Latest / Crypto Coin Show / Decentralized AI and SLM’s will beat OpenAI? w/ Assisterr | Blockchain Interviews
Transcript
- 0:00Speaker 1: I'm Ashton Addison from the Crypto Coin Show and today
- 0:03Speaker 1: i'm Blockchain Interviews with Nick Havriliak, co founder and CEO
- 0:07Speaker 1: of A Sister.
- 0:08Speaker 2: Nick.
- 0:08Speaker 1: Welcome to the show and thank you for taking the time.
- 0:12Speaker 2: Ashtin, thanks for inviting me, and hope the next Spanish
- 0:15Speaker 2: Thar two minutes would be really insightful for people who
- 0:18Speaker 2: are going to listen to us, and for you personally
- 0:21Speaker 2: and for me, it's a great opportunity to take a
- 0:23Speaker 2: stage and share what we're building and as sister, Yeah.
- 0:26Speaker 1: I'm super interested in learning more through your insights about
- 0:30Speaker 1: the intersection of AI and Web three and more specifically
- 0:34Speaker 1: decentralized AI or community driven AI, and how we can
- 0:39Speaker 1: get as big or bigger than the big tech guys,
- 0:41Speaker 1: which AI is getting pretty useful, but I would love
- 0:44Speaker 1: to have decentralized versions as the go to, and I
- 0:48Speaker 1: think a lot of people would when they realize as
- 0:51Speaker 1: we dive into that in the conversation today, I would
- 0:53Speaker 1: love to start off our conversation just a little bit
- 0:56Speaker 1: on your background in AI, blockchain and what got you
- 0:59Speaker 1: interested in this field.
- 1:03Speaker 2: Yeah. Thanks, My background is basically into consulting, so I
- 1:07Speaker 2: started my career into it consulting. I was building and
- 1:12Speaker 2: the teams, manageer and the teams mostly in AI and
- 1:16Speaker 2: initially it direction. Then later on I speech fully to
- 1:21Speaker 2: webs three Space was doing exactly the same, so building teams,
- 1:25Speaker 2: organizing them, delivering some products. Then twenty twenty two, for
- 1:30Speaker 2: two years I joined AI startup where I was CPO
- 1:33Speaker 2: and I was also responsible for Errant team. We were
- 1:38Speaker 2: using SE. We were using recommendation algorithms and also computer
- 1:42Speaker 2: revisional algorithms too. Uptomates a very pragmatical part of business functions,
- 1:48Speaker 2: sales and design function for small medium businesses. It goes well,
- 1:52Speaker 2: it establishes a working business. I moved to a LA
- 1:55Speaker 2: from Ukraine and originally from Ukraine. I launched the office there.
- 2:00Speaker 2: I hired a couple of BD folks. Everything went well,
- 2:03Speaker 2: but then the war happened and like I just make
- 2:07Speaker 2: a decision to sell portion of my company to my
- 2:09Speaker 2: co founder. Moved to UK got stipant from National Bank
- 2:13Speaker 2: of Ukraine to starty at Cambridge Business School. Took maybe
- 2:17Speaker 2: one year. I just spoused to think about life, what
- 2:19Speaker 2: they want actually to do, and that's where I found
- 2:22Speaker 2: myself that I want to combine webs three element of
- 2:25Speaker 2: my journey with AI element of my journey, and that's
- 2:28Speaker 2: where my sistere comes from. So initial idea was, hey,
- 2:31Speaker 2: let's take AI technology and just up to made very
- 2:35Speaker 2: routine and where we boring and very complicated process of
- 2:38Speaker 2: developer on boarding and support from webs free productol sites.
- 2:43Speaker 2: We did it. We established it as a business. But
- 2:45Speaker 2: then at some point of time, a couple of months ago,
- 2:47Speaker 2: we figure out that like hey, all the infrastructures that
- 2:50Speaker 2: actually we build just for one use case can be
- 2:53Speaker 2: scaled and offered for many many people, individuals, companies, and
- 2:57Speaker 2: organizations to launch the own specific domain specific use cases
- 3:02Speaker 2: and earn money, monetize the knowledge, experience and build something
- 3:06Speaker 2: really cool and amazing. So right now we basically position
- 3:09Speaker 2: ourselves as a network of community owned small language models.
