AIMinds #030 | Nicolas & Mathieu of NovaSquare Ltd

About this episode

Mathieu, a seasoned software engineer with 12 years of experience leading startups from pre-seed to Series B, is passionate about creating user-loved products. He's also a musician, playing guitar and saxophone in various bands. Mathieu has built online communities around his music and now streams his jam sessions on Twitch.

Nicolas is an experienced software engineer with 8 years in startups of various sizes. Formerly a rising French tennis player, he transitioned to computer science and developed a passion for esports, engaging in commentary, playing, and streaming on Twitch. Over the past decade, Nicolas has supported numerous community leaders in building and energizing their online communities.

Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.

In this episode of the AI Minds podcast, host Demetrios and guests Nicolas and Mathieu dive into Alicia, an AI platform that boosts community engagement for streamers and creators. Mathieu, co-founder of NovaSquare Ltd, shares his journey from navigating AI in the workforce to overcoming challenges while streaming his saxophone lessons on Twitch, highlighting the difficulty of keeping viewers engaged as a musician.

Nicolas, also a co-founder and experienced Twitch streamer since 2014, discusses the challenges of maintaining viewer engagement, especially in smaller communities. The episode explores Alicia's evolution from a basic chat tool to a sophisticated platform that mimics personalities, interacts proactively, and adjusts streaming environments like lighting. Originally designed for small to medium-sized communities, Alicia now also serves larger audiences as a community mascot.

The podcast delves into Alicia's technical progress and the broader implications of AI for online interactions and community building, offering a look into the future of AI-enhanced community management.

Fun Facts:

Mathieu decided to learn saxophone during the pandemic in 2020 and took to Twitch to stream his learning process, aiming to build an online community around learning new musical instruments.

Nicolas, an early adopter of Twitch, started streaming his gameplay back in 2014 when making a living from streaming was rare and reserved for the very top streamers.

Show Notes:

00:00 Built big data platform, learned English, AI.

05:24 Received saxophone, learned by ear, created community.

09:03 Struggling to engage audience, seeking advice.

12:35 Reduced streaming hours lead to AI replacement.

13:30 Building community with interactive voice and video.

17:58 Configure Alicia to match community with diverse characters.

20:51 Focus on small and medium streamers globally.

23:48 Expand community engagement across multiple platforms, including Discord.

More Quotes from Nicolas & Mathieu:

Transcript:

Demetrios:

Welcome back to another edition of the AI Minds podcast. This is a podcast where we explore the companies of tomorrow built AI. First, I am your host, Demetrios. And this episode, like every other episode, is brought to you by Deepgram, the number one voice API on the Internet today. Trusted by the world's top conversational AI leaders, startups and enterprises like Spotify, Twilio, NASA and Citibank. In this episode, we are joined by none other than the two co founders of Alicia, Nico and Mathieu. How you all doing today?

Mathieu:

I'm doing good. Thank you for having us.

Nicolas:

Doing great. Happy to be there.

Demetrios:

So I want to start this off by getting to know you both a little bit more because you have such a rich history and interesting background that has led you into the streaming world. Let's start with you, Mathieu. I know you were working in AI. You fell in love with coding at a very young age and that almost like led you into the AI field doing cool stuff. Stuff. But maybe you can tell us a bit about the different startups that you worked at when you got into the workforce.

Mathieu:

Yeah, sure. So yeah, I started to work for a startup and actually the first one was 2012, which was connected health. So how do you help, you know, taking care of some patients and making sure that you have the right information as a doctor when they are actually staying at home. So it was all about connected devices and how do you get this data and you show it in the right way to have some insight, like any problem dialecting at first, like if there is any issues, which was my first experience and I was in a very strong training with very hardcore developers and engineers there. So it was quite interesting. I would say it was the best score I could ever have. That was an interesting one. I moved then for short amount of time for a smart city company.

