Flora Tasse on The Robot Brains Season 2 Episode 9
Transcript has been edited for clarity with several areas unable to be recorded
Pieter Abbeel: This week, we welcome a guest who has been on a remarkable journey. Flora Tasse grew up in Cameroon. She completed her bachelor's in Mathematics at University of Buea, received a second bachelor's and a master's in computer science at Rhodes and Cape Town in South Africa. She then earned her Ph.D. in 3D Computer Vision at Cambridge in the UK. Since then, she has co-hosted several of the Black in A.I. workshops at NeurIPS, filed a few patents and likely most importantly, founded her own company and later sold it. She is now leading AR research at Streem, the company that acquired her first startup, Selerio. Streem offers A.I. and AR powered communications, bringing artificial intelligence to the front lines of customer service. Welcome to the show, Flora. We're so excited to have you here with us.
Flora Tasse: Thank you. Thank you. It's such a pleasure to be here. I'm very excited about this.
Pieter: So, of course, I'm really curious about your startup Selerio and you know the journey there, but I want to start a bit earlier than that. You grew up in Cameroon and started your university journey in mathematics. How do you know how you choose mathematics and how did things evolve from there into computer science?
Flora: And I think, yeah, that's let's go way back. So I'm actually from a very young age. I knew that I wanted to do graphics, but there was no computer grapics. And more importantly, I want to do research and graphics. And so that meant a path towards computer science. But then there was no bachelor in computer science, and I studied English in my country at the time. So if I'm going to come to science, the next best thing was for mathematics. And so that's how I go to University of Buea and I do mathematics. And then I think a few courses in computer science and for me, I just found it was just easy to accomplish things, but with a plan that I will do things after. So that was how I got into mathematics.
Pieter: But then you have to move countries to switch to computer science. How did that feel?
Flora: Yeah. So it's yeah, I think it was a 10 year plan from the time I decided. So I watched Jurassic Park, the movie and I was like, Wow, how they're making these dinosaurs kind of, you know, just be in that environment. I want to do that. And then realizing that there's no way I can do that in my country, right? And so there was like a 10 year plan for me to study many things, then go abroad, then do these tests and ultimately to my computer science, hopefully in the studio graphics. And so, yeah, it was just I just knew that I was going to have to travel outside. It was just a matter of where would I go, who had the best program in computer science. And at the time, we decided that we had a very strong year having strong programs in computer science. They were doing great work in graphics, some of the best work in Africa. And so I just kind of decided that that's where I wanted to be. The. So emotional then, yes. Yes, I did. So I did my masters at Cape Town after my stay at university and then I just got to a point where I wanted to go to a university that had a bigger, a bigger research group in other countries and that was Cambridge University. So you were just, you know, thinking, what is the next step? Where can I go from there? What's the next big thing I could do? And yet Cambridge finds that they can move to the UK.
Pieter: Now it's interesting you really wanted to get into graphics inspired by watching Jurassic Park, a movie I also watched a long time ago. And I really was, you know, impressed by how realistic it was all made to look. You wanted to get into computer graphics, but actually you ended up in computer vision, which is the reverse problem of computer graphics. And yeah, I'm curious, how did that happen?
Flora: So, OK, so I wanted to make pretty pictures. That's how the whole thing started. And then realizing that I had no artistic skills. So how was I going to make these two models if I couldn't draw, you know so that turned into the problem of how do you make it easier for people to create content, which is one of the most serious parts of the whole graphics pipeline. And so and so realizing that, well, we have objects under our laws, why can't we just digitize them and then go from a picture of something in space to a digitized graduated and then insert into a 3D experience? And so that's how you then they start saying, OK, I need to then take images and then turn them into graphics into the contents of. And once I have that, you can then create the pretty pictures of whatever application is needed. So that's how I started studying Graphics Garden with 2D vision, and then I seen its signature.
Pieter: But still with graphics in mind, it seems still with creation of visual artifacts in mind.
Flora: Yes. So I'm now, I'm in augmented reality, and we'll talk about that later. But I'm now in the augmented reality space, which runs with this because you get to work on this full set of spectrum.
Pieter: Yeah. One thing anecdotally that's kind of interesting is one of our other guests. You might, might know him, Sergey Levine. He also started wanting to do graphics and ended up in AI. It seems like there's, you know, a pattern here of among our guests that they might start out with wanting to do graphics, but ending up in AI. And now when when we look at your Ph.D., there's one one. More specifically, I'm pretty excited about which is 3D Computer Vision, which I think captures some of the ideas that have become pretty common in a lot of AI which is turning something into a vector, right? And then that vector becomes easier to work with. So can you say maybe a little bit about, you know, your motivation behind the shift effect work, but also then give a little bit of an explanation of what does it do and what can you do with it?