- 3:13Speaker 2: We relying on small language models instead of large language models,
- 3:16Speaker 2: and we relying on a community element, which we see
- 3:19Speaker 2: is it very beneficial if we want to create a
- 3:24Speaker 2: high level of curate data sets behind of each model.
- 3:29Speaker 1: Very interesting backstory, and now this part about the small
- 3:33Speaker 1: language models, when most people have heard of LM or
- 3:38Speaker 1: large language models, but they don't really know what that
- 3:41Speaker 1: means and I think CHGPT and the main AI functionalities
- 3:45Speaker 1: are lms. What is the main difference between small language
- 3:49Speaker 1: and large language and does it matter?
- 3:52Speaker 2: Yeah, So on a core it's pretty same technology. It's
- 3:56Speaker 2: like from technological perspective, they're pretty same, but it's a
- 3:59Speaker 2: core differenterrence is that with large language models, they usually
- 4:03Speaker 2: have a lot a lot of billions. It can be
- 4:05Speaker 2: eighty billions, it can be one hundred billions of parameters,
- 4:09Speaker 2: while with small language models it's usually up to eight
- 4:12Speaker 2: ten billion parameters and it can be even in law
- 4:17Speaker 2: like zero point two zero point three billion parameters. And
- 4:20Speaker 2: because of this smaller size, we can do two things
- 4:24Speaker 2: like first we can put these models on edge, so
- 4:27Speaker 2: it's mobile phones, laptops, small small devices. Then second part,
- 4:32Speaker 2: we can focus on highly curated data sets that's going
- 4:37Speaker 2: to be act as a brain behind this model. And
- 4:41Speaker 2: for curated and small data set, it's easier to find
- 4:44Speaker 2: this data. The question is like where to take for
- 4:47Speaker 2: which use cases and so on, and that's where our
- 4:51Speaker 2: webs real element comes in. So we are using these
- 4:54Speaker 2: small language models, but then we invite in community of
- 4:57Speaker 2: professionals who have expertise let's say in some product development,
- 5:01Speaker 2: in marketing, in in like whatever you can think about
- 5:04Speaker 2: from the day to day tasks that you are working on,
- 5:07Speaker 2: you can launch a small language models that would be
- 5:10Speaker 2: trained on specific data set and it's performing does highly
- 5:14Speaker 2: accurate and highly good. If you're talking about large language models,
- 5:19Speaker 2: everyone already tried. They're funny, they're cool, they're interesting, and
- 5:23Speaker 2: like initially it was very excited, but right now you
- 5:26Speaker 2: understand if you're trying to feed to use this model
- 5:30Speaker 2: for the main specific task or automation business automation, it's
- 5:33Speaker 2: not good enough. It's very average, and it's had a
- 5:36Speaker 2: lot of hallucination. So that's kind of a problem we
- 5:39Speaker 2: addressing with small language models with the created data sets.
- 5:44Speaker 1: It's very interesting. So are you saying that we could
- 5:47Speaker 1: make sort of like chat GPT for each industry or
- 5:51Speaker 1: for each business use case and just make the language
- 5:54Speaker 1: model just for that and then make it ten times
- 5:57Speaker 1: smarter than it would be if it had to learn
- 5:59Speaker 1: everything in the world world and try to be the
- 6:01Speaker 1: best at everything.
- 6:03Speaker 2: Exactly. So it's like it's as our brain is working,
- 6:06Speaker 2: like we don't have one cell that is doing everything,
- 6:09Speaker 2: but we have like different sectors of the brain. We
- 6:13Speaker 2: have different like region and like a lot of like
- 6:17Speaker 2: complex complex combination, and each of these parts responsible for
- 6:21Speaker 2: one task or one function or whatever. The same way
- 6:25Speaker 2: we see our system, our networks that we're going to
- 6:27Speaker 2: have maybe tens of thousand, maybe hundreds of thousand, maybe
- 6:30Speaker 2: even millions small language models trained by group of people,
- 6:35Speaker 2: and these people have in sent to an alignment to
- 6:38Speaker 2: keep this model up to date. Why it's important because
- 6:41Speaker 2: in the future, if you will think like what exists
- 6:44Speaker 2: in the market and how people are solving their problems,
- 6:47Speaker 2: there would be three options. Like first option it's still
- 6:50Speaker 2: old old school way with Google search, with other search engines.
