Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Forecasting the Future: Jordan Miller on AI, Blockchain, and the Power of Decentralized Prediction Systems
5 Plays3 months ago
What if artificial intelligence could predict the future with astounding accuracy? Join us for an insightful conversation with Jordan Miller, the innovative mind behind Moontree and the Satori network, as we delve into the intriguing world of distributed AI systems and blockchain technology. Jordan reveals the inspiration behind Satori, a revolutionary project aimed at predicting future events by harnessing the brain's natural forecasting abilities. Learn about the complexities and breakthroughs in developing a scalable proof of prediction algorithm and how blockchain technology ensures a decentralized and transparent network.
Explore the transformative potential of Satori, from forecasting viral videos to stock market trends and even sports betting. Jordan explains how Satori's comprehensive world model sets it apart from other AI services, providing more accurate and reliable predictions. Discover how the Satori token facilitates decentralized control and aligns economic incentives, making the system a vital tool across various industries. As we discuss future applications, you'll gain a deeper understanding of how Satori could impact everything from environmental monitoring to financial markets.
Lastly, we dive into the community-driven ethos of the Satori project. Jordan shares insights on the Satori Association’s nonprofit mission, its phased development plan, and the importance of community involvement in shaping the platform. Learn how you can join the Satori community, contribute to its growth, and stay informed about its progress through their website and social media channels. This episode is essential listening for anyone fascinated by the future of AI, blockchain, and the power of decentralized technology.
Recommended
Transcript
00:00:02
Speaker
All right. Testing, testing, one, two, three. Hello, everyone. Welcome back to the show.
Introduction and Background
00:00:09
Speaker
This is James Paris, and today we have a very special guest, Jordan Miller. How are you doing, Matt? I'm doing well. How are you? Okay. Great to hear from you. So can you kind of start this off by giving me, you know,
00:00:31
Speaker
of who you are, what you're about, what you're... Yeah, absolutely. So I'm the lead developer of a crypto startup called Moontree. And I'm the founder of a distributed AI, a crypto AI project called Satori.
00:00:57
Speaker
So Satori is a distributed AI. It's meant to distribute the production, benefit, and control of AI through future prediction. What the Satori network does is predict the future. So that's kind of the two things I'm involved with.
AI and Brain Predictive Nature
00:01:20
Speaker
Okay. And when did you first discover that
00:01:29
Speaker
Like first discover kind of the Satori kind of relationship, AI and crypto relationship there? Yes. Yeah. So I've always been interested in how the brain makes intelligence. And one of the major operating principles of the brain is that it always predicts the future of everything that it's going to receive, all the data it's going to receive.
00:01:57
Speaker
through the skin, the eyes, the ears, all of our senses, it's actually subconsciously predicting the future of what it thinks will come in. And this is so that it can give the rest of our brain, the rest of our intelligence, the ability to anticipate it and be ready for it when it comes. So time is a very important thing to the brain. It's always predicting this global time.
00:02:27
Speaker
But in a lot of AI systems out there like LLMs, all the latest stuff, they're mostly focused on not time, spatial patterns, spatial patterns within the language domain, spatial patterns within the images domain, stuff like that, so that they can translate one domain to another.
00:02:52
Speaker
So noticing that nobody's focusing on time and prediction, and also noticing that all things are correlated in time. So it's already just perfect for implementing on an actual network.
00:03:16
Speaker
That's when I realized we could implement this on a cryptocurrency, on a crypto network, and basically allow anybody to turn their computer into a future predicting AI Oracle. That's when I noticed it. It was probably about 2012, 2011 is when I was having these thoughts.
00:03:44
Speaker
And actually, at that time is when I first started learning about blockchain technology. So it was at that time that I was thinking, okay, let's make it a network of AI bots that are all predicting the future. They can help each other with those predictions.
00:04:03
Speaker
And as soon as I had these thoughts, I thought, well, I don't know how to make this network. And then I discovered blockchain and realized blockchain is a way that you can make an open distributed network. Anybody can join, anybody can participate. We're all doing proof of work so we know that we can trust one another.
Crypto Token and Network Control
00:04:23
Speaker
So I actually tried to implement this idea very naively back in 2013-ish.
00:04:32
Speaker
And I tried to implement it at the lowest layer of the distributed consensus stack, which is making a new consensus algorithm proof of prediction. And I wasn't able to get that proof of prediction algorithm to scale. So I kind of put it on the back burner. And about two years ago, I came back to it and said, now I know how to implement this correctly. And that's what Satori has become.