Flora So yeah, that was what I doing in 2016 and the issue I probably were trying to solve was that we wanted to take. We wanted to create what if we get 3D content, but using images or using hand-drawn sketches? And so and then you said, Well, we have all of these big repositories of 3D models that are already out there. Why can't we just search for what you want using an image as a query, for instance? And so now you want to use something in one space to search for something in a different space so you have different modalities. And then the question was, how do we bridge the gap from natural images, all of the sketches to 3D to the objects? And so that means taking a shape and then turning that into a vector. And so you could take a 3D object intelligently vector and also take an image and turn that into a vector. The idea is that all of these vectors actually live in the same space. And so because they live in the same space, you can then use. …[Unintelligible}...So that was really inspired by what you read and and the way to do a lot of work at the time around doing this for images and that was really amazing. It was really it was really popular and then they wanted to do it. You said that those ideas and then apply them to the ships and then you have this thing where you can use you have words, images, 3D shapes, all live in the same space and you can use search one. What use anything as a query to search any of the ships. So we kind of just want to set in kind of what the problem was and what was the inspiration behind a solution. And then the way that we actually created ships should do well was to use semantics as a way of tying up all of these different languages. Because if you think of an image of a chair and then a 3D shape of a chad, what we did things together is that semantic meaning. And if you could create vectors that can capture the semantic meaning, then you are able to really bridge the gap bridged the gap between these different words. And so should do that is really semantically a way descriptors of. Yeah. So that's how you did what
Pieter: You said earlier you had trouble maybe drawing things too. You know, that you found, you know, artistic enough to be happy with. And so you want the computer to help write. And is it fair to say that with shape, the fact that you don't maybe don't need to draw that accurately, you can just get something out, even if it's a poor drawing, and it might find and retrieve something else that's much higher quality that enhance what you're looking for.
Flora: Exactly, exactly. That was the goal. And and I think at the time we compared to any method that was out there, I think we would something like that. They are we kind of jump from I don't remember the exact numbers, but we did this well. We added 30 percent more accuracy compared to the previous work of the council using hand-drawn sketches to search 2D shapes. So it was like a big gap and it was super exciting. So we're just really well, we're doing the experiments. You couldn't believe it's like, Is this true? Like, is is this working at this resort real? And you have to run it like multiple times
Pieter: And then make sure you were in testing on the training data?
Flora Yeah, exactly. So it was really it was really fun. I think it's one of my best work.
Pieter: Yeah, it's absolutely amazing work. And now if we can talk a little bit deeper here, maybe one more level. Our audience is at this point, quite familiar with the notion that you can train a large neural network if you have an input and output, and a network will, from enough data, learn the pattern to go from input to output. And can you maybe describe it in your setting the shape of X setting? How is that set up? What's the input on the neural network and what is this supposed to output to train?
Flora: So the idea here was we wanted to create semantically or semantic AI were descriptors. So let's say for hand-drawn sketches, the input will be an image of a hand-drawn sketch. And then the output of this for the trip, for the for the machine, for the intended word will be the word chair is if your hands. If your sketch was a chair, then you want to train a network to give you a vector that correspond to the work chair. And so by doing that process, you are basically encoding you are hand-drawn sketches into descriptors that have some sort of semantic meaning that is that kind of explain what is in that image. So you do that for the word of sketches and you repeat that process for the world of images, natural images. You did that for 3D shapes. In the case of two déchets, what we do because they're not two dimensional, they're actually in 3-D dimension. What you can do is then you take multiple random renderings of the shape. So now you have netted 12 images for your 2D shape and you go through that same process where you are encoding them into a semantic away descriptor. And so because you've done that for each of those modalities, you now have a descriptor for any anything that you add. So you have these ensembles of of train networks, and you can then use that at any given moment in time to contradict this kicked off for a given query. So that's how the mission that came into.
Pieter: It sounds like they're in some sense the labels that you're relying upon is that the images and shapes you retrieving from the internet to train on might have some annotation with them in mind already, say, chair or a table, or maybe something even more detailed, allowing you to have it desired output that the new network is supposed to generate.
Flora Yes. And so. And so that's how you can kind of. Because it will be really difficult to put all of them in the same training in the same training process just because they are very, very different type of objects. I mean, back at the time, now these days you can do much more complex things that they use nuclear missile networks, words you can do more interesting things. But at the time, it was one way of reading between two different types of input by using different networks for each of them.
Pieter: And so you trained a separate network for processing 3D compared to a network you use for processing a sketch versus another network to process an image. But it all needs to lead to the same embedding space where things that are related are close together. Yeah, that's really cool. Yeah. I mean, some of the latest results now, five years later. I mean, you must be very excited about those because it's kind of sometimes a continuation on your work, like Opening Eyes clip, which is embedding language and images into the same space as in many ways, very, very similar to what you're doing there with with the ship to back.