- 6:54Speaker 2: Then we're going to have these l alms which can
- 6:56Speaker 2: cover actually a lot of tasks. Helped to craft some
- 7:00Speaker 2: tax articles, whatever else. But with llms you have the
- 7:05Speaker 2: pin six twelve months on the training data set because
- 7:08Speaker 2: they cannot allow like four to retrain the model all
- 7:11Speaker 2: the time. And if you just remind when it was
- 7:14Speaker 2: googling something and looking for some solution, you're not asking
- 7:17Speaker 2: like tell me what was RELO one year ago? You're
- 7:19Speaker 2: asking like what you do now? Like I need to
- 7:22Speaker 2: solve some problem Google, you have all the time problem
- 7:25Speaker 2: was the search engine that all the suggestion from them,
- 7:29Speaker 2: they come from the d model. So it's not all
- 7:32Speaker 2: the time best for you, but it's definitely best for
- 7:34Speaker 2: someone who paid to show you. This. With LLMS, we
- 7:38Speaker 2: redesigning it because if you're looking for something with LLAMS,
- 7:42Speaker 2: you will definitely find the latest possible solution for any problem,
- 7:47Speaker 2: any domains on that you're googling, like looking for or quarrying.
- 7:51Speaker 2: And then if something is missed, to have a community
- 7:54Speaker 2: maybe fifty, maybe one hundred, maybe thousand people around each
- 7:57Speaker 2: model that would be aligned to keep this model up
- 8:01Speaker 2: to date. So if they are professionals, they can like, okay,
- 8:04Speaker 2: there is a querius that model cannot se automatically. Then
- 8:10Speaker 2: some people will start competing how to feed additional data
- 8:14Speaker 2: to make this model smarter and to earn I don't know,
- 8:17Speaker 2: ten cents from you.
- 8:20Speaker 1: Very interesting So I'd love to dive into that economic
- 8:23Speaker 1: model more on the web three side, So creating a
- 8:26Speaker 1: decentralized AI. Anybody or maybe people who have the right
- 8:30Speaker 1: credentials based on that industry of where the small language
- 8:33Speaker 1: model is learning, they can contribute their knowledge or they
- 8:38Speaker 1: can query back and forth to help train the model
- 8:42Speaker 1: and then they're going to be rewarded with crypto or
- 8:45Speaker 1: some ownership in the network.
- 8:48Speaker 2: Yeah, exactly. So the quality is that we want several
- 8:51Speaker 2: Hakaptons in Denver recent with Bindance Labs, with okaex as
- 8:56Speaker 2: well in this specific AI web three incentives. How to
- 9:00Speaker 2: incentivize people to share the data which actually already exists.
- 9:04Speaker 2: If you will think about any domain zone that you
- 9:07Speaker 2: can think about. It can be even like B two
- 9:10Speaker 2: C marketing in China, let's say, or be too ce
- 9:13Speaker 2: bit too be marketing on some specific you know, like
- 9:15Speaker 2: country or region or whatever. There is a group of
- 9:19Speaker 2: people already who has this knowledge and expertise about this market,
- 9:23Speaker 2: but they have zero incentive to share these data. So
- 9:25Speaker 2: this data is all all the time stored on their
- 9:28Speaker 2: laptops and their mobile phones, in their accounts in different
- 9:31Speaker 2: software and so so what you're trying to do, we
- 9:34Speaker 2: try like and that's what we're building. We offering them. Hey,
- 9:38Speaker 2: if you have expertise and you want to create a
- 9:40Speaker 2: model based on your data, like, let's kick off this process.