00:05:08
Speaker
So why crypto? Well, there's two reasons really. So first of all, you can make an open distributed network that anybody can join. And the second reason is you need to have some kind of unit of control that the people can have in a distributed manner in order to distribute the control of the AI.
00:05:35
Speaker
So you need a unit of control. And that just is the token that's minted by doing the work, you know, by making predictions. So as a computer is running the Satori software, it mints a token.
00:05:51
Speaker
That token is actually a utility token that says, I now have voting rights on what the network at large should look at, pay attention to, predict the future of. I can direct the attention of the network.
00:06:08
Speaker
And then that's how you keep it distributed. Because if you don't have that, then there's some small organization, some group of people that are deciding this is what the Satori network should predict. And you don't really want that. You want that managed by the community. So that's why you need kind of a crypto solution.
Blockchain and Decentralization
00:06:33
Speaker
So for novices and people who might not be too familiar with a lot of this lingo, what is the blockchain? The blockchain is a technology that allows us to come to consensus without having any centralized authority.
00:06:52
Speaker
So normally, I mean, money is the easiest thing to implement on a blockchain because you only have to keep track of one number. And that's why Bitcoin and then all of its competitors came about first. But you can use it for any kind of consensus, really.
00:07:11
Speaker
So, but to give an example of money, if you say, well, we're all going to transact with one another, so we create this centralized authority, and that's the bank. We say, okay, bank, you keep a ledger of, you know, everything we send to one another, and then we'll trust you, you know, we'll trust you. And the bank could lie. So the bank could say you didn't send money that you actually did, or whatever else. And so,
00:07:41
Speaker
You have to have this trusted authority. Well, decentralized distributed consensus, that allows you to all know the ledger without having a trusted authority that manages the ledger. The actual technology of blockchain that you're all running this protocol, that will manage the ledger for you guys.
00:08:04
Speaker
So that's kind of the high level reason or what blockchain does, why it should exist, like what it is.
Satori's Predictions and Future Plans
00:08:17
Speaker
What types of results can a consumer of this product maybe expect? And what would that consumer look like? The consumer of like the Satori product
00:08:31
Speaker
Satori, since it's an open distributed network and everything, it creates predictions and it broadcasts them for free. So at this stage, anyway, the Satori is producing free predictions. This is called our public good.
00:08:49
Speaker
So, a consumer of this would be anybody who wants to know the public good predictions, which would be things that are applicable to everybody, such as, you know, large scale government statistics, metrics on the environment, metrics on the economy, kind of the prices of the S&P 500 or something, you know, large scale stuff.
00:09:18
Speaker
So anybody who wants to know or get kind of a glimpse of the future of those things would probably be a consumer of the Satori output.
00:09:34
Speaker
As we scale, though, once we get to a point where it's predicting those large scale things relatively well, it can get more niche and more specific. So at some point, we'll be able to build a marketplace of prediction on top of this layer.
00:09:53
Speaker
that will say if you want the Satori network to de-silo your data and kind of bring it in line with everything else and see what it can learn and predict your particular data stream, maybe it's your quarterly sales next month or something.
00:10:10
Speaker
Well, it can bring it in, associate it with all the information that it's already predicting, and then it can provide that back to you for a fee. So that would be not a public good. That would be a private good. So I do see that happening eventually as it grows and evolves, but we're not building that right now. So I would say those are the two types of consumers of the whole project.
Competitors and Unique Position
00:10:43
Speaker
Just to say something interesting here, this is not to take a launch of you, but there's an old saying that if a product is free, then you are the problem. So when you're dealing with these types of consumers and they're using your service for free, is there any type of data they have to give away? What's sort of the trade-off of the free service?
00:11:08
Speaker
There is a trade off. So the trade off is this. The thing is printing a token, right? So we have crypto out there in the world and we have a bunch of projects and they all print their own token and basically they're trying to print money, right? So what gives them the right to print that money and to have purchasing power? I think if you're going to do that, you should be providing something for free.
00:11:38
Speaker
So this is what Satori provides, is future prediction. Do you think Satori may have competitors? Because there's a lot of AI services out there. You have blogs, you have Gemini, you have Chat GPT. Do you think any of these
00:12:02
Speaker
especially Chaching PT, a potential plugin that might be able to act as a competitor to this.
