17:48spring up of people that want to train artificial intelligence AI LM learning language models,
- 17:55and that is a huge piece and I'm we're very excited about that and
- 17:57the technology we're building up Fluence is relevant for that in the next phase.
- 18:02But we're focused on CPUs and CPUs are relevant in two ways to AI.
- 18:07One is that you need to clean the data before it gets into the model,
- 18:12and to clean the data you need CPUs. That's more the generalized compute,
- 18:18it's not the GPU side. And then your FLUDE is also relevant because
- 18:22you want to be able to prove that your model was trained on particular data
- 18:27set. And you've see in the New York Times suit open Ai claiming copyright
- 18:33infringement because they claim that open ai trained on New York Times. Maybe didn't,
- 18:37maybe didn't. I don't know, but I can promise you that if
- 18:41you're building an LM now, you'll be worried about copyright infringement and lawsuit risk.
- 18:45And so you're going to be able to provably validate the data or provably
- 18:51demonstrate what your model was trained on, and that influence is capable of that
- 18:55as well. And the last piece is then once you have model, I
- 19:00think it's important that models will run in a decentralized way so they're less subject
- 19:07to manipulation and to opaque manipulation. And so you've already seen chat GPT,
- 19:15you know, start to have you know, that's just a questionable answers.
- 19:19You know, can carbon may give me five reasons why carbon is good for
- 19:23the human race? You know, you may not get the answer you want.
- 19:26You know, there's there's a whole number of those what we'll say about
- 19:30I don't know in Pakistan versus India. You know, I don't know,
- 19:33But they're going to centralized companies are going to come under a lot of pressure
- 19:38to generate specific answers or forced the answers and models depending on where they are,
- 19:45and that is going to universally undermine credibility. And so a decentralized model
- 19:51will certainly have its failures. But if it's trained in open, transparent way,
- 19:55verifiably trained on certain data, then you can answers are a provably generated
- 20:02from that data set, you at least can't claim it was manipulate. And
- 20:07I think that will lead to those models will be the most trusted models eventually.
- 20:14And I give the obvious example here is people say, well, you
- 20:18know open Eye has you know, ten billion from Microsoft and this an enormous
- 20:22company, happened anyone compete with it? But you think about Linux had I
- 20:26don't know, Microsoft had a ten year head start, I'm not even sure
- 20:30and pretty much unlimited resources, and Linux just through the you know, just
- 20:34through the open global developer community, has become the default standard for operating systems.
- 20:41And so I think we have a model there to show what open source
- 20:45can do the power of it, and I think that's going to be applicable
- 20:48here with but it also has to be run on these open systems in order
- 20:53to help prevent manipulation down the road once it's complete. Yeah. No,
- 20:59that's a great and very important point, Tom, because people are starting to
- 21:03use chatchipt more and more every day, and you have to not just rely
- 21:08on the answer it gives as the absolute truth. You know. I feel
- 21:11like from the first version of chatchipt that came out, you could ask the
- 21:15same question to it now and the answer is probably different, whether they've updated
- 21:18their algorithm with new information or just new bias. It's hard to tell where
- 21:23the data sets are and what's the bias is, you know, whether it's
- 21:26about climate change, countries, political things. Yeah. So I feel like
- 21:33decentralized computing networks are definitely going to play a role in that. I mean,
- 21:38actually and just that point, as they become viewed more and more used,
- 21:44they become inevitably viewed as somewhat of the source of truth the way Google
- 21:48does, except instead of Google giving you links, is giving you an answer,
- 21:51right, So that is going to make governments in particular super focused on
- 21:56it. You ask the question, for example, imagine during COVID being should
- 22:00I wear a mask? You're telling me that the government isn't going to care
- 22:04what that answer is right, like and when when that answers right. So
- 22:07that's just one example, you know, whether mary mask makes sense or not,
- 22:11just as an example of the type of very specific question for which aren't
- 22:17links, there's not data, it's just an answer, and I think you
- 22:21will see that more and more focused on that going forward. Yeah, no,
- 22:25it's a it's a great point. So now with more of a decentralized
- 22:30LLM or open source of TAGIPT competitors, how does that work where they go
- 22:36to Fluence for example, as an open network they can host the application and
- 22:40the data sets public or how would that work a couple of different ways.