- 9:44Speaker 2: Register and a sistorm, create a profile, choose the role
- 9:47Speaker 2: that you want to perform, if it's model creator, data
- 9:50Speaker 2: validator or data contributor. You would have different paths and
- 9:53Speaker 2: different abilities to contribute and monetize your time and knowledge.
- 9:57Speaker 2: But also you can organize people around yourselves so effectively
- 10:00Speaker 2: you like, we're trying to create this new wave of
- 10:04Speaker 2: entrepreneurs where this entrepreneurship would be directed to creating models
- 10:09Speaker 2: and contributing data to this model instead of creating let's say,
- 10:12Speaker 2: applications on App Store or Google Play market. So we
- 10:17Speaker 2: try to play this game, but instead of one owner
- 10:19Speaker 2: of application, a company who're going to be owned this application,
- 10:23Speaker 2: we're offering this ability to group of people. So initial creator,
- 10:27Speaker 2: he initially will have the biggest take of the ownership
- 10:30Speaker 2: over this model. But the more people contribute, the more
- 10:33Speaker 2: we track in and allocating some portion of this model
- 10:36Speaker 2: to different people. And like if there would be a
- 10:39Speaker 2: thousand people, they would have kind of a fractionalized ownership
- 10:42Speaker 2: of this model, and they would be sharing profits generated
- 10:45Speaker 2: by this model as well.
- 10:48Speaker 1: And with small language models being able to narrow down
- 10:53Speaker 1: into so many different industries, where.
- 10:56Speaker 2: Do you start?
- 10:57Speaker 1: Is there a focus for a start to start on
- 11:01Speaker 1: certain areas, whether it's development or web three or do
- 11:04Speaker 1: you just open the playing field and say hey, any
- 11:06Speaker 1: experts of their field can come in and start working
- 11:09Speaker 1: on the industry that they like.
- 11:12Speaker 2: Yeah, that's a very interesting approach because it's all about
- 11:14Speaker 2: the go to market strategies. So what we started with
- 11:18Speaker 2: it was like, hey, we know a really well AI
- 11:21Speaker 2: space and we know really well Web three space, Like
- 11:24Speaker 2: let's target this area and like what we actually can
- 11:27Speaker 2: build an automat in this area. That's why we started
- 11:30Speaker 2: to use developer support automation for Web three productols. Effectively,
- 11:34Speaker 2: we build with the smartest copilot or as system developer
- 11:39Speaker 2: assystem for many many top producols which can assist with
- 11:43Speaker 2: smart contract developments dub development. We is navigating into the
- 11:47Speaker 2: system how to build, how to deploy something, and it's
- 11:50Speaker 2: already really strong ground layer for us. Why it's important
- 11:54Speaker 2: because we see that in Web three space we have
- 11:57Speaker 2: a problem with consumer face and application. Why we have
- 12:00Speaker 2: this problem because layer one layer two they accumulate all
- 12:03Speaker 2: the value on the side and opportunity cost to launch
- 12:09Speaker 2: consumer face and application. It's approximately same as launch layer
- 12:13Speaker 2: one layer two, but like the output could be very different.