00:12:11
Speaker
Oh, sure, sure. So they already have made LLMs that they're trying to get to be able to predict the future or talk about the future. Right now, it's really bad at it. And the ones they've created, I don't know if they're any better. It's kind of like guessing. It's kind of like saying, well, I'm going to use the patterns I find in language itself to kind of maybe extrapolate a good answer for that question into the future. But it really
00:12:40
Speaker
Has a hard time with that particular domain the future and So they are trying to build these large language models that specialize in that however, I Don't see how they're going to continue to make it viable Really well without kind of an underlying ecosystem of future prediction
00:13:01
Speaker
So I do see kind of that technology having a place, but I see it as kind of an interface into a larger, kind of like an iceberg. The bottom of the iceberg is what Satori is. It's this huge massive network predicting the future of anything that people care about.
00:13:23
Speaker
And on the very top of above the water, maybe there's a little LLM, could be a big one, but it's aggregating all those predictions into a cohesive model that it can then talk to you about. You can then have a conversation about it. So that's kind of what I see as the right way to do it. Now,
00:13:45
Speaker
I do see that Satori would have kind of indirect competitors, but it already does because everybody's trying to predict the future of something. We all predict the future. Businesses, CEOs have to foresee the next five years. We're always trying to predict the future of our particular domain.
00:14:08
Speaker
And what we don't do is kind of bring all those predictions under one roof and kind of see the whole picture. And that's what Satori is. It kind of brings it all together. So we do have weathermen, you know, today that are predicting the future for everybody.
00:14:24
Speaker
but
Privacy and Data Integration
00:14:25
Speaker
only in one specific domain, you know? And so I don't see that being very profitable to a private industry. I do see it being the appropriate domain of a distributed system. So we don't have any direct competitors at all, but we do have kind of these kind of
00:14:53
Speaker
I could imagine Fitbit coming out and saying, we gather everybody's data, we throw it through models, and we're creating predictions of everybody's data. So if you want to know what your heart rate is going to be over the next 30 days, we're going to give that to you.
00:15:09
Speaker
And so I see these kind of specialized domains that are kind of economically incentivized because they're so specialized and they're, you know, this kind of data or whatever. I see that as, as growing in the future. I think we're going to have more and more of that as AI gets better.
00:15:29
Speaker
But I don't see any project or any company saying, why don't we make a world model? Because that's kind of a utility. It's kind of like a public domain kind of thing. I don't think they know how to make any money off it.
00:15:48
Speaker
So I do see those particular solutions though, they might want to plug into the world model, to Satori, because if all of a sudden your water supply is tainted or something, that's going to affect your health. And that's kind of what your biometric data is reflecting, your health. So they're probably going to want to be able to connect out to any kind of
00:16:17
Speaker
data, all things are correlated in time, but they're not going to want to have to manage the overhead of that. So I think they will plug into something like Satori or Satori for that solution. So I don't see any direct competitors for Satori at this, you know, on the horizon.
00:16:48
Speaker
Does this software specialize specifically in cryptocurrency or do you think this will eventually carry over to other things like maybe predicting viral videos or predicting the stock market or even something like sports betting or gambling? Yeah, I think a lot of people will want to use it for the beginning and for a long time.
00:17:16
Speaker
for direct economic incentive, gambling of various kinds or whatever. So I do see that being a large part of what Satori is predicting because that's what people will point it towards.
00:17:35
Speaker
We're going to try to get it to predict large scale things like government statistics. We're going to have a domain for that. But I do see people wanting to use it for that. So what we're going to do, we're in beta right now, so we don't have this feature built in.
00:17:53
Speaker
But what we're going to do is make it so that you can use the Satori software as a specific tool. If you want to just use it as a tool, you can do that. So right now you download it. It just goes out and starts grabbing data streams that the network is sanctioned and saying, OK, I'm just going to predict these random data streams.
00:18:14
Speaker
Well, if you wanna use it as a tool, you're gonna be able to say, no, I'm gonna point you to a specific data stream. You're not gonna broadcast the result out to everybody. You're just going to show it in the software locally. And so that way, you could point it to some sports betting or whatever you want to predict for yourself.