Mathieu:

We basically built a big data platform with some help from the DevOps team there where I could learn a lot more around servers, development, deployments, all of that which helped me later on to be able to build what we have today. Then I moved to the UK without being able to speak a word of English at the time. I was able to articulate a bit of the technical aspects, which is how I got the role. But when I started my first job, they explained to me, oh yeah, today for the first week, you will be doing this and this and this. And I was like, I have no idea what you just said, so you have to write it down. And when it was written, I could actually understand what I was doing, but I'm glad they didn't give up. He kept helping me. And that was the first real instance where I get very close to AI through the gunshot detection system that was analyzing audio to basically detect if it was a gunshot or just someone with a hammer hammering a metal bar, for example.

Mathieu:

That was super interesting.

Demetrios:

And so these were AI models on embedded systems, right?

Mathieu:

Exactly, yeah.

Demetrios:

And this is fascinating to me. What were the hardware components of this?

Mathieu:

Well, you're asking me a bit much. I'm not the electronic guy, so I can't really tell you much about the hardware itself.

Demetrios:

Let me rephrase that. Let me ask about what were the devices that the models were embedded on. Was it just a microphone that was put into random parts of the world? How did that actually look like in practice?

Mathieu:

So it was kind of a smog detector that you have in your home or something like this. They had built and designed a specific small computer which was running the model there with a small battery as well and connected to master. So it was installed directly into schools, unfortunately, stadium companies, banks, stuff like this. And it had so microphone in there, listening all the time, but just having a threshold about the level, because obviously a gunshot is, you know, quite a lot, quite loud. So as soon as you reach this level, then we pre recorded, or we recorded the last, like, 2 seconds, something like this. And then we passed through a pipeline that would analyze it to your way and then pass it to the model to detect if it was a gunshot or just, you know, big bang.

Demetrios:

Okay, so in 2020, you decided to learn saxophone. And not only did you want to learn saxophone, you said, I want to learn it with other people. So I'm going to stream myself doing that. But that ended up being a bit of a mistake. Why is that?

Mathieu:

Yeah, so actually. So I got gifted a saxophone in December 2019 and kind of sitted there for a while. And, you know what happened in 2020? I think everyone knows. So, yeah, at some point, I was just thinking that would be finally cool to get something else than just watching tv or drinking some wines or anything. So I went to grab the saxophone, and I was trying to learn some stuff by ears because I was playing some guitar before now. I thought it would be cool to do that directly on twitch and start creating a community, an online community of people interested on how to learn new instruments. Like this turned out to be much more challenging than I thought it would be. So I had to call a friend who might have a bit more experience than me on that field.

Demetrios:

Okay, well, this is a great point to bring you into the conversation. Nico, because you had experience with Twitch since 2014 am I, correct?

Nicolas:

Yeah. So I discovered the platform as a normal user around 2012, and I started watching whatever was happening there. It was pretty new. Like, I already had like watched some content, like video game related on YouTube before, but I didn't know people were doing that live. So I started looking at it that somewhere. I said, hey, maybe I could do that too. Maybe I could stream and show also my game plan, connect with other people, because I was already, you know, in some Skype groups back in the day with some friends or people I played with, and so already had this kind of community aspects in addition to the game. And Twitch felt like a natural step.

Nicolas:

Natural next step.

Demetrios:

You started streaming, you built up a community, you continued streaming, and this wasn't your full time job, though. And back in those days, I don't think there were people making a living off of streaming, am I correct?

Nicolas:

Very few. Very few. It was very difficult, like, you had to be among like the best of the best to really get something. And if you were just average, because I think at that time I had maybe 20 30 viewers on average when I was streaming, which by today's standards is still very big for Twitch, but by then it was just not big enough. You don't get any kind of monetization opportunities or anything. So just you do that for fun, which I was doing that for fun. I was not making it or trying to make it a job, really.

Demetrios:

And then what was the conversation that you two had when you got together and Mathieu was like, hey, I'm playing saxophone. I'm not sure how many people like it. Can you help me build up a community around it?

