Guide to Building AI Products Quickly & Easily with Aladin Ben

But in today's world, you have to be generalist, like saying, no, I'm, my expertise is in something, you know, is simply not enough to really be a successful entrepreneur. And you have to fight for that. And I believe that, you know, being able to quickly build something yourself, validate it with your initial customer, with your initial community, whatever, it's going just to be such a benefit every single developer that is going to join your team and work on the scalable product will have a better understanding of what you are trying to achieve if you already were able to deliver it in a prototype. We are Memorality, and what we are building is a platform that enables everyone to basically turn simple ideas into software. The idea is we all, you know, are industry experts in our niche. We understand problems that other people might not even be aware of. But it's like pretty hard to get the solution going, specifically when it requires software and technology. Usually in the first step, uh, that you need to do is like hire some expensive engineers to help you like to, to prototype or build your MVP. It's like taking a lot of time and it's quite inefficient in terms of costs. Um, you might potentially be spending tens of thousands of dollars and months months of work and you haven't even really validated your idea yet. So we are trying to like figure out how can we use AI for that. But with token limitations and hallucinations, which a lot of you might be aware of, it's pretty hard to create scalable projects. If you're like trying to build some sort of a simple HTML front end, AI can do that for sure. But if you're like trying to build a more complex problem, it becomes like really, really hard the more software you're basically producing, the more likely it is that the AI will do a mistake or two. And you are not even able, like, to really debug it because of code complexity. Even the AI at some point isn't really aware of the contextual data anymore, and therefore debugging this code can become, like, really hard unless you use Memorelity. And basically what we are doing is we are building a domain-specific programming language. So you might be familiar with like Python or HTML or C++. Those programming languages are made for humans, not really for machines. You would probably use binary for a machine. Now with large language models, they can try to use these programming languages that have been made for humans, but it's not really made for LLMs. So we have to rethink the way we build a programming language from scratch. Specifically for LLMs. And on average, Dew, which is the domain-specific language that we created, is 90+ percent shorter than traditional programming languages. And basically the way it works is you can wrap any sort of code inside of a Dew sequence component and then call that using what we call nodes. And that is likely to reduce the errors because Each node is a component, a validated component by itself. And if an AI creates that specific component for you, then essentially because it's limited in size, it is less likely for it to do an error and for you more likely to build on top of. Now, Do-Itself— sorry, this is— there we go. So Do-Itself is not really made for humans. While it is human readable, it is specifically made for large language models. What we have built on top of it is what is called DoVisual. And DoVisual functions pretty, pretty simple. And you will see in a second why. You have a playground. And in your playground, you have a library of nodes. And these nodes are encapsulated components. They can be machine learning models like ChatGPT-4, or they can be connectors to different platforms or spreadsheets. But you can also create your own with your own logic or having your developers or AI create them for you. Once you understand what you want to build, you just have to drag and drop the different nodes, connect them, configure them, and create an entire workflow like the one that you are seeing right now. It's using an Arabic speech-to-text model that we created. It understands Arabic dialect, then creates a transcript out of that audio file, sends it to OpenAI to generate a response using GPT-4, and then takes a response, sends it to ElevenLabs, to basically create speech out of the text that we've received from OpenAI. And basically you can build this in a couple of days with like an engineer, but with our platform you can actually build it in a couple of minutes. And I'll show you later how we are actually doing it. The idea is that when we can create software on scale, a lot of these components become reusable. We are working with one of the largest hospitality companies in the world called TUI. They are managing over 200,000 rooms worldwide to ship an on-premise assistant for hotel guests so that you are able to order room service without