00:18:40
Speaker
you could also route your own data to it, like your own spending habits or your own biometric data so that it's private, you get private predictions, and it's only running on your machine, it's not running in some server. So we're gonna build that in so that it's a tool that people can use. But yeah, yeah, I don't know what people are going to want to use it for, but I'm sure a lot of it will be,
00:19:09
Speaker
predicting markets of some kind.
Beta Phase and Legal Preparations
00:19:15
Speaker
Now another thing I'm thinking about is what about nut jobs hypothetically? What if you have someone that's using this software to maybe predict something and they might have some type of addiction or they might be in some type of bad financial situation and the prediction software
00:19:40
Speaker
It might be accurate, but it might not be totally accurate. And they end up losing a lot of money, not really making as much money as they wanted. And then they try to sue you. Do you have some type of user license agreement or some type of legal protection for that?
00:19:57
Speaker
Yes. So while we're in beta, I mean, we make no guarantees, you know, I mean, the model's just doing the best it can do, but the world is crazy and chaotic. And so there's absolutely no guarantees that this software provides. So yeah, while we're in beta, we're getting all those legal ducks in a row. So we're going to launch July 1st and we'll have all that taken care of.
Defining Value in Predictions
00:20:29
Speaker
So what should we expect of the July launch? So if you download it right now, it just starts running. It's automatic. Everything's going to look pretty much the exact same.
00:20:49
Speaker
But the difference will be that it will be issuing the mainnet token. It will be issuing the token that gives you voting rights, voting power over the network, that utility token. So basically launch means
00:21:08
Speaker
We've been building this under beta. It's been issuing a token because we had to be able to test that functionality. But that token is just some test net token that doesn't matter. It doesn't have any abilities. It's useless, right? But once we switch it over, it'll be the main net token that is actually useful, has a purpose.
00:21:30
Speaker
So that's basically it. That's what launch means, that it's officially on mainnet and it's running in the global network. So the token is essentially how many commands you could put in? The token is an abstraction of utility.
00:21:59
Speaker
So when you say how many commands do you mean like how many data streams I Was thinking in a much more practical sense like credits Like how some of these free softwares they'll say you have 30 credits. These are the amount of credits you could use per day Is that what the token is? What is the token? you can think of it as credits, but it's a credit that I
00:22:24
Speaker
There's really no spending the credits. So if you download the software and you want to do some special thing, at this point, maybe we will in the future, there's nothing you can spend the token on. What you can do, though, is you can say, I have 100 tokens and I want to put them all on a particular vote. So you're staking it. You can't move it. It's going to stay there until you're ready to take it away. And then you're not voting on that anymore.
00:22:51
Speaker
So you could say, well, I'm really interested in the environment or something. So I'm going to put all of my tokens on predicting sea temperature changes or something like that. So and then as long as that money is just sitting there on that button,
00:23:09
Speaker
it's not going to you know you're not losing it right but it's stuck there like it's doing its work of voting for this is what you think the network should pay attention to and then as soon as you're like well um i want to throw this stuff away or whatever i want to get rid of it then you can take it off of that button
00:23:31
Speaker
no longer are you voting for sea temperature changes and if nobody else is, then the network will not predict sea temperature changes. So that's kind of the point of the token, is to say this is where the network should look. This is what it should look at, this is what it should pay attention to, this is what it should understand and this is what it should predict. Right now,
00:23:57
Speaker
That's the only place that the human is in the training loop with this system. In LLMs, we create the language and we train it on our language and we curate the language, we throw out hate speech, we do whatever we want to, the training dataset. And then when it's not trained exactly the way we want it, we retrain it correctly.
00:24:21
Speaker
And so every step of the way on these other AI systems, the human is in the loop, the training loop, and is implementing our bias on every step. And so these other systems, like LLMs, they can't approximate truth, because what they're actually approximating is our bias, our interpretation of the truth.
00:24:48
Speaker
Which is fine. I mean, that's what language is. We got to do that. But if you have raw data coming into the system, like the future of some numeric metric that was measured, you have the raw data coming in. You're predicting the raw data automatically. There's no human in the loop, the training. There's no human in the training loop there.
00:25:12
Speaker
But we do have to direct its attention. So we don't implement our bias except in the domain of saying, well, we don't care about how many termites there are on the Earth. What we care about is the price of Google or something. So we direct it according to what we find valuable. But then it gives us the truth of that domain. It tells us.