- 12:17Speaker 2: So what we're trying to achieve we want to give
- 12:19Speaker 2: this AI AI CO pilots or AI assistant to many
- 12:24Speaker 2: many people who doesn't know how to develop application fully,
- 12:27Speaker 2: who doesn't have enough resources to build a team of
- 12:30Speaker 2: five or six people, And if they be able to
- 12:33Speaker 2: launch this application on a mass market many many applications
- 12:37Speaker 2: and maybe invite just closest friends, maybe twenty, maybe fifty,
- 12:40Speaker 2: may bee hundred, it's already enough for them to earn
- 12:43Speaker 2: let's say twenty thirty fifty bucks per month. And for
- 12:46Speaker 2: many people in the world, it's a huge money, you know,
- 12:48Speaker 2: like there is a vanderstand which we like people earning
- 12:53Speaker 2: a couple hundreds bucks per year, you know, So if
- 12:55Speaker 2: they would have ability to launch something in use and
- 12:57Speaker 2: benefit which makes sense for them, they would be trying
- 13:00Speaker 2: and doing and playing with this. So the next step
- 13:03Speaker 2: was accept this. Like Debrail and websree focus, it's like,
- 13:07Speaker 2: of course, we keep building. We just announced today in
- 13:10Speaker 2: the morning that we're going to be opening one hundred
- 13:13Speaker 2: like we're going to be supporting on hundred to web
- 13:15Speaker 2: three productols like top hundred Web three producols in the
- 13:18Speaker 2: next two months. And next step for us we are
- 13:21Speaker 2: launching AI Lab. So with first use case and like
- 13:25Speaker 2: first approach, we were doing onboarding. We were doing like
- 13:29Speaker 2: establishment of this model on our site, so like our
- 13:32Speaker 2: core team was doing it. The next step was is
- 13:35Speaker 2: to give no code functionality for anyone who join in
- 13:38Speaker 2: our work system, to launch these small language models, to
- 13:41Speaker 2: organize all the contribution process, to distribute these models to
- 13:45Speaker 2: the communities on their own. So we don't want to
- 13:48Speaker 2: be a bottleneck. We want to provide the instruments and
- 13:51Speaker 2: then ecosystem that supports with distribution, with marketing, with many
- 13:55Speaker 2: other things. But these creators, developers, this should be a
- 14:02Speaker 2: people with entrepreneur sheep mindset a little bit who wants
- 14:06Speaker 2: to make it accessible.
- 14:08Speaker 1: M h yeah, very well, said I'm looking forward to that.
- 14:11Speaker 1: And I also saw on a sister roadmap there's these
- 14:17Speaker 1: autonomous agents. Is that different than the assistant for developers.
- 14:24Speaker 2: No, it's like basically, what is the core of our
- 14:27Speaker 2: solution eats a small language model, so a smaller and
- 14:30Speaker 2: engage model. It's a brain kind of operational system. But
- 14:34Speaker 2: then this operational system can be connected to different UI interfaces,
- 14:39Speaker 2: and this QUI interface can be like different childboards, conversational
- 14:43Speaker 2: chat boards, web vegets. It can be integrated to web
- 14:46Speaker 2: platforms like whenever you're thinking about, you can do this
- 14:49Speaker 2: integration because this integration, it is only interface, but the
- 14:53Speaker 2: core magic happening if you have strong brain that's that's
- 14:57Speaker 2: basically a course and this edge. We already have several
- 15:02Speaker 2: several UI interfaces which I mentioned, like different conversational chat
- 15:06Speaker 2: boards for discord Telegram. Right now it's slight under development.
- 15:10Speaker 2: We have web digits that can be embedded to any
- 15:12Speaker 2: web platform and so on. The next step that I
- 15:15Speaker 2: see in the future that we would love to work
- 15:18Speaker 2: and focus on, but it's going to be a latest
- 15:20Speaker 2: stage for us. It's to launch personal a assistant that
- 15:24Speaker 2: would be able and capable to go to the open
- 15:27Speaker 2: web and search the best small language model or combine
- 15:31Speaker 2: several small English models. If the user or like customer
- 15:35Speaker 2: or owner of this assistant want to achieve specific results
- 15:39Speaker 2: and to make it possible, this agent should or personal
- 15:42Speaker 2: assistants should have some balance in crypto to be able
- 15:45Speaker 2: to pay for small language model, large language models, or
- 15:48Speaker 2: will search or even humans. It's defense like who will
- 15:52Speaker 2: be reaching out to get service that you need and
- 15:56Speaker 2: our goal to have this network of south and maybe
- 16:00Speaker 2: hundreds of thousand models small language models that would be
- 16:03Speaker 2: winning calls, a kind of API calls from this personal
- 16:08Speaker 2: assistant that I believe everyone going to have is just
- 16:11Speaker 2: a question of time.
- 16:12Speaker 1: M hm.
- 16:13Speaker 2: That's very exciting.
- 16:15Speaker 1: And what do you think about the perspectives in the
- 16:19Speaker 1: small language models and decentralized AI catching up to the
- 16:24Speaker 1: user base of the large language models and you know
- 16:26Speaker 1: the amount of people that are using chat GPT right now,
- 16:28Speaker 1: how quickly will this be able to catch up?
