Perceptron Network cuts AI training data costs by 92%

0:00I'm Ashen Edison from The Crypto Coin Show, and today on Blockchain Interviews with Peter Anthony, co-founder of Perceptron Network, a high-quality AI training data, which can get information at a fraction of the cost of what centralized providers charge. And also, you're having that ability to earn from data that you're contributing and running through a decentralized network

0:22also, which is super exciting. I'm excited to dive into this, Peter. Data is so important with the growth of AI. The biggest companies in the world are paying lots of money to get high-quality data, and people that contribute that data should be able to earn from it. So, thank you so much for taking the time to come on the show.

Thank you for having me. I'm looking

0:44forward to talking about data in the world of AI.

Definitely. It's always nice when you can earn, especially if you're sitting on AI all day. It, you know, back in before Web3, you know, web people were getting sort of tired of Web2 and being like, well, you know, I sort of feel like I am the product. I feel like that's more so the case even with AI on these centralized

1:05platforms. Everything you type and respond to is used to make their platforms better, and you have to pay them for it. You're not really earning anything.

Absolutely. I mean, every single minute of every day that you and I or everyone out there who's using a mobile phone or a laptop, we're contributing about 1.3 megabytes per second of information.

1:29Now, all of the data that's in existence right now has already been scraped by the existing AI models. So, OpenAI, Claude, Anthropic, et cetera. They've all been able to gather the existing data. That's something like 750 trillion tokens of data. Token is effectively 3/4 of a word. Now, that's going through all the forums, going through all of this on X,

1:53going through all the things on Quora and Reddit, whatever else. And that gives you obviously a portion of the internet, portion of information. But that's only a fraction of really the kind of information that's available. The majority of information that is useful for training the future of AI is what you and I are doing on a daily basis, how we're interacting with

2:14Claude, how we're interacting with Twitter, whatever that might be. Now yes, we're contributing or yes, we're producing data. A lot of that data is just going to waste. What is missing is perhaps being able to gather that data and make use of it and reward the people who are contributing that data.

Definitely. I saw that Google and OpenAI had paid multiple

2:37hundreds of millions to Reddit and to get access to some of these parts of the internet where people are contributing their own information. I'm guessing none of that actually went to the users of Reddit. It'd be nice to have like a little payout, but I feel like that's something that needs to be fixed and decentralized networks should have that

2:56capability of contribute contribution and then earning from it. So is that how Perceptron network is laid out?

Yeah, so at the moment, you and I when we when we're contributing data that's being used from Google or whoever else, obviously they're packaging that data up and selling it on for millions if not hundreds of millions of dollars to other

3:17companies. Now what Perceptron network allows you as an individual user to do is to be able to run a node. Now the node can be run on a computer or on a Android device and this is effectively you sharing your excess bandwidth. It's giving us a vantage point of the internet. It's not us gathering your data, but it's us being able to look at your

3:41your access point of the internet, your IP address to be able to gather a package of data. Now we are able to do this across our network. We have around 800,000 downloaded nodes. Each and every one of those nodes gives us advantage point where we can gather pieces of packets of data and build them into meaningful data sets. And by running a node, you will earn

4:05you'll be able to earn points. Points in the future will will translate across to Centrom tokens. In addition to that, when you gather the data, you know, raw data as as it is isn't always as valuable as it can be, but you can data sets. You can annotate different data pieces, etc. which kind of upscales the value. Kind of like if you think about milk. You know, milk's very cheap, but

4:33if you go and buy cheese and ice cream and whatever else, it's the same product, but it's kind of made more expensive by kind of adding annotations and creating more value in the data, you can create more value from the end current. What an individual user in our network can do is they can contribute not only by running a node to contribute to the bandwidth, but they're also able

4:55to participate in things like annotation tasks as well as contributing to data sets. So, perhaps we need pictures of certain types of people. Perhaps we need to have images of certain places in certain cities of the world, whatever it might be. You can then go out in your city, you can take photos, you can contribute to data sets, and ultimately you can then earn rewards

5:19for these different data tasks.

That's very cool. And so, from what I understand, there's sort of a passive part and then an active part. You know, if you are a busy person, I think everybody has excess bandwidth on their phone or the computer they're not using especially at night time unless you're running AI agents 24/7 at maximum capacity, maybe the 0.1%, but you can

5:44run a node and not have to actually be actively contributing the valuable part of data. It still provides information with earning potential, but then if you go full active mode and you actually contribute the exact type of data that these companies or perception network is looking for, then you boost your earning potential. Is that Is that right?