Mathieu:

So first I was quite sure how many people liked it was near zero. So I think there is an element to state specifically on Twitch is that 97% of the people streaming on the platform have less than five viewers in average. So as soon as you start having one, two, three viewers in average, you start being close to the top 3% and you get closer and closer. And like Nico was saying, like 20 viewer average. I reached that point recently and went back to less. And it is actually top 1.5%. So it's really high. It's a bit tricky.

Mathieu:

So the first element, I think when I started or when I thought about the project, I talked to Nico and I said like, yeah, I'm most probably going to do that. I was like, oh, yeah, good luck. It was kind of normal, like, yeah, okay. And so I spent a few hours basically streaming and, like, no one was there, no one was chatting. And then. So I basically, after, I think, two weeks like this, two or three weeks, I reached out to Nico and said, hey, actually, I must be doing something wrong, or there is, like, something happening, because I don't know what it is, but is really crap experience. I can't interact with people even if there is someone nice. I'm usually playing saxophone when they arrive, so I'm not interacting with them, so they're not staying.

Mathieu:

And so, yeah, I was just. Yeah, it was a weird thing. I was doing it more like, okay, I know I have a set of 2 hours where I'll be streaming and so I can practice saxophone. But in the same time, I really wanted to build this community. So we chatted about it, and actually, Nico offered a different kind of help. I might let you explain, Nico.

Nicolas:

Yeah. So basically, what Mathieu kind of wanted was some kind of co host in the chat. So have someone that would interact with the new newcomers, interact with people already there. And so, for example, he wanted to jam on music. So one of the key elements for that is to get people to suggest songs that Mathieu could then try to play on or butcher, depending on the song. And so I would be there and encouraging people, hey, is there any song you would like to be playing next? And then making sure there is a list always populated so he has music he can play, and there is kind of no real down time in the stream. Rhythm is very important, as in music. It made sense to have that, in a way.

Demetrios:

It was like you were trying to engage people.

Nicolas:

Yes, exactly. Yes. There's really making sure engagement is always at a high level.

Demetrios:

And now you all took this idea, you saw something there and decided to create a product out of it or try and productize that experience of the engagement.

Mathieu:

Yeah. So it went through multiple processes. Nico. As soon as Nico joined, the difference was really like day and night or night and day. I started. So, first of all, just for myself, knowing that someone was in chat where I could actually interact directly with him, say, like, hey, what do you think about that? Should I play this song? Blah, blah, blah. He would basically interact in the response. So it helped me really keep the right energy level through the stream and help me to keep speaking.

Mathieu:

That's also a big challenge when you're streaming for the first time, is being able to always interact and entertain during your. Your first hours. And he was there to welcome the new and the first viewers to just welcome them and suggest that they would interact, so they would stick around. So I started to have my average viewers higher, but also they were interacting a lot more directly in chats. And that really triggered that getting affiliated, like having some of those streamers showing around. So started to kind of build a community between people interested in music, interested in saxophone, interesting in learning, but also other music streamers. So that helped a lot through the first two years. So basically, from 2020 to 2022.

Mathieu:

And what happened is that in May 2022, I was streaming about like 15, 20 hours a week, which you might guess when it's not a paid job, if you have to stay in front of your computer for that amount of hours to watch someone play the same song that you've heard many, many times, it is not as enjoyable as it might be. So Nico asked me if he could basically just step away a bit rather than doing that. We kind of laughed around and I said like, yeah, okay, that's fine, I'm just going to replace you with an AI. And that was kind of the trigger from there. So May 2022, I think, you know, the landscape, but all the Chan GPT element didn't really. Wasn't really something. So I had to find ways to do that. But I had the first version of Alicia running on my channel, only interacting through messages at the time.

Mathieu:

And that's when we started kind of to shape it up, because we also had a few streamers from my community starting to ask questions about what was this and how it would work, what was the purpose about it. And for me, with undeterminated at the beginning was more like, I want that interaction, like I want. Because that was kind of the key element to be able to build a community. And I didn't want to lose it. And that was something we kept working on and just realized that, no, actually it is much bigger than that. It's something completely different. It's not just having a chat, chat interaction in your stream. It has like, voice interaction, because you can have this cohost animation part, can have integration not only for music, but for example, we have integration with video games like Microsoft Simulator, which is going to interact directly.