- 16:33Speaker 2: Yeah, I think with chat, jipit and a couple other
- 16:36Speaker 2: players in the space who have the situation when big
- 16:40Speaker 2: tech companies they are just firing billions of dollars to
- 16:45Speaker 2: shift the user behavior. So every time when you know,
- 16:49Speaker 2: like there was a big invention, is it a steam engine, electricity,
- 16:53Speaker 2: mobile phones, Internet, there was a huge you know like
- 16:57Speaker 2: initial huge investments should have and to change behavior of
- 17:01Speaker 2: the people's perception of the people, like how I can
- 17:05Speaker 2: achieve specific results because like there was old way to
- 17:08Speaker 2: do things, there is new way and they're going to
- 17:10Speaker 2: be you know, like even better way to do things.
- 17:13Speaker 2: So I think that's not direct competitions. I think all
- 17:18Speaker 2: these guys they are educating the market and educating the community.
- 17:22Speaker 2: The next step for everyone would be logical. Okay, I
- 17:25Speaker 2: can achieve this level of accuracy or this level of
- 17:28Speaker 2: automation or this you know, like amount of money I
- 17:31Speaker 2: can extract if I would be working with these tools,
- 17:34Speaker 2: And okay, there is alternative. So what we are doing.
- 17:37Speaker 2: We're not saying like, hey, we're going to compete directly
- 17:40Speaker 2: with any of these big players, because direct competition is
- 17:43Speaker 2: like we will lose most likely, it's almost impossible. But
- 17:47Speaker 2: to create an alternative where maybe ten, maybe twenty percent
- 17:51Speaker 2: of the market will choose this alternative would have other
- 17:55Speaker 2: way to contribute data, monetize the dat monetize the data,
- 17:59Speaker 2: establish the business use cases, build a community owned small
- 18:03Speaker 2: language models as you know, like small business. That's the
- 18:06Speaker 2: way is that I want to go. That's the way
- 18:08Speaker 2: that the sisters gonna be evolving over time, and I
- 18:11Speaker 2: expect that there would be a lot of community involvement
- 18:15Speaker 2: in this process. So right now, we have one hundred
- 18:19Speaker 2: fifty thousand registered users, which we get over the first
- 18:23Speaker 2: months after we launch a platform, which is a huge
- 18:26Speaker 2: So what we're doing right now we launch a huge
- 18:29Speaker 2: quiz use the research where we're asking them what type
- 18:33Speaker 2: of models you're excited about, what use cases you want
- 18:36Speaker 2: to see on a platform, what type of data you own,
- 18:39Speaker 2: what is your domain expertise on? So based on like
- 18:43Speaker 2: many many other questions. So based on this information, we're
- 18:46Speaker 2: going to have a huge demand side expectations. So it's
- 18:51Speaker 2: like there is huge user base who expecting to use
- 18:54Speaker 2: this type of models. And then by having this information,
- 18:57Speaker 2: first we can customize the product itself, make it better,
- 19:02Speaker 2: more precise for this audience that already is there, the
- 19:04Speaker 2: community that is there. Second, we can go to any builder,
- 19:08Speaker 2: any company in this space and say like, hey, do
- 19:10Speaker 2: you know how to distribute your product? How do you
- 19:13Speaker 2: distribute your models that you're working on? And in most
- 19:15Speaker 2: cases if you I just came from Brussels and I
- 19:18Speaker 2: was mentoring one hakatn and I chatted with five teams
- 19:22Speaker 2: and I chatted with guys like, hey, it's really interesting
- 19:24Speaker 2: what we building, but like, how are you going to
- 19:26Speaker 2: do a distribution of this platform like or model or application?