00:25:37
Speaker
Well, according to just the math, this is what's going to happen tomorrow. So I see Satori as a better approximator of truth than any of these other AI systems out there. And this might be a bit of a deep question, but how does your company define
00:26:05
Speaker
value when it comes to what the actual software focuses on? Value is basically value is kind of like the dollar is an abstraction of value. So the dollar is only valuable because somebody else wants it.
00:26:26
Speaker
And if somebody else wants it, then I can get them to move some product for me in order to get this dollar. So I can get other, it's an approximation of other humans' labor.
00:26:41
Speaker
So that's of what value is. I can move something in the universe by giving this dollar away. Whether that's moving product literally from the back room to the shelf or selling or whatever, I can move something. I can make a change in the world. It might be, hey, I can change the governments and senators and congressmen, sell their
00:27:11
Speaker
and influencers sell their influence, right? They can change people's minds. They can implement a change in reality. So that's what value is. It's a change that we would like to see or someone would like to see, somebody who has money. So it seems to me like what we're creating here with building up this AI expertise
00:27:39
Speaker
is we're building the ability to understand the world. And with that, we're building like computerized labor, right? Or a particular form of computerized labor predicting the future. And so it seems like value is a share of labor, whether it be human labor, computerized labor,
00:28:03
Speaker
animal labor, you know, it doesn't matter. It's a share, it's a value is a share of labor. So that's how I kind of see value. Now, what people will want it to predict
00:28:19
Speaker
is dependent on how they think they can use it, right? Maybe they're just, hey, they just like the environment. They think it's good for humanity to kind of understand the environment. So let's predict the environment, right? So there's that value, but there's also the value of saying, well, I can get rich on sports betting. So why don't we predict sports betting? And so we're kind of letting all the flowers flourish. Like whatever people find valuable is what the system will predict.
00:28:49
Speaker
Now there's only so much bandwidth. There's only so much compute power. So those, what are they called? Those competing interests have to compete. So yeah, so I think they will. And whatever people want the most will rise to the top and get predicted.
00:29:15
Speaker
Did that kind of answer your question about what value is or what we think it is?
Growth and Community Involvement
00:29:21
Speaker
Correct, it does. From what I understand, the value system is going to be based entirely off really what the consumer and the market plans. Yeah. Yeah. It makes sense. You know, it is a business.
00:29:40
Speaker
Right. And so there's going to be a Satori association. In fact, there already is. We're incorporating. But the Satori association is mining, is running nodes, is doing things. And so it will have some token. It won't have most of the token, but it will have some. And its mandate is to build that world model. So all of the data streams that it predicts
00:30:10
Speaker
or wants to predict all the data streams that votes for are going to be whatever is not voted for by the community that will help inform a world model. So I don't think the community cares too much about demographic shifts. So I'm sure the Satori Association will have to put its voting power towards those kinds of things.
00:30:42
Speaker
This is all quite interesting. And what I'm thinking of now is, this is a lot for a free piece of software. So what do you think the cost is going to be? Because I'm familiar with chat GPT, you know, they have a premium service and then sometimes what I'll do is I'll pay for the full workspace service just because it has more power, more storage, even though I'm the only person using it.
00:31:11
Speaker
There's a lot it could offer and it's worth the payment. This seems like something that's worth a monthly subscription. And I'm saying this too, not to just help your company, but I feel like as a consumer, if I'm paying money for a service, I feel like I'm paying for additional privacy, security, more control on my part too, if that makes sense. Absolutely.
00:31:36
Speaker
We have not considered that kind of stuff yet. So there is a lot of ancillary services or stuff like that that we could implement and actually charge for.
00:31:53
Speaker
because we are able to mint the token itself and because the Satori Association's mandate is to help that ecosystem flourish rather than make a profit. It's a nonprofit organization. We haven't explored those realms.
00:32:12
Speaker
So I think these add-ons will eventually get to, which will be good because then we'll just be able to build it even bigger. But at this time, really, it's a distributed system. So we don't have to pay for too many servers. We don't have a lot of overhead.
00:32:33
Speaker
We don't have to buy huge AI computers because we're distributing this onto anybody who wants to run it. So being able to kind of distribute that load means we can do things very leanly and we don't really need too much.
00:32:53
Speaker
We do need a little bit, and that's why the system has a dev fund that automatically gets minted. But the community is in charge of the percentage of that dev fund, in fact. So if the community thinks that the Satori Association and the devs are not doing a good job at improving the system or whatever else,
00:33:17
Speaker
then they can just vote that down to zero and say, well, you guys aren't doing a good job. You don't deserve any money, right? So the community is kind of the board. The community of token holders, that's how I kind of view them. They're in charge.