Mathieu:

And was one of the first tests that I've done in August 2022, about that again a while back. And yeah, it started to become like the key point for us to, if you want to build a community, having a cohost is basically one of the best solution. And this is already what people do. Like, when you build a community, you will have volunteer community member that will step up, help you interact with people on your discord or on your Twitch channel or any other place, and that is the kind of element or your friends like Nico, and that's the kind of element that we got inspired by. And see how Alicia was working on my channel plus the other channels, we started to shape it up in that instance.

Demetrios:

Okay, so this was pre chat GPT, what and how did you build that first iteration?

Mathieu:

With a very more expensive model. It was still some of the elements with OpenAI we had to get approved at the time. So I've spent multiple hours chatting with them to get approval because they didn't want basically anything that would be public surprise. But twitch is public, so it's kind of something we had some issues with. And also at the time, it was basically just impossible to roll like a proper financial model that would work at large scale. That was something where at the time to basically be running Alicia for 10 average streamers. So I'm nothing talking about a very, you know, 8 hours a day, all day, all week, et cetera. It's more like normal streamers, you know, 10 hours maybe per week.

Mathieu:

It would cost us more than $250, closer to 300 for more, for just ten. And so financially it's not, it's not viable. What we are doing is something for smaller streamers, for new streamers. It's not something that basically people will have 60 quid or more to spend per month on this. It's something that should be available for most of the streamers and content creators and community leaders that are doing that. And it has to be on an affordable way.

Demetrios:

And so how has the platform evolved since the inception?

Mathieu:

A lot have changed.

Nicolas: First iteration of the platform, it was like, so I would say there are two components. There is the twitch component, where there is just Alicia joins the chat and then she's interacting. So there is no platform for that aspect. But then there is the kind of dashboard where they can control personality, where they could get their voice listened by Alicia, so that could have this stronger interaction. And so it started with just that, you know, one page with the login button in the same level as everything else. And then, yeah, it was one field to write, you know, if she would be nice or sassy or whatever people had in mind, and one button to start listening. It was only available for English. That was, and that was that for quite some time.

Mathieu:

Until April, most probably April 2023, something like this.

Demetrios:

And explain to me now, what does it look like?

Mathieu:

So now it's a full fledged dashboard. So we basically have a very, like, I would say, even sometimes a bit too complex way to be able to configure Alicia on the way that will match your community as close as possible. So with a lot of different settings that you can interact with. The main idea for us is creating characters. So it's basically you think about your community. Who do you want your community to be joined by? So it can be. For me, I love, like, the grumpy, grumpy grandma. So it's kind of an old person joining in and just, you know, complaining about paint and the fact that sometimes I'm playing electron music and that she hates it, or just offering, like, weird cookies when I'm just, like, saying, I'm hungry, or do you want a cookie? A cookie with, I don't know, some weird stuff in it.

Mathieu:

And so we have, like, integrate interaction through that. It's also something where now she's interacting with the environment as well as a streamer. You have multiple times, you know, connected lights or different commons that would do stuff in my stream. She is connected with the lights. So I would, you know, I'm streaming and I'm saying like, hey, Alicia, can you, can you just change the lights to green? And she'd be like, I think you can do it yourself. I'm like, yeah, nice. That's a sassy, very grumpy grandma doing that. And then after, she would just send a few messages and just after changing to red and saying, do you like it that way? I'm like, so it has this very strong personality that we are trying to really build on.

Mathieu:

And so all of that can be configured properly through the dashboard. And that was kind of the main element. But on top of that, we added voice volatia. So now you control, like, many different voice, more natural and human. One more basics. It depends on the plan that you are on. It's something that you really have a proper voice to voice interaction. So she can hear you, but you can hear her.