6:08

Yeah, effectively effectively that it's you know, like like anything in the world the more difficult it is to deliver a particular thing obviously you know, the value of that becomes more becomes higher, but there is obviously the passive kind of contribution that people can do by running by running nodes. And effectively what we do is we take data. We then sell it on to clients.

6:29Obviously clients who are building things like AI models. We also gather gather kind of like live data sources for sentiment analysis particularly in crypto that's very useful thing can then be sold on to exchanges, quant desks, etc. And this then you know, creates revenue. Revenue then goes back into gets distributed amongst the contributors, but also kind of goes to

6:54developing future products and building up the building up the business as a whole to you know, onboard more clients, you know, get the whole ball rolling.

Definitely. Well, it sounds enticing especially with this job market. People are always looking for side hustles or you know, extra income, but of course there's probably a lot of variables in

7:16how much you can actually earn based on the passive methods and then actually how what kind of data you're contributing that's needed. How does that mechanism work in determining how much potentially you can earn from contributing the data?

Good question. At the moment as we move towards completion of our data questing app, there's going to be certain data sets

7:42that are obviously much easier to contribute to because you know if we are English speaking for example we need something written or sorry read out in English it's very easy task lots of people are going to be able to do it a lower value task. Let's say however that we need a rare language from somewhere in sub-Saharan Africa and maybe we've only got five or 10 users on

8:04our network who are able to actually read out that particular passage. Rare example but that would obviously command perhaps a slightly higher reward for doing such a such a process. The mechanism via which you know points and things are gathered is obviously a complex one and obviously it's quite heavily regulated by AI but we've just acquired

8:34The a new company or sorry well an older company in crypto called router and this company has effectively built a framework that allows us to identify each and every user and how they've been contributing to the network. So it's effectively done automatically no one's really sort of sitting down working out individual contributions.

Wow that's really cool and

8:56I'm sure there'd be some

[clears throat]

those few sub-Saharan Africans would be more than willing to contribute some of that rare information. I would love to dive into the app and see what are the things right now that my expertise can can fill in those gaps. So is that with the earning potential you said right now you can earn points and then

9:19later that moves into tokens. So can people go to the website to start or is it Android and iPhone applications?

So the best thing to do right now is if you go across to Twitter in fact actually if you go to the website which is Spectron's with an S dot XYZ. Over there, you're going to be able to download the application either for a Chrome browser, like the one I'm

9:44using, or for Android device, iOS device, etc. So, once you've kind of downloaded it, effectively right now, because we are pre-TGE, you're going to be earning points. In future, once the token has become live, those points can can translate to being worth tokens. And obviously, as the network grows and as the network brings on more paying clients, that gets distributed

10:10throughout the network.

Mhm. Yeah, it always pays to be early on, you know, the next big thing. So, I can leave those links in the show notes below. Some of the points that I was looking through on perceptrons, it was saying that, you know, the cost is much lower to acquire this information. And I'm curious, how does that economics work versus centralized AI collectors? You

10:37know, they paid 200 million for Reddit information. Would they pay an equal amount for perceptron data?

So, right now, like effectively one of the major problems in AI is the monopolization of access to data, compute power, etc. And this is predominantly due to costs. But one of the major highlights or one of the major things that's

11:05highlighted the actual problems of monopolization happened the other day. US based Anthropic, which is where you run Claude, etc. were allowing people to use this new model called Fable 5. And then all of a sudden, it shut down access to Fable 5 for everyone outside of the United States. Now, the That's great if you're in America, no problem, okay? But if you're outside

11:30the United States, it highlights that there's a company out there that's heavily relied upon by a lot of different people that suddenly stopped giving access to people who might need it. That can have a really big impact in the future because it dictates these people are going to be allowed to do building and whatever else, but these people are not going to

11:49be able to. That's the huge problem with monopolization.

Mhm.

Another problem with monopolization when it comes to data is the access to data and how you do it. Now, traditional methods to being able to gather data has come through access to APIs for Twitter, for Reddit, etc. Now, obviously, the data is extremely valuable. And because it's extremely valuable, if

12:13you own those companies, you can charge huge amounts of money for it. Now, the problem with doing that is if you charge huge amounts of money, only OpenAI and Anthropic and a handful of others are going to be able to afford to pay those data costs. If you are a new upcoming AI project, you're never going to be able to access that quality of data,

12:31therefore your products are not going to be very good. You could build the best product in the world, but it doesn't have access to the data, it's not really very useful. Okay, so it's like you could be the smartest person ever, but if you've never read a book, you're not going to know much, right?