00:33:33
Speaker
They get to decide where the money goes. They get to decide where the attention of the network goes. And any more fine-tuned stuff than that, well, that's the devs or that's the system itself. But they're kind of the overarching boss of the entire thing. So I guess that's how we're initially getting any kind of funding is the dev fund.
00:34:04
Speaker
But beyond that, like if we ever need to, we can add these ancillary services that we could probably charge for. Interesting. So what do you think is sort of the future of all of this? What do you think is kind of the next step after
00:34:31
Speaker
July. What are your expectations for your business and your brand? We've mapped this out very broadly in basically four phases. So we're in phase zero right now, which is development. So we've been developing it for two years. We're still in beta and we'll exit beta soon. So at that point, phase one will begin.
00:34:57
Speaker
And in phase one, the point is just scale it. The more people join, or the more computers, the more compute power we have, the more things we can look at, the more models we can associate with each other, and the better all the predictions become. So scaling is the first step.
00:35:18
Speaker
The second step is kind of a marketplace and saying, I kind of mentioned it earlier, the saying, okay, now that we've grown this computerized labor force, we can bring it to market on specific privatized prediction kind of platform.
00:35:38
Speaker
So that's phase two. So the first phase is scaling so that we can get this world model, this public good. The second phase is add on the marketplace so that people can get their own data predicted. And the third phase is kind of out into perpetuity, continue to try to make this system as decentralized as possible.
00:36:02
Speaker
So that's part of the mandate of the Satori Association is to keep the system as decentralized as possible. So that's why we've built it in from the beginning that the community controls how much the Satori Association can take, how much the dev fee is. And we've tried to build in controls from the beginning, but it will be a continual effort
00:36:29
Speaker
that will never end. So you always have to try to decentralize something that's going to naturally become centralized. So those are the three phases that we can see. They're very broad strokes. When you expect to see
00:37:01
Speaker
in your community exactly in the future. Did you say that? To sort of evolve and kind of adapt with the system. Like how do you expect consumers to sort of use this product and how do you expect them to maybe evolve or change as time goes on? Do you want to sort of create more so just a bunch of regular users or do you want to make something more similar to like
00:37:30
Speaker
participatory culture where the community is a lot like chat GPT where it builds a market, it builds plugins, it begins to build up the community on its own in this more open source freelance way rather than you just grow the company and the community just consumes and uses what you plan for.
00:37:53
Speaker
We want to make it as open source and community driven as possible. So it's a community project. I mean, I guess you could call it that technically. It's not sponsored by any corporation, right? It's that kind of thing. So it's a community project and
00:38:11
Speaker
So we're going to need the community to continue to help us, right? So at this point we have a very small community and they have already helped us with like beta testing, making sure all the features work appropriately, that kind of stuff.
00:38:26
Speaker
But as we go forward, we're going to want more things. So we're going to want people to be able to write their own AI algorithms and then share them across nodes or in the marketplace so that people can experiment.
00:38:44
Speaker
We wanted to go automate first. So we want this to be able to be downloaded and anybody can just install it, run it, and then walk away, right? So we wanted to do that first so that we could make it easier to scale.
00:39:05
Speaker
right they download it they run it they walk away this is this is crypto mining right and so since it's a crypto project we felt the first people that would get involved are crypto people and then maybe ai people after that but the first people will be crypto people and they just want to run it on their machine and walk away those are miners so
00:39:29
Speaker
We built it that way first, and we want to build kind of, we want to build in the ability to modify things, put in your own AI algorithms, stuff like that, second, and make it a tool, a tool that people can use with their own data, that kind of stuff, second. And so we do need the community to be there,
00:40:00
Speaker
and to be kind of helping develop these kind of things. But we'll just have to see how that plays out. That's interesting. So you know, at first when I was looking at your company, I was seeing it as more of a Microsoft, but it really is more of a Linux. If you think about the overall theoretical kind
00:40:29
Speaker
look at the entire thing.
Privacy in Data Streams
00:40:31
Speaker
But, you know, I think the next question I need to ask is, what about the privacy? You know, what type of data do you think will go around, like, let's say, potentially, right?