Mathieu:

You also have the fact that she can now build a long lasting memory, so creating more relation with the people, because again, it's about community. So if someone comes in your community starting to say, hey, next week, I'm going to have an interview in a week time, and they come back to your chat, she's going to ask, how was your interview, by the way? Or if she's very grumpy or sarcastic, she might just say, hope you failed your interview, because again, it depends which personality you'd like to have in your community.

Demetrios:

And so you mentioned that when you were first starting out, this was a tool for streamers with very small communities to help them grow and nurture their communities. Has it adapted who you are now having as people that use alicia?

Mathieu:

Yeah, so it's a, it's an interesting element where we. So our focus, and it's still, it's on small and medium streamers. So I'd say, you know, below, like 2025 viewers as kind of the target we have. However, we have many streamers with more like, you know, hundred, 100, 5300 thousand of viewers. And we never really thought that Alish here would be a good fit in those environment because there is already a big activity, a lot of people talking, and actually people loves it. It kind of becomes the mascots of the community, and people are interacting with her all the time, and so it creates gimmicks within the community, and so the jokes about it, and so it helps, again, strengthen the community together about that mascots and what the stream is about, all the, the community is about. So it definitely kind of shifted away on some aspects, but our focus is still towards, like, more small and medium, medium content creator or community more globally.

Demetrios:

And does she have different names depending on the community that she's in?

Mathieu:

Yeah. So now it's an option. Hasn't been for a long time. I think for a year we restrained ourselves not to do it because we needed to have a visibility of Alicia, so we didn't want to change the name for her. But now it's a possibility. And it goes along with some work that we are currently doing about reworking the profile and the settings for Alicia to now create characters. And a character basically has a personality, a backstory, a name and a voice. And this character can be shared between streamers.

Mathieu:

So if you're in a community of streamers, you can have the same character joining all of that, sharing the same memories. So that's kind of the idea, yeah.

Demetrios:

And I could see how in the beginning, your assumption would be, this isn't going to be for communities that are already well established, that have a lot of people in them because they may have someone that's doing this.

Mathieu:

Exactly.

Demetrios:

Or the community itself does this, so you don't need it. But when you talk about the capabilities that it has and if it can tap into your light source or it can be a little bit snarky, then that sounds like fun. And you can always count on, like, Alicia to respond to you, maybe anyone else that's out there, they've gone and started making breakfast. And so they're not. They're logged on the stream, but they're not actually at the computer. Right, but Alicia is going to be there whenever you call.

Mathieu:

Yeah, exactly. It says she's all the way there as soon as you start streaming.

Demetrios:

And now what is next when you think about where you're going with this?

Mathieu:

So for us, it's basically, and weirdly, in a sense, but getting away off twitch and live streaming communities, or not getting away but expanding out, that's most probably the right phrasing. The idea is to go on discord and start implementing the cross platform elements. Like you meet someone on twitch to join the discord and you remember, you still keep the conversation and the knowledge about this person directly on Discord. So to keep them engaged with the community even when the live is off. So that's the idea. We also have a lot of discussion about more brand communities where we have companies that are managing and creating communities about what they do, the same way that we have a community around Alicia on discord. And so they need ways to keep the community engaged and active. So the idea would be having Alicia to help them do that directly and so on, that you basically have to think about the number of platforms.

Mathieu:

So if we talk about just live streaming community, you obviously have YouTube, Kik, TikTok, Instagram, which are the platform that we will be looking into integrating at some point. And if you look into more management or community management or offline communities where you want to call it for Discord, it can be discord, Slack teams, Facebook groups, WhatsApp, all of that. So the idea for us is going to be around expanding those different platforms for Alicia to be present in all of that and again in a very continuous experience. So if you're on Twitch and discord or Slack or WhatsApp, you should have or have the ability to have an Alicia, which is going to be consistent in between the platforms.

Demetrios:

Incredible. Well, guys, I'm super excited by what you're doing as a community man myself, it is fascinating to think about how you've created this, how you are helping others get engagement and nurture their communities. So I love what you're doing.

Mathieu:

Thank you. Thank you. I appreciate that.