- 22:45So first the data. First, data can be clean influenced. They come
- 22:48to Influence, they ask us or they don't ask us, but they go
- 22:52on the to a protocol and you know, upload a lot of data and
- 22:56basically find hosts that are willing to or certain whatever price is clean that data
- 23:03and make it consistent. Then they would train that data, but that data
- 23:07training on the model would be a GPU based, but not on Fluence,
- 23:11at least not in this iteration of FLU. Influence will be good for training
- 23:15will be GPU, let's say compatible in version two, which is probably about
- 23:19a year issue away. I'm not sure exactly, but everything we're building is
- 23:23GPU compatible. We just started this before GPUs were as relevant. And then
- 23:30you can also then come back and say, okay, we had this model.
- 23:33Can you validate it was training the correct data and prove that and then
- 23:37ask hosts who will host this data and then run and then pay to run
- 23:47queries on this data. And so it's basically the running of the model.
- 23:53It's the computation the querying of that model that hosts on Fluence are student to
- 24:00do. And that is and that is the important part. And this is
- 24:03another nuance to what decentralization means. Decentralization doesn't mean I have five providers and
- 24:11I pick one, and there's some people in the ecosystem that do that,
- 24:15like I can do that with my web browser. What decentralization means is I
- 24:19pick or the algorithm, the network, or meat pick some provider. But
- 24:26if the provider goes down, the application automatically fails over to another provider.
- 24:33Without having to reprovision it. And so decentralization means effectively self healing. It
- 24:37means resilience, and that's the important piece here, because if you just have
- 24:44to go and keep reprovisioning, it's less usable, it's less user friendly,
- 24:48and it's you know, they're barriers to the to using and adopting it.
- 24:56Well, I'm very looking forward to trying some of those competitors, and you
- 25:00know, it's even chat Gibt is so new and it's growing so quickly.
- 25:06I feel like the decentralized competitors and hopefully more competition in general, will be
- 25:10able to be spun up quickly and compete, you know, headed head to
- 25:15head with some of the more centralized LM systems. Yeah, I mean I
- 25:21think that. I think that's going to happen. You know, it's it's
- 25:25all all in progress right now. So now about Fluence specifically, you mentioned
- 25:30there about version two. I want to roll that back to sort of the
- 25:33present day and what's coming up in the near future in the coming months.
- 25:37Here, I understand that the main net launch is coming soon. You talk
- 25:41about where exactly you're at and what does the main net launch mean and what
- 25:44will happen at that point. Yeah, main net launch we're scheduling for eighth
- 25:48Denver on February twenty seventh, So we're going to do you know IRL in
- 25:52real life event at eathh Denver. Very excited about it, and that means
- 25:57that we'll have a network usable, stable enough for people to use. It
- 26:03also means this is relevant that we will launch the Fluence token. And the
- 26:07Fluence token is critical because we have to pay people to provide their compute to
- 26:12the network. And so just like file coin offers fill rewards for the storage
- 26:18providers sps they call them, we provide the FLT reward to providers who offer
- 26:26their compute to the network, and then as more and more useful compute enters
- 26:34the network, that useful compute pays you know fiat I mean in stable coin,
- 26:41but you know denomina and FIAT prices to the providers who then take that
- 26:47revenue as opposed to the FLT reward. And so our view here is we
- 26:52want to make sure that providers are compensated fairly for their for their services and
- 27:00have downside protection. So we're pricing this in a in a FIAT way as
- 27:07well, but also benefit from the upside as this as it has the use
- 27:12of the network grows and as and as we hope the the amount of locked
- 27:18fluence increases significantly as well. Wow, very exciting, and that's right around
- 27:22the corner. What is the best way for people to follow along with with
- 27:26those updates and if they can't make it to eat down specifically, is there
- 27:30online community or the updates and how it can follow along with the main net
- 27:33launch? Yeah, I mean listen, I think first I would certainly go.
- 27:37You know, our website is is Fluence dot Network, Fluence Underscore Project
- 27:41is our Twitter, and I think that's probably the easiest way to follow us.
- 27:48We're also Fluence Network on Telegram as well, but I think that on
- 27:53on the Twitter is probably the fastest, easiest way to stay and stay in
- 27:59touch with us. Amazing. I'll put those links in the show notes below.
- 28:03Appreciate your insights into all of computing and blockchain, Tom, It's a
- 28:07very interesting discussion looking forward to the main net launch and we should definitely follow
- 28:11up in the near future after that. Great. Well, Thank you so
- 28:14much, really appreciate it and look forward to forward to coming back. Thank
- 28:17you again, Ashton