taking the phone, but talking to an Alexa-like system that is trained on the hotel data. Or one of the other things we did is work with the government of Dubai for legal tech to basically compare the different jurisdictions and their laws or frameworks in terms of like crypto and virtual asset regulation. And similar to that, we can also like ship our entire own products. Like for example, we've built a shipping and logistics workflow that enables through natural language to find the fastest route by sea and then inquire through personalized emails for pricing, but also like space if you want to ship like multiple containers or and so on. Our business model is pretty simple. We've been focusing from day one on on-premise work. So we want to make sure that everything that we develop can kind of stay in-house, which is like one of the largest challenges across different industries. So for example, if you're like in finance, insurance, or health tech, it's more likely that you require an on-premise solution rather than a cloud solution. And we are trying to stay as competitive as possible with our on-premise solution as well as with the cloud solution. But we are not here to talk about our traction or business model. We are here to talk more and more about what can we build. When I was younger, people told me like, hey, you gotta focus on one idea. And I like to think of my experience in the video game industry as the following. In video games, everything is code. So people that create character design or level design or networking solutions or anything else in video games, Fundamentally, all of that is code. To ensure that designers, marketing managers, and so on can contribute to a game, game studios built what we call a video game engine. It's an engine that basically creates a visual layer on top of the code to enable them to contribute to the development of the video game. And that's what we have done with AI. So we created an AI engine that enables you to quickly build and exchange those different nodes. Um, I'm gonna jump a little bit further just to tell you a little bit more about our team. So we are a very small team. Um, so back then when this presentation was done, we were 7. Now we are 9. Um, we are trying to be as efficient as possible. Um, so we hire only out of our network. We have people that have been in senior positions at companies like Amazon, Deloitte, and Samsung. My co-founder is a PhD at the University of Michigan in probabilistic programming and machine learning. And I have that experience from the video game industry. In fact, today, larger scale companies, they have trouble with really iterating quickly and moving quickly and understanding the pain points of the customers. And that's why you kind of like every now and then see like one of the largest companies acquiring smaller startups because they are far more efficient in solving certain problems than those other companies. And the reason why I'm sharing this with you is because, you know, I've been through Founder Institute myself, I think in 2019 here in Tunisia, and it is very clear the founder journey isn't just about building products. I think building a product should be the easiest aspect of all of it. What is more crucial is building a community and building a team. And I can only recommend to every single founder out there especially with the emerging technologies, and I'm not just talking about the ones that we as Non-Morality are working on, but generally speaking, what they should focus on is community and building the right team. And building, especially in an entrepreneurial environment, means that you have to be mindful of resources, but you also have to push beyond of what is available. I hope that is clear so far. So I'm going to switch back to the screen just to see if there were any questions meanwhile. Um, let me check here. All right, so great. Uh, right now you can actually give it a try yourself. Um, if you visit memorality.com, you will find DoVisual, and DoVisual is the platform that enables the building of the different notes. Unless there are any more questions, I'm more than happy to share my screen and we jump right into it. All right, I don't think that there are any other questions. Oh, maybe there is. Yeah, fantastic question. How does Memorality handle variable passing from one node to the next? I think it's maybe best to jump right into it. Basically what we do is we do mapping. So we enable the user right now himself to map, and we are using in the background reinforcement learning through human feedback to basically train a machine to do the mapping by itself. So I'm gonna just open it here really quickly, and then like each one of you will be able to create an account on DoVisual and go try it out themselves. We are still kind of like in stealth beta mode, so while everything is online and people can try it out, we are not production