00:40:49
Speaker
I don't know how this software works, but what if I have a Kraken account and I put my Bitcoin wallet ID into the software? Do you now have access to it? Does anyone else now have access to it? Where does that information go exactly? Because it's free. And obviously I'm going to assume because it's free and it's still new, you're going to be using a lot of this to collect data.
00:41:16
Speaker
So the data is interesting. So it's a series of data streams. Right now, any data stream you publish will just be open and public. And so not only could the Satori Association collect that data, but anybody could collect that data. They can subscribe to that data.
00:41:45
Speaker
At the beginning, that's where we're at. As we go forward, though, we'll want to be able to say, OK, now this data is private. I have the Satori software running on my machine. And so I'm going to route this directly to my instance of my software.
00:42:04
Speaker
And if I do it in that way, I can specify this doesn't get broadcast out. It not only doesn't get collected by the Satori association, it can't because it never leaves my machine.
00:42:19
Speaker
So we're gonna want to build that pretty soon as part of that free tool vision of saying, anybody can download this, have their own future predictor under their desk, and all the data is local. One thing that, since we're doing future prediction, one thing that I think is pretty cool is we don't have to share the models.
00:42:45
Speaker
So we share the output of the models, the predictions, but we don't actually share the models. So if your computer's churning on some, you know... Could you clarify what a model is? Is that a prompt or is a... A model would be like the chat GPT-4. So that's a particular kind of model. Okay.
00:43:05
Speaker
the actual LLM. So your computer, as it's running the software, it's listening to some data streams and it's trying to predict the future. Well, the way it predicts the future is to say, I'm going to listen to the data stream, build a model of what this thing is, recognize all the patterns inside of it, recognize how it relates to other data streams, the patterns across data streams. And it's going to build this model for each data stream
00:43:34
Speaker
And it's going to say, this is on my computer. This is mine. Nobody else has control of this. I don't send this little machine, essentially. I don't send this up to any servers. This is my little asset that I'm developing as my computer listens to more data.
00:43:56
Speaker
So, and it's always doing that. It's churning 24 hours a day. It's looking, basically searching the model space for what would be a better model of this data. So, and then that lives on your machine. Then when it sees a new piece of data come in, the very first thing it does is say, how would my model have predicted this? What's my accuracy?
00:44:22
Speaker
And then it runs that data along with the history. It runs that data through the model and gets a prediction. So predictions are the output of the little model engine. And then that prediction is what gets broadcast out, right? So you're never giving away the cow, you're giving away the milk.
00:44:43
Speaker
So you get that prediction out and you broadcast it, but the model stays on your machine. So in our system, unlike anything I've ever seen, models are private. They belong to the person running the software. And if that person turns off their computer, the network no longer gets those predictions. It no longer can use that model because
00:45:09
Speaker
it's not shared ever. So as far as model privacy goes, I think that's kind of cool. They could be shared, but the honest truth is they are getting recreated all the time. So you might, you know, you might start listening to a data stream and you start looking for models or you start building models and your model
00:45:35
Speaker
Very quickly, within an hour is superseded by a better model. Your computer found a better one. So we throw that old one away. We replace it with the best known model so far. And then we continue to try to make a better one. And so what you have is that at the beginning stages,
00:45:56
Speaker
you get a new model very quickly often. And then the longer you're watching this data stream and building these models, the more it kind of slows down. Like it'll take you a whole week to find a better model than the one you have currently. And it'll take you a whole month. And so the best known model, um, kind of the longer it's been living is the longer it lives. So,
00:46:24
Speaker
As far as data privacy goes, we want to route that to your machine so it can be private. We're going to do that soon, but the models are already private. They already just live on your machine and they don't get shared. This is all quite interesting because I feel like that would be something that would put a lot of strain on the
00:46:47
Speaker
computer itself in the computer memory, but at the end of the day, if you're a Bitcoin miner, you should already have the hardware that could be able to handle something like that. So it makes sense. I don't think that's an issue, but this doesn't really seem like a product that the regular consumer would want to use. This sounds like a very specialized tool that
00:47:15
Speaker
high-end Bitcoin miners based around Australia, China, those types of areas. You know what I mean? Well, it seems like that. But let me make a distinction between Bitcoin mining and intelligence mining.