Flora Yes. I mean, they seem to have seen since that is a work, I kind of evolve into more 3D augmented reality work. So I'm not I'm not sure we're up to speed with the work that's happening around around. That's basically bad. Yeah, I've seen amazing demos of life. It's it's kind of like it's how the space has moved. It's crazy. And so you look at work from like two years ago and it's it's like, it's like, this was 20 years ago, like it has moved that fast. So it's quite amazing what I'm seeing now in that space and the new datasets that are coming up. It's because I know how hard it is to create a dataset, so I just have my respect for these teams. That would be good at, I said, to really push to really push the field. So it's amazing.
Pieter: And so you're working on your Ph.D. and a lot of people on this list. And also, you know, when you did your peers a few years ago when they work on their Ph.D., most people with a PhD end up at a big company. These days, some of them become faculty. But you, you decided to do yet something else, like how did you converge on deciding to start your own company?
Flora It was not an easy decision because all my friends, when all my colleagues, when I worked for the companies Google, Microsoft Research. At the time I had just finished my internships at the HoloLens team at Microsoft Research, and they were doing amazing superchargers and things. So obviously there was a temptation to go back and work there. So, so but on the other side, I kind of also was thinking, I've been thinking about commercial entrepreneurship for a long time. There was a lot of activities happening around the air space and entrepreneurship. And and I feel like it was now or never. If I wanted to ever do my own thing, then I had to do it now because this was it was the ideal time for me because you're going to you're going to do the job. It's it's difficult once you kind of settled somewhere to then make the big change. So it was a combination of, you know, seeing that it was now a never or and also off of the work to make work than just talk of talk about like the result was so amazing. I was like, OK, do I really want to stop there? I feel like, you know that there's more to explore, especially from an entrepreneurial viewpoint, and I want to try that. So I did. I try to.
Pieter: Yeah. So what did you start building?
Flora: So the the main question after today we will say, okay, we could take is guess you could take an image or you can find a kids ship for you. Can you then take that concept and extend? That just seems pretty since. So you're not just looking at one object, you're not looking at it, his face like, see your living room. And so could you then some an image, maybe a video reconstruct and digitize your whole living room, and then you could use that for different applications. And so Selerio was basically our motto was bridging the gap between what's real and what's virtual. And it was this whole idea of digitizing everything around you so that you can augmented with with virtual experiences before. So that was itself when you said yes.
Pieter: Now that's a very technological. I'd say starting point where there's this new opportunity, something that wasn't possible before to turn reality into a digital counterpart thanks to 3D vision that you're leading the way in that scenario. Did you also have some specific applications in mind or were you thinking of providing this as a tool for others to build applications on top?
Flora So we we had a few applications in mind, mostly around augmented reality. And there were also some applications in VR. But really, what we saw serve as a platform on which people will build their experiences and ideas that we can take all this research out of these very technical bits and hide them behind API behind a platform to directly to empower anyone was in augmented reality in the visual arts space to kind of build on top of that and just benefit from all of these advances without having to necessarily understand what's happening in the backend. And so, yeah, today was very much like an R&D start up, very, very tech driven. And then what we do is that we would be partnering with various companies to basically see what type of application for best leverage to detect that we had to.
Pieter: And what are some specific applications that you started building or started partnering on?
Flora Oh, yeah. So I think that there were things like, I mean, there are some companies that were looking at it from me, some lenses like Google Lens. Think of places. Where did you go? Let's see what you what you are doing with face filters. You could do that with your space. You could because you had context, because they had context away. Are you could kind of trigger some experiences based on where your camera was shooting at. So one of the pictures was really intact in men and filters. But status for your space? Let me for your face. And so that was one and then the one that we also found very, very interesting that the males to where we are today was a more utility. Driven. Application way, because we know what you are looking at can get such a success. We can tell you how to use it so that that was a much more interesting one.
Pieter: But it's like us now. When you decided to start Selerio, how did you get the team together to start the company?
Flora So at the time my co-founder was at Amazon in a really good job so I started talking to him. He's also my brother. So it was quite easy to say, Oh, I have this idea, what do you think? We have no idea how it's going to work. There's no funding. So we just leave that job and it's going to get this thing out.
Pieter: Did I hear correctly that your brother is one of your co-founders?
Flora Exactly. That's so exciting. I mean, I did tell him I couldn't think of anyone else who had kind of the expertise that he had. I was coming from an academic background. He had an industry experience. You had to build put up a sample of the customers. And obviously, because I've known him all my life, there was a trust there, which I think is one of the key things when you are starting a company is picking your co-founder is is that trust is having that choice because sometimes, you know, startups fail because the founding team.
Pieter: Now you said, Okay, why don't you quit your job? And why don't we start this company together? And you had no money for the company yet? What did he say?