ready yet, uh, for the public. So please feel free to go to the platform and just try out like one of the different playgrounds. Let me share my screen. Uh, so this is Duo Visual, um, and as I mentioned before, it's basically to automate different workflows. Um, you can use what we call nodes. Nodes can encapsulate all sorts of code, and by kind of like knowing what you want to do, like what you want to create, you just have to drag and drop it into place, connect it. To build that functionality out. So let's see how that looks in theory. And for this specific example, I'm going to even create the nodes from scratch. So we start by new playground. Let's say Founder Institute demo and building a cool chatbot. So as you can see already here, this is our playground. We have a bunch of node modules here, and each one of these node models already is kind of validated by yourself. So if we take, for example, the Groq one here, which is a hardware manufacturer for AI models, an extremely fast one, we can see that in the configuration we can pass on an API key, we can pass on some memory, we can select the model, we can prompt it and even give it a user input. And when we do that and we run it here, we can see that it just rendered for us a response. So let's start by maybe telling it, you will act as a personal assistant for startups, make funny and short recommendations. And then in the user input, we could say something like, uh, hey, do you have any suggestions on what I should work on? So now if we are trying to render it in terms of a playground that already exists within DoVisual, and we just want to quickly create a workflow to use something to solve a very specific problem. We can just execute it as DoVisual, and it will basically bring the entire response here to us. But there's also a development mode. So let's say, for example, instead of getting a text response, we want to have an audio response. One of my favorite text-to-speech models is a platform called Unreal Speech. It's similar to ElevenLabs or PlayHT or what you might be familiar with. Now we are getting to the question that has been asked earlier, which is how are we doing the JSON mapping? Let's do this really quick. We drag and drop the Unreal Speech, we connect these two nodes, and now we know that our next node, because if we go into the detail, we can literally see it, it's expecting a text to render. What we want to do is we say whatever the text is supposed to be should be replaced by the response from the LLM. That's how we map the different outputs. Let's say now, for example, we want to render this. We say executeDoVisual. I might hit some limitations from Groq because I think there's some rate limits on their API call. So let me try this again in a second. Or actually, I could use another API key. I guess it's a demo effect. Let's try it one more time. There we go. So as you can see, it didn't just like create the text now for us here, but it also generated the entire audio. Sure thing, Memorality team. With here are some output of the LLM. Now, if you don't want to use this through DoVisual, but you want to integrate this in your own platform, and I will go into detail a little bit later, but for now, I just want to make sure that we are displaying it. We can also just execute it through an API call. And for that, I'm going to share my entire screen with you because I need you to see Postman. Great. So you should be able now to see my entire screen. So now what we can do is instead of saying, hey, we want to execute this in DoVisual, we want to execute this through an API. An API is basically a way for the backend to communicate with— sorry, for the frontend to communicate with the backend. Most of the things that you actually see in the internet, they're not being processed on your computer or on, on the website itself that you are visiting, but on a server somewhere. And for that front end to communicate with the back end, you're using basically an API request. And for me not to build right now like a complicated API request, I can just use a client which is Postman, um, to do an API test myself. Now when I click here on Get Docs I can get— sorry, I can get the entire documentation of how to use this specific playground from any other front-end or application. And all I have to do is select this endpoint, I paste it here into my Postman. It's a POST request, and then I copy the payload into the body here. And I say, for example, instead of, uh, do you have any questions and suggestions, I say like, uh, hey, my name is Aladdin, what's yours? And when I execute it directly from here You can see I received a response and this output that we received right here is basically the playable audio file which I can now display on, let's say WhatsApp or on any other platform. To show you really quick how easy it actually is, I'm going to just stop presenting one more time.