Yeah.

so what we do that's different is that

12:48we look through the eyes of an individual user and we do that hundreds of thousands of times, which enables us to gather the data and information that would be shown through APIs. Now, this is something that recently went through court and has been identified as legal in that it is ident- accessing the internet without any having to go through paywalls, etc. So,

13:17if right now you're able to access Twitter or Reddit from where you are in the world, then I can also access Twitter or Reddit from where you are in the world and I can do that hundreds of thousands of times over. To be able to gather the same meaningful data sets that any of these other companies can do. And I can do it for a fraction of the cost because I can do it I can

13:36gather the data for effectively free. And then obviously, you know, we make we need money to build network, to run the network, to distribute to users, etc.

Mhm. Mhm.

So, we're breaking down the monopoly effectively by being decentralized.

Yeah. No, that's one of the most important factors of decentralization is the global access and

14:00equality for everyone in the world. You know, one thing I'm I'm curious about with regards to the data quality, you know, of quality of data is super important, probably much more so than than quantity. If there's all of these data questing tasks in Perceptron, and I go in there and I'm thinking in my head, I'm just going to do all of this as fast as I can. I don't really care about the

14:24quality myself. As long as I put stuff in there, get paid, and get out. How does the network, you know, and the AI mechanisms behind that determine the quality of data and ensuring that you're actually getting paid for putting in quality information and not just junk?

Good question. So, that and actually one that I was kind of talking to

14:46someone about today. First of all, we've been looking at this with regards to the data questing platform and how to effectively identify, you know, serious contributor versus a non-serious contributor. Non-serious contributors who are perhaps are not, you know, who are trying to kind of do every different task that are not necessarily doing it properly, and they're quite

15:09clearly not able to do the task that they want, will obviously have some sort of a penalty for not contributing data effectively. But also, we have right now and what we So, for example, when we scrape, let's say someone wants to go and get us Someone's requested that we want to have 500,000 pictures of cats or 500,000 images of whatever it is. We've been

15:36asked for some some strange things. Now, we obviously individually are not going to go out and find those items because it would take a very long time. But, what we'll do is we'll build an understanding through AI of what a cat looks like in this scenario. Then, they will have The AI will train on this particular information. Once it's trained on the information, it is able

15:57to gather that information, it will then go and gather huge inventories images of cats. Now, once that image thing gets fed back through, there will be certain criteria that the AI will then identify within that information and what it's looking for in terms of the annotation, which means that if certain items that have been annotated are not in line with what

16:22the AI is expecting, effectively they will be rejected. If it's rejected, effectively that then will be rejected for, you know, one of two reasons. It's rejected because it's not been fully done properly, but you can see that they've made an effort. This is something we need to kind of figure out. Or it's been done, you know, with no intent to kind of do

16:43it correctly, in which case then penalties should apply. So, people shouldn't be in fear of contributing data sets because they might get it wrong. As long as they kind of, you know, making efforts to do it correctly. But, there should be penalties [clears throat] in place to kind of ensure that people are doing it properly.

Definitely. I saw that Perceptron also just launched

17:05a $10 million AI data fund to fund different teams and startups for AI data. This is super interesting. How how Can you explain the data fund and the goal behind it and how it ties into the greater perceptron network?

Absolutely. So I've been in crypto for a long time now.

17:31Originally I was the founder of the House of Crypto, which is a YouTube channel. And over kind of, you know, my time in crypto, I've watched a lot of different projects come and go. And really what kind of led me towards building Perceptron was to identify that actually I really like crypto and blockchain technology and what it can really offer the world, but

17:53it's very rare to kind of think of a way that in which blockchain isn't an essential component of a particular aspect and can actually improve the way that a service or business works. I think when it comes to AI, you have a few different ways in which blockchain technology is the quintessential technology to be able to kind of push things forward, right?

18:19And because of that, I want to kind of showcase blockchain technology and what it's able to achieve. And by by building this particular business, I've been able to kind of gather, you know, to solve a real issue, shall we say? And what better way to showcase the issue being solved by giving the opportunity to a lot of crypto sorry, AI projects

18:45to come and use our quality data basically from that. So, eventually right now, it's to one give shine the light, shall we say, on blockchain technology and why, you know, it can be the future and actually decentralization works far better than centralization in one way, like this. And then also to showcase sorry, to enable AI projects to

19:12build out into the future cuz obviously they will be future paying customers, etc. And yeah, so sort of move forward, give something back to the community, etc.

That's very cool. What's the timing on that data fund? And if there are teams that are looking to get funding or, you know, get their startup out there into and tie into Perceptron,

19:38what's the best way to get started on that?