- 19:30Speaker 2: And they said, like, oh, people will found how cool
- 19:33Speaker 2: this is, you know, like and like they will start
- 19:36Speaker 2: using it. No way, Like it doesn't work that way,
- 19:38Speaker 2: Like if was early nineties, maybe maybe two thousand, but
- 19:43Speaker 2: not right now. So what if we can give any
- 19:45Speaker 2: of the builder. We can say like, hey, we already
- 19:48Speaker 2: have fifty thousand users who are excited about Web three
- 19:52Speaker 2: analytics and they are ready to use it, and also
- 19:54Speaker 2: they are ready to contribute the data there. Or I
- 19:57Speaker 2: can say like, hey, hundreds and users who are excited
- 20:01Speaker 2: about some social twins that is trained on the data
- 20:05Speaker 2: from the community chats, and like they are ready to
- 20:09Speaker 2: play with this, and you would have a lot of exposure.
- 20:11Speaker 2: So it's highly vailable. Sinks with the distribution and organize
- 20:14Speaker 2: an all ecosystem and community around it, and that's how
- 20:18Speaker 2: we focus. Yeah.
- 20:20Speaker 1: No, very well said, and I completely agree. There's there's
- 20:23Speaker 1: so many projects now. You can build the greatest thing
- 20:26Speaker 1: in the world, but if you don't get the right
- 20:29Speaker 1: channels of distribution or go to market, then it's hard
- 20:32Speaker 1: to get adoption on it. So I'm glad you guys
- 20:36Speaker 1: are building this out now. I saw that and you
- 20:40Speaker 1: mentioned it earlier. There was a one point seven mil
- 20:43Speaker 1: pre seed that was just closed. Congratulations on that. I
- 20:45Speaker 1: wanted you to just talk about some of those those investors,
- 20:50Speaker 1: the partners and the partnerships that from this raise that
- 20:54Speaker 1: you're looking to partner with or capitalize on to take
- 20:58Speaker 1: a sister to the next level.
- 21:01Speaker 2: Yeah, thanks for translations. It's actually very big at human
- 21:06Speaker 2: for all our team because we started this project when
- 21:09Speaker 2: it was a bear market. We started and we revive
- 21:13Speaker 2: over the Borough market. It was really complicated, like wherever
- 21:16Speaker 2: we were starting started a company, you know, like between
- 21:20Speaker 2: these like green cycles. Definitely understand how painful it is.
- 21:25Speaker 2: And even like Webstree investors that we were chatting earlier,
- 21:28Speaker 2: they were saying like, no way crypt is dead, like
- 21:30Speaker 2: there is no developers, like, no no reason to loch
- 21:33Speaker 2: edits in there. It's like whoa really bad? But so
- 21:36Speaker 2: far it sounds like right now it feels really good.
- 21:39Speaker 2: So we secured this one point seven million from a
- 21:43Speaker 2: pretty good like group of investors like Websuite, dot Com,
- 21:48Speaker 2: muonfil Capital, Cantanga Guys, Outlier Adventures, ex ventures like a
- 21:53Speaker 2: lot of great angels from Zero Gravity, Michael Ice here, Swissbourg,
- 21:58Speaker 2: eigen Layer, Bay, Particle Network, and so on. Right now,
- 22:02Speaker 2: our focus is to scale the product. That's why we're
- 22:07Speaker 2: hiring a lot of people. I just made a post
- 22:10Speaker 2: maybe ten minutes before our stream that we're hiring right
- 22:14Speaker 2: now our system lead. We're going to do a lot
- 22:17Speaker 2: of effort in that direction. I want to have more
- 22:20Speaker 2: vertical small builders and I'm going to be mentioning like
- 22:23Speaker 2: we're going to be mentioning it with this user research
- 22:26Speaker 2: and expectation from our community just to you know, like
- 22:29Speaker 2: to build something that people really need. And except that,
- 22:33Speaker 2: we're also adding more people onto AI side because we're
- 22:36Speaker 2: doing a lot of research. So right now we have
- 22:38Speaker 2: three three engineers in our team point who is constantly
- 22:42Speaker 2: doing research because the AI space right now is just
- 22:45Speaker 2: changing every week, like every week something happening, and it's
- 22:48Speaker 2: not something you know like itself. It's developing, so fars
- 22:52Speaker 2: that you need to keep the space. You need to
- 22:54Speaker 2: you need to understand what is going on, how you
- 22:56Speaker 2: can commercialize something that you know, like just was releasing
- 23:00Speaker 2: so on, and also of course developers just to keep
- 23:03Speaker 2: building releasing everything. We establishing processes, we scale. We're going
- 23:07Speaker 2: to scale a team even more. Accept this like a
- 23:10Speaker 2: huge product milestone for us. It's AI Lab, so it's
- 23:13Speaker 2: no called solutions, so a lot a lot of people
- 23:16Speaker 2: will come and with their creative ideas, we can expect
- 23:20Speaker 2: to explore and see even more creative and interesting use
- 23:23Speaker 2: case at on our platform and over the old network
- 23:26Speaker 2: in general. And also we want to move fast, so
- 23:31Speaker 2: I expect that somewhere autumn time we might go for
- 23:35Speaker 2: another seat round if we will need our targets in
- 23:41Speaker 2: terms of the community, not community but user base, and
- 23:45Speaker 2: also in terms of the verticals that we're going to
- 23:47Speaker 2: be adding by relying on the AI.