And then there you go.

Now let's say, for example, we want to receive the, the response on WhatsApp and create like a WhatsApp agent. For that, we can use, for example, Twilio. And all we have to do is connect this here again. And we can check directly in the configuration what it's expecting. So we see here it's expecting the body. This has to be like some value that will be posted to the user. So what we say is, hey, whatever the body is supposed to be, we want it to be the response of the LLM. And then we update this. And I think I have to share my whole screen again to be able to show you my WhatsApp. There you go. Do try to not read my personal messages, but I just want to show you. So this is the Memorality Twilio account, which is already preconfigured here, as you can see. So we put in our API key from Twilio, which is a service provider for virtual phone numbers. And we received a Twilio phone number that we are using for the Memorality account. And here I'm putting in my personal WhatsApp number. Obviously, this can be completely remapped depending on what kind of front end you are using or if you want to use it by building directly a WhatsApp smart AI chatbot if you want so. So now let's see if we actually get a response to WhatsApp if we try to execute this playground again. So Groq gave us a response. Unreal Speech created the audio, and we can see that the response from the LLM was perfectly formatted and sent to us on WhatsApp. And similar to this workflow, we could be sending out automated emails like I shared earlier in the presentation where we are doing personalized email outreach. Also, it can be used for WhatsApp, it can be used for your own website. And to jump in and show you like other tools, not just like the ones we are using, we are going to use OpenAI competitor called Anthropic, which is building an AI called Claude, which I'm a huge fan of, to build a very simple front end and connect it to the chatbot that we just built. And to show you how simple that is. We can visit the Cloud website and we can ask it, hey, I want you to help me today to build a chatbot for my customer using an API that I will provide. This will be a simple demo. HTML, JavaScript, and Tailwind will be enough. Here is the API that we want to connect to. Make sure that everything fits into an index.html. So we give it this simple prompt and then we can delete value again here because we are not using WhatsApp right now. So we have our Groq, we have our Unreal Speech, we get the documentation here. So we have the endpoint, which we will just copy. We'll give it to Claude and we take our payload example. I say payload example. I'll push it here. So now that cloud magic will happen, and while it is producing the code for us, what we are going to do is open a new project in a, in a simple editor. So visit my desktop, fi-demo. We create a new file, we call it index.html. We can see that it produced the code for us, so we just copy the code. We put it into our index.html and we save it as, just in case, demo.html. Done. Now we can open it in the browser. Now we have a startup assistant chatbot. We say, for example, hi. So now it encountered a problem, and I believe the reason for that is that it doesn't really know how to render the response. But because we did that before and we kind of like know what the response is, all we have to do is take this from our Postman and in Cloudways say, hey, so the response is going to be to be an MP3 file, which will look like the following. So make sure to account for it. I hope everyone is still seeing my screen. I should maybe monitor this more than I am. Hey, is everyone still hearing me? I think my browser might have frozen here a little bit. All right, great. So let's see what's going on with Claude. Let's try to reload this. You have no idea how happy I am that Claude is the one that crashed and not our platform, even though we are the ones that, you know, like, are the smaller startup by far. So let me just exit this page for a second. And then we just visit again. There you go. So now it's producing the updated code accounting for the response format from our backend. Um, and while I don't want to share too much, but we are kind of working on a solution that can completely automate this process from frontend to pushing entire backends to actually going live and publishing your product, um, through us instead of like using multiple different products. So if we just take this code by now and put it here, we save it, we refresh this page and we say hi, we can see that it now responded with the audio file here. Now obviously we can optimize this a lot. And by learning a few techniques, it is pretty simple today to set up even a more complex project, for example, using Next.js and React. Essentially, you have the smartest piece of technology ever created by humans, which is AI. And you're able to ask it all sorts of questions. Today you can ask, what is the best framework to build a website like a personal assistant or a chatbot? And it will tell you, for example, hey, you could use React or Next.js or whatsoever. What we are mostly focused on is the backend aspect because we feel like the backend is a significantly larger problem to solve than trying to solve design issues. Specifically, not every app kind of like requires an individual frontend. So some apps shouldn't be apps but more like a WhatsApp chat. Or Google Spreadsheet or a simple email that you receive and are able to send. Um, so what we are focused on is the workflow that is in between kind of like the individual website and whatever exists already out there. And while the example that we just had a look at is pretty simple, um, the use cases that we are working on can become very complex. So one of the use cases, for example, and I'm more than happy to share post this conversation a lot more is, for example, an on-premise for hospitals to create medical reports. That's actually quite the challenge to solve. And you have some startups like Nabla, for example, that raised $27 million plus to solve that issue. We were able in a couple of weeks through our platform and different technologies to solve that issue. Nabla works on the cloud. We've built the first medical report assistant fully on-premise without the data leaving the hospital premise. We are able to fine-tune models, we are able to generate datasets and create everything in between. I hope you enjoyed it. Please feel free to go to dev.do-visual.com. I'll post it here in the chat as well. You can create an account, you can check out the playground that we just created, or you can check out any of the other playgrounds um, that we have, as well as the notes. Uh, and for anyone else interested, more than happy to schedule a follow-up call and go through the building and creation of different nodes, as well as individual use cases and how we potentially can help you to launch. Thank you. Thank you, Kevin.