Intelligence Mining vs Bitcoin Mining
00:47:31
Speaker
So blockchain, the way in which the blockchain is created and everything is every computer is looking for a very random number, right? And there's no way to figure out what this number is going to be. So this is the proof of work algorithm. It says, there is a number out there that matches this kind of data from the last block.
00:48:00
Speaker
it matches, but nobody knows what it is. And so what we have to do is we have to randomly create a number, see if it matches, and if it didn't, we throw it away and we try again. So what they're doing is hashing, right? And so that's what hashing is. So they're randomly creating these and it's guess and check, guess and check, guess and check, guess and check. And so
00:48:23
Speaker
What that means is they can all use the same algorithm because there is one most optimized way to guess, to create this random number and then see if it was correct. So what they can do is they can
00:48:39
Speaker
not only put that, um, they can program it in a CPU course, right? But they could also program it on something a little more specialized like a GPU. So now they can do it in parallel and it's a lot faster. Well, they can even do it on a six, which are like the hardware itself can do nothing other than that algorithm. And so it's the fastest at doing that algorithm. Cause that's the only thing it ever has to do. So,
00:49:07
Speaker
My point is, if you are doing guess and check hashing proof of work, it centralizes very quickly because it commoditizes extremely fast. Now with intelligence mining, there are thousands of different algorithms you can use to build a model. And they all have their pros and cons of how much
00:49:37
Speaker
cost it takes, how much RAM it takes, how much CPU power it takes. So they're all different. And they all have their pros and cons according to, well, this data stream has a lot of entropy. So you need to use something more like this. This data stream has little entropy. It's easier to predict. So you can use these kind of models. So what I see happening, we have this engine. The software is basically a wrapper around this engine.
00:50:05
Speaker
The AI engine, its job is to say, what kind of hardware am I working on? Well, okay, it's a laptop. This is how much hardware I have to work with. I only can use these algorithms, but they make models. They're not as detailed as models I could build on some massive GPU setup, but they are actually models that have some predictive power.
00:50:34
Speaker
So I don't see this intelligence mining as commoditizing nearly as fast.
00:50:40
Speaker
And so you can run it on a laptop and have it make valuable models, even if it's not as valuable as the ones that a GPU miner could make. And you can also pause the engine. You can say, hey, look, I want you to run all the time because you're a tutorial and you're building models and that's great, but I need to use this computer for the next eight hours.
00:51:06
Speaker
So I want you to pause, not go heavy on my CPU or whatever. And then you can resume, right? And so that way we can have this software running all night for people. We tried to make it as easy to run as possible on any hardware. It's for Mac, Windows, Linux, whatever.
00:51:35
Speaker
Do you think Bitcoin mining as a whole is moving forward?
Future of Intelligence Mining
00:51:40
Speaker
Do you think it's this thing like intelligence mining? Do you think it will completely replace maybe the old school methods of mining where you need loads of hardware, all these fans and all these small computers digging and digging profusely for, you know, It's possible.
00:52:02
Speaker
It's possible. Like I said, though, at the very beginning, I tried to implement this in the consensus algorithm.
00:52:11
Speaker
way back in the day and was unable to make the match. So until somebody can make the match at that level, we still have like in Satori, we have a blockchain, which is that hashing algorithm. And on top of that, we have layered the algorithm for intelligence mining.
00:52:33
Speaker
So we have a dual system. I would love to have been able to make it as just one thing, but it's more practical at the stage of the technologies right now to make it in two layers. So if that ever can be combined, condensed down to one layer, yes, I do think intelligence mining will just take everything over. But until then, we have to keep it separate.
00:53:05
Speaker
excellent this has been a really good and you know informative interview and you know i definitely learned a lot and i'm assuming the audience will also learn a lot too um
Conclusion and Community Engagement
00:53:17
Speaker
Thank you again for being on the show. And just to close this off, are there any final words you'd like to say to the audience, suggestions on, you know, again, you have your website right there, anything else you would like to let the audience know, social media, et cetera? Sure. Well, thank you for having me on. And these questions have been very thoughtful and really good.
00:53:44
Speaker
Satorinet.io is the website. You can go download it now. If you just want to put in your email to get notified when it's up and running as a mainnet, you can just put that in. We also have social links on the join tab for Twitter and Discord. Our community is in kind of Discord, so if you want to actually get involved programming and anything like that,
00:54:10
Speaker
You can get on Discord and just talk to us directly. All right. Well, thank you again for watching the show. I will be seeing you all next. Thank you.