- 23:50Speaker 1: Lab amazing And on the community side, what is the
- 23:54Speaker 1: best way for people to contribute or try out on
- 23:58Speaker 1: the platform and see where they might be able to participate.
- 24:04Speaker 2: Yeah, of course, like right now, it's the easiest way
- 24:07Speaker 2: to participate. It's we launch incentivized test nets. So right
- 24:11Speaker 2: now anyone can go through different user flow choose his
- 24:14Speaker 2: role right now do you want to be model creator,
- 24:17Speaker 2: data contributor or data validator. Also you're going to have
- 24:20Speaker 2: a lot of you know, like ability to test different functionality,
- 24:24Speaker 2: participate in some valuable activities with us, Invite to your friends,
- 24:30Speaker 2: claim some synthetic talk, and how we call it uses
- 24:34Speaker 2: synthetic talking to query different models, interact with different models
- 24:39Speaker 2: that are already on our platform, create some tickets, and
- 24:42Speaker 2: many many other things that you already can do and
- 24:44Speaker 2: already can participate. And also think about which type of
- 24:48Speaker 2: models you would love to create a contribute to, because
- 24:51Speaker 2: if it's another app store, generative AI powered app store,
- 24:56Speaker 2: being at the beginning at this, you know, like wom
- 25:00Speaker 2: some sense, it is highly valuable because then you can
- 25:03Speaker 2: choose different direction which is empty and available that no
- 25:06Speaker 2: one is actually buildings there, no one is present with
- 25:10Speaker 2: some solution. It's something that could you know, like post
- 25:13Speaker 2: fortune for you in the future if you will use it.
- 25:17Speaker 1: Yeah. No, it's very exciting to be able to contribute
- 25:20Speaker 1: to the next AI model that is in your expertise
- 25:25Speaker 1: and be rewarded for it as well. And I think
- 25:28Speaker 1: is very valuable because it could be used by millions
- 25:31Speaker 1: and millions.
- 25:34Speaker 2: Exactly, exactly, very cool.
- 25:36Speaker 1: Thank you so much Nick for it for the insights
- 25:39Speaker 1: into decentralized AI and the merge of Web three. I'm
- 25:44Speaker 1: personally excited to have more small language model alternatives to
- 25:48Speaker 1: the LMS, and I think it's something that it's going
- 25:52Speaker 1: to be very big. AI has just been huge in
- 25:55Speaker 1: Web three. We're finally seeing how the two industries together
- 25:59Speaker 1: are a creator value and how other people, not just
- 26:02Speaker 1: developers but community people can also get involved in it.
- 26:05Speaker 1: So I'm looking forward to following along with the growth
- 26:09Speaker 1: of a Sister. I'm going to check out the community
- 26:12Speaker 1: platforms and I'll put a link to the platform and
- 26:15Speaker 1: the socials as well in the show notes below to
- 26:17Speaker 1: be able to can contribute and all the best moving
- 26:19Speaker 1: forward with a Sister and everything in AI and web three.
- 26:24Speaker 2: Thanks such then, how good to time, Bye bye