Thank you so much, Aladin, for the presentation. That was very insightful. I hope everyone, um, really liked it. Uh, it It was very— I like the format of a workshop. It's better than just presenting like PDF presentations, as you said. So we have— we don't have like a lot of questions, but let's try to take some. Otherwise, you can interact with the audience at the networking lounge. So we will have after this session a networking lounge, so please stick around. So we have a question here from Daniel, and he's asking, can we build real-time AI products, for instance, real-time phone call audio transcriptions and translator or real-time video processing?

Yes, absolutely. So one of the things that we are working on right now is the integration of native WebSockets so that you are able to do the processing in real time. Essentially, it always depends on where are you hosting, as that is kind of like significant as well. Obviously, if you have your tech stack hosted on a stronger infrastructure, you are usually able to kind of like render the responses a lot faster depending on the different technologies you're using. However, In regards to the phone call, we actually have a demo ready. More than happy to have a follow-up conversation with you, Daniel, and show you the exact use case and hear also a little bit more about what you have in mind. Also, I would love to clarify something really quick. So we've been using this a lot to build use cases as consultants for other companies or to partner up with industry experts, people that understand the medical field like Dr. Alexander Mbaza, who's a vice dean of the University of Paris, or with TUI directly for the hospitality, and many, many other parties involved. However, the reason why we are planning on giving people access to this and making it as cheap as possible, specifically for entrepreneurs, and one of the programs that we created together with the government of Saudi Arabia. So if you visit memorality.com, you will find KSA Labs. And KSA Labs is a $40,000 grant that we can help entrepreneurs to receive by the Saudi government to build at least the MVP and take the first steps towards entrepreneurship and building a product. Our goal isn't to sell this to entrepreneurs. In an ideal world, we could make this absolutely for free and make more than enough time to really contribute to each one and each individual journey that you guys will be on. So please don't be hesitant to reach out and really like seek a conversation with us. We are more than happy to help you solve it. We've been entrepreneurs, you know, for as long as we can think, but it's always been the people that pushed us to really reach out and receive the mentorship that we need to take the next level. And like a barrier, like the feeling, hey, maybe this guy is trying to sell me something. No, just, just reach out. Like, we won't bite, and the worst case scenario, we will tell you, I'm sorry, we won't be able to help you. But best case, maybe your use case is less complex than you might have in mind right now, and we will be able to quickly help you.

Great, thanks, Aladdin. So let's take another question, uh, here. Um, it's a pretty technical question asking, is there a limit to using each service, like the nodes? In memorality?

It depends. So for example, Groq, if you don't have a paid plan right now, you have rate limits. So for your API key, you are limited in the usage. But because you are bringing your own API keys, you're only really limited to whatever the company that provided you with the API key is limiting you to. So if you're using OpenAI and you have like, let's say, a few thousand dollars in credit on your OpenAI account, you won't really hit limits that quickly. But if you are using, let's say, a free plan of 11Labs, then obviously your limit will be whatever the free plan is enabling you to do.

Great. I'm going to try to give you a question myself. It's more like a general question about AI and what you presented. I'd like to ask you, since you have pretty huge experience like using AI products. So what are the key challenges you see in AI adoption among these startups? And how do you, how you as Membrality, help solve these challenges? If you can elaborate on this.

Just to make sure that I got your question correctly, like how many different use cases can we cover with what we are doing?

Yeah, exactly.