So, basically, right now, we are open to taking on boards for sorry. Right now, we're open to onboarding AI customers who are looking to kind of make the most of that fund. Projects that are working in the crypto industry, perhaps in Web2 as well, who are looking to gather data in order to use for building

20:04out their models, etc. Those companies probably best if they reach out through X, our Twitter account. That Twitter account can be found through the website, which is I think you're probably going to link it down below. Definitely. Down [clears throat] below here, maybe the Twitter page as well. So, best to kind of give us a message on Twitter. And then from

20:22there, we can connect you through the team, discuss the kind of requirements and needs that you're looking for, and then kind of kind of process through that.

Awesome. I'd love to you know, we don't have a lot of time left, but I'd love to hear a little bit of the long-term vision and then the short-term heading towards the token launch. So, in the

20:42next 12 to 24 months, how do you hope and expect that Perceptron will grow sort of be a fierce competitor in a AI data space?

Good question. So, our main kind of goal right now is to build out this kind of data questing platform. And also to move more towards live business intelligence. Now, right now there are existing companies

21:11within crypto that are gathering data and selling data on to other companies. Okay, so they might be gathering huge numbers of tweets or [snorts] state images for training text-to-image like mid-journey type models or video, whatever it might be. That's fine. That's cool. But what I'm kind of particularly interested in where I think decentralization and having huge

21:36numbers of vantage points actually becomes really valuable is in building business intelligence building business intelligence models. Now, this is very useful because where you're looking at the internet from right now compared to where I'm looking at the internet from, we will see different stuff. You go on Amazon, you go on Facebook for example,

21:58meta. If you go on Facebook right now, they will show you ads based on the credit card that you personally have, based on how much money you make. That's meaning that lots of people are seeing lots of different things, lots of different information. That information is very valuable to people. It's very valuable to e-commerce companies who are selling stuff. They

22:17want to know where their products are ranking on Amazon today in New York compared to where it's ranking on Amazon in Dubai versus Brazil, wherever it might be. To gather this kind of data and information is something that's very, very difficult for a centralized model to be able to see it through all these vantage points. And it's where our real kind of moat, our separation from the

22:36rest of the market comes in as a decentralized model because it's easy to build the tech to build out a node, the nodes that people are downloading are not particularly difficult. But what is very difficult is to build a network of people in the number of hundreds of thousands. And I've learned this, you know, through my YouTube channel, building a

22:57following of 300k over, you know, it took me five or six years to be able to build that YouTube channel. It's a difficult thing to do. Something that centralized companies are not able to do. So, yeah, building towards business intelligence products, stuff that could be used for e-commerce, comparison sites, quant trading in crypto sentiment such a big important

23:19thing. Gathering sentiment analysis, selling it to crypto exchanges, to trading desks, huge.

Yeah, that is huge and level the playing field for people to earn in AI and not just contribute all day to the next version of Claude and Open AI and being being the product. So, I'm wishing you and the team all the best on it. I know it's live right now. I'm

23:45going to leave all the links to the site and the socials and the apps. In terms of heading towards the main launch and eventually people converting their points into tokens, is there any tentative timeline that you that you're seeing? Is this something that's happening in the near future?

yeah, so we moving towards kind of TG. I said to

24:12you before like it's a very dangerous thing. I know everybody everybody always wants to know when. It's the same in crypto. Everybody always always wants to know when. It's like we all really know that Bitcoin right now is going to go up at some point. Everybody wants to know when is the best time to buy. Actually the most difficult question. We see a TGE coming very soon. It's

24:32difficult for me to say exact dates because that hinges on exchanges, how long it takes for them to do their due diligence, etc. So, you know, I'm not going to promise anything because I'm waiting for them, but we're looking at the next kind of eight eight weeks, eight to 10 weeks to be generous.

Okay. Well, that sounds great.

could be much it could be much sooner

24:51than that, but I'm just

Okay. So, what you're saying is

Give me some Give me some touch reason and such

with the point system, check out the data set better earlier than later.

Yes, absolutely.

Awesome. Well, I'm definitely going to check it out. It sounds enticing. I don't know how many cats I can take photos of, but maybe there's some part

25:12of expertise of kind of data AI information that applies to me to get extra points. Otherwise, I'll be going full passive on all my extra phones and computers to earn while I'm working away on the show. I appreciate your insights into this and making AI more fair for everyday users, Peter. It's something that we really need and tying in decentralized

25:35networks super important. So, I'm wishing you and the team all the best moving forward towards launch and in the next few years in that grand vision that you gave and thanks for taking the time to come on the show.

Peter, thanks very much for having me. Really really enjoyed chatting about AI and data. Look forward to Look forward to seeing seeing if the

25:55future trends and fees will come come on the show again.