So basically, We are industry agnostic. Obviously, there are a lot of low-level technologies that won't be solved through our platform. For example, you won't create a new way of generalizing 3D assets through machine learning through our platform. It's very unlikely. But for most ideas that exist today and most platforms that people are working on, they're very similar to what already exists. So let's take as an example 5 different platforms that act as a SaaS, each one for a different vertical, each one with their own unique proposition. Now, if you're thinking about stuff like authentication so that the user can sign up with like Google or Facebook or their own email, or if you think about stuff like the integration with Stripe so you can collect payments, or you think about I don't know, a simple newsletter. All of this isn't really a unique differentiator. It's not really a key differentiator. What makes you unique is the functionality that only you kind of have developed. Everything else you share with everyone. So why should you spend potentially millions of dollars on reusing existing components if these components are already out there? And even if you have to build them from scratch, Why would you require engineering hours for your product team to iterate or prototype relatively quickly? So what we are trying to create with DoVisual is a platform where code components become reusable. And most of the things today are reusable. If you're using an LLM, then the only differentiation might be the prompting, or are you using RAG or not? Are you rendering the outcome as a text or as an audio? All of these can be simple configurations and don't have to require engineering hours. To use this to also answer a question that has been asked earlier, which is, it's clearly that we have a limited set of nodes. Yes, what you see right now today on DoVisual is our public demo. Obviously, it would be a little bit risky to give you access to the version where we are building things for actual customers, which will make it like a lot more complex. So the, the, the, the DoVisual platform out there depends really on the contributions of the different users. In some use cases, it's relatively simple. You could use Claude or ChatGPT to generate a node. So you tell it, for example, here is the OpenAI API, I want you to create a node out of it, and it creates a node. You host it on Replit or on any other service integrated to DoVisual, and now you can use that node kind of like forever. Um, but the idea is to have a shared library where each one kind of contributes different nodes. So the least technical people, the ones that don't really want to bother about hosting or creating, they just find a library and whatever they are seeking, they can search for it. Hey, I'm looking for the cheapest LLM. Boom, you have a node, you drag and drop it. I'm looking for the, uh, fastest way to generate speech out of text, you find a node for it. You are looking for a connection between a very niche product and whatever you are building, potentially there's someone that faced the same issue and published a node on our platform for you to use. So the idea is to bring the power of the collective of the open source community and have the different specialized model providers, but also like engineers that you know, don't want to bother recreating the same components again, to just provide it as a node for you to use. I hope I, I was able to answer 3 different questions at once here.

Yes, thanks, thanks a lot, Aladin. We have like a question coming up from Kevin here, and he's asking how— what, what is your quote to state the quality of the platform?

I'm not sure what you mean by my quote to state the quality of the platform. Obviously, we are still in early stages. So, Memorality as it is today, we're not even 6 months old. So, we started around the 15th of January and we only announced it on the 1st of February. And obviously, like, a lot of our customers right now are like not going through the platform. For example, KSA Labs, which is a Saudi acceleration program that we are building with the Saudi government. So we obviously have to scale slowly. However, what we want to make sure is that we are doing is that whatever learnings we have with our customers, that we translate them to the platform. While the algorithm today, which is the orchestration between the different nodes, is very solid as we are using it internally for everything else as well, The nodes might not be that solid because they are hosted as community projects. Most of the nodes today are hosted using public API keys or public infrastructure. So I wouldn't really rely on building a fully-fledged product on it. But if you are looking to build an MVP or just validate an idea or just create an internal workflow like I have for myself to be able to use YouTube, transcriptions with Google Spreadsheet, for example, that is relatively easy to do and you can already do that today.

Cool. And could you maybe state like some of the most impactful applications that you built with using Memorality, if you could give us like more—

Yes. —details on that? Yes, no, absolutely. So one that I mentioned earlier is about creating medical reports. We are able to even integrate models, different models from our partners, from our community into it directly. One of the models being able to do dental vision. So it takes a doctor, for example, around about 30 minutes to go through your X-ray and find all the details that they need. Together with our partner, Decamet out of Saudi, they developed a model and we are helping them with the data around it. Is they are able in within less than 30 seconds to create a treatment plan, find caries and dental issues as well as bone density. Now they have a very specific use case for it, which is to integrate it directly with the software in hospitals. But what about potentially all the other use cases? So it is great to see them contributing their models and their work as nodes on our platform, which will help other people to come up with creative ideas. For example, maybe you want to build a consumer application to help people to understand blood analysis better or to validate. I mean, I assume most of us here are not dentists, right? So there is this huge trust that you need to have to go to a dentist and let him actually perform his work. The problem is that a lot of the dentists are incentivized through monetization to do bigger work. So it is really a trust-heavy situation. And if you're a software developer or if you're working in tech, there's this term which is trustless, trustless environment that everyone is like continuously using and loving. I would love to see that across multiple industries more. And I think, I like to think that our platform can help to kind of like democratize access to different technologies for people to come up with their own cool ideas to create a more trustless environment where we don't necessarily have to rely on the word of, you know, one doctor or of one person, but rather like can consult across, you know, the community, across machine learning models and so on. This being one example. The other one is when we announced Memoreality in February, I was presenting on stage at AI summit here in Tunisia, and I checked into my hotel at around 12:30, so after midnight, and I was quite hungry because I was flying in from Saudi. And I tried to call them to order some room service, and it was a terrible experience. Like nobody picked up the phone, and I got really frustrated. So quickly I hacked together a demo where we would build onstage Alexa in Arabic for hotels. And that ended up getting us a deal with TUI. So we are integrating with one hotel, today in Morocco, which has like 200 rooms. But as I mentioned earlier, TUI is managing worldwide over 200,000 rooms. The majority of our revenue is coming today from TUI just because I was able in the middle of the night to create a prototype to solve an issue that I had at the hotel, which, you know, some other people might just think, yeah, whatever, like then I'm just sleeping hungry and it is the way it is. But I ended up being able to create Alexa in Arabic for hotels in less than 15 minutes and I presented it the day later and I was able to land our first customer outside of approaching and going through the hassle of convincing them and so on by just being able like, look, this is Alexa in Arabic, we are building it now together in like 20 seconds because we have the notes ready. So let's make it happen and someone happened to work at TUI and be like, hey, we would love to integrate it. I like to think of that as a potentially quite impact as well. And if we can make it even easier, faster, and more reliable, maybe, you know, a 10-year-old in Tunisia will be able to create really a billion-dollar company. Like, I know Sam Altman and a bunch of others are constantly talking about that. We don't believe that LLMs will get there because of code complexity and hallucination. But if you create easier ways for people to build scalable stuff, with the power of the collective, which is a community, then we will truly might be able to see like some new conviction turning in from young people, old people all over the world.

That's a great story. That's a great story, Aladin. So I have a question myself for the like many startups, like they hear a lot, they hear a lot about this AI, it's a hot topic and they are hesitant to dive into it because of the lack of expertise or the non-technical background. And what advice would you like to give to these kind of startup founders and how can they maybe leverage Memoreality to bridge this gap?

Fantastic idea. So I think I briefly mentioned it earlier. Today we have the combined knowledge of humanity in our pockets. Like think of Who Becomes Millionaire, the TV show, and you get asked a question that you can't answer and you try to call someone and now it's like luck to, will that person even pick up? Will that person know the answer to my question? You know, like you can't kind of know for sure, but with these technologies you kind of have like someone incredibly smart with the combined knowledge of humanity that you can ask all sorts of questions. If you are building a startup today and you're not consulting with AI, you're doing something wrong. And programming and maybe like even some of the steps that I showed earlier. And trust me, I am trying to do it in the easiest way possible. Yet I understand some people even struggle to understand what JSON is. You can ask an AI. Like you should consult with an AI and if you are able to even have a surface level answer for your question, it will help you at least to communicate it easier to your team. How are you expecting to manage developers even if you have a co-founder and CTO? How are you expecting to be able to communicate with your CTO if you're not even able to create the simplest version of your prototype kind of yourself? A few years ago, I understand, pretty much impossible. Your strength is in business and business development and writing business plans and presentations, whatever, right? Like you're not an engineer, that's why you have a co-founder who's focused on engineering. But in today's world, you have to be a generalist. Like saying, no, I'm, my expertise is in something, you know, is simply not enough to really be a successful entrepreneur. And you have to fight for that. And I believe that, you know, being able to quickly build something yourself, validate it with your initial customer, with your initial community, whatever, it's going just to be such a benefit. Every single developer that is going to join your team and work on the scalable product will have a better understanding of what you're trying to achieve if you already were able to deliver it in a prototype. You will understand how much time it's taking them because you have struggled with it yourself. And again, you have some of the smartest piece of technology that ever existed in the palm of your hands that will be able to kind of like help you navigate through it. If you don't know what JSON is, ask an LLM. If you don't know how long your presentation should be, ask an LLM. Worst case, you just dismiss it as a stupid answer. Best case, it will open up millions, if not billions, of new doors for you to explore.

Yeah, definitely, that makes so much sense. Let's take one last question from the audience and then we can move to the networking lounge. I can see a lot of technical questions in the Q&A section, so Aladdin will maybe dedicate a few time for like few minutes from his time to the networking. So feel free to ask all these technical questions to him. I don't like to like put them on screen, honestly. So, um, yeah, so let's take this question from Daniel. He's asking, one of the challenges for AI products is the data privacy. Does Numerality comply with ACoC? I don't know, honestly, these, uh, these kind of, these keywords, uh, but they, I guess they are frameworks, considering it is still an early stage product.

Um, so as I mentioned earlier, right now all the nodes that exist on our platform are self-hosted by, by people, by the community, and so on and so forth. Uh, we have a disclaimer where we make clear, hey, this isn't for production, this is for you to use. We barely communicate the platform at all, um, and every content is shared across everyone. We're not monetizing it or anything behind it. The idea is to just have some people like play around with it, keep the community small for now, just to make sure that we don't run into any compliance issues because we are not hosting the nodes ourselves, at least not on the community version of the platform, which means that if you're sending your data to a service created by someone that you don't know and without any validation, I can't guarantee if they are like following strict laws such as GDPR. So I would always like create my own nodes and then connect them to DoVisual to make sure that my data kind of like stays with me. In the broader picture. Now, giving you guys a sneak peek of something that I'm extremely excited about that we are working on. We are working on something called an anonymity engine. The idea is that you have something built on top of the network protocol that uses a machine learning model to anonymize data. So basically, you know, hospitals, for example, they kind of always use legacy technologies. Because they require everything to be on-premise. But what if you can out of the box anonymize all of the data that leaves their premise? So there is something that we are working on and something that we are planning on integrating into DoVisual so that you can feel safe even when you use third-party nodes of community members that have not been validated. You know that your data gets like anonymized enough for their model never to really be able to understand you know, outside of the, the context, the key that you and only you have, uh, what that or any of that actually means, but yet still benefiting from the latest models and technologies.

Great, thanks, thanks Aladin for the session and for answering all the questions. Um, I think we should, it should be, we should close the session today. Um, but we will, the networking lounge will be open, uh, so me, Aladin, uh, we will all be each one of us will be at a table. So feel free to ask your questions and let's network there. Thanks, Eladin, for the presentation, for the workshop, for being here today. It was really insightful. Thanks everyone for being here. Tomorrow we have another event. It's about launching a US company. Feel free to register. We do host this kind of events on a weekly basis. So yeah, happy to have you all and have a good day.

Thank you so much, everyone. Thank you for having me, Silvia.

Thanks.