Flora: I had been hinting at it for quite some time now. And then, yeah, this guy, I tried to try to convince him. But to be fair, he was. It was that thing where you tried to convince the other person and then when? And that's OK. We like, I'm not saying we should do that. And then he convinces you. And then so at any point that if one person was like, I'm not sure that this was like this, let's do it. Let's do this. The idea is amazing. The tech is awesome. Our team is very strong. They just know this. There's no harm in trying to give this one year Rhug estate when you have your life, did it get to this end and you see where it goes? So yeah, he was very much in the he, I think without his expertise and his experience and the way we are to do so, I was very lucky that he said yes.
Pieter: Now that's such a beautiful story. Is there a moment that is the moment where it was just like, OK, we're doing this?
Flora: Well, because we had no money, we had to kind of figure out how you're going to feel around for those first two months. So I had to figure out how I was going to pay rent. So we applied for this accelerator called Entrepreneur First and then. And so there were very different from that attitude at the time because they invested in the team, and this is really the idea. And so you can just come to them and say We are. This is us. We kind of want to work in this space for after you. I'm not sure you know exactly what the code is going to be, who the economy is going to be, but just give us money to figure this thing out. And so that was really good. So they accepted us. And so that was kind of the point where it felt like it was validation. And we kind of knew that at least for the next six months, we will have enough money to pay rent. And so I think that was the moment when they said, Oh yes, you're going to a program like this, it will get you really doing this. This is for real.
Pieter: So then you dove in and you start building and at some point. You get acquired. How did that come about?
Flora: So it was about two years and a half after we officially started. And so in that space of time we got really sick of those programs. I did some fundraising, hired our first employees. I think there were times where we just found no matter what happens, we are just so grateful to have had this thing because the whole team was just so dedicated. All the ups and downs. And then towards the end of twenty eighteen, we have we have finally, we have our first we have a better version of this city. So some of them are dead. They can really use it. It's sad moving things around. So that was an exciting time. And then one of the people who we're trying our testing, our bid, our stream, and so they the associated for a cause like this is exciting this job when you call in this list of love. And so that's how we start talking about the application we've maybe actually talked about Streem does. So searching was really one of the first are companies or media companies that we're doing are interactive video. So think of your typical Skype, but we are on top of it. So the idea was that you could enhance customer service by using it. Are and how people think it was like fixing that fridge, learning how to get started with their washing machine and just came in if you have video, but also you have augmented reality tools to kind of guide you around the treaty space. So they were looking at our tech as a way of recognizing what the camera was looking at and then just activating the right excuse for for, for, for that machine. And so when they saw the API, they were super excited. And so the color, the colors and then they ask, OK, is that because we are also looking at what is the right application for these? How do we increase the impact of our work? How do we kind of go fast into this? And so it just felt just right that they had the right application and they said I had the right technology that really kind of increase the impact and accelerate our growth.
Pieter: So then they approached you and they did acquire your company? Right? I mean, yeah, as you said, it's exciting because you can have direct impact and you can see the past impact. But at the same time, you're also giving up the ownership of the thing you've been building. I mean, it must have been I mean, must not have been an easy decision.
Flora It was a lot of discussions and this was with everyone, even the employees around whether it does. Treme was so it was just a matter of, you know, getting up quite a bit. What is this Iraq company for us is a cultural right? Do we feel like what the mission of what we are doing? Do we still continue that building stream? And we did support that. So there was a there were a lot of questions that just just will this keep us happy because you don't want to get acquired and then just lose yourself in the process. So we and so strange tune is amazing. The Stream founding team, the I think because I think we're talking up to other companies wasn't just a dream that came to us and say, Oh, maybe interested in the like, is mission bad? You feel extreme, had the right, just the right culture and also the right vision of augmented reality. I mean, for us, it was all about being really blurring these gaps where you feel you can easily move on really to the virtual world without having any disconnect and where things feel, things feel magical, almost that it's magic. And so we wanted to keep fueling the tech. And Jim was basically was was odd in that. And so you can view the deck knowing that you could test that deck in the applications that were put out there. And that's it. It was it was still a bittersweet moment, right? Because this is almost like you hated being in that kind of. To be fair, you will never you will no longer be in control. And so it wasn't easy. It wasn't easy letting go. I don't think that I think looking back was just the perfect decision. It was it was the right decision to make. So so what we're trying to do?
Pieter: Well, congratulations. So it's actually checking out some of the demos on the Streem website. I mean, you're heading up computer vision there. You're heading up, essentially. You know, the special sauce that that's being built is what you're heading up right now. So it's checking it out. And I saw in the demo that essentially you can use it on a phone. You don't need anything but your phone and you don't even need to install an app. You can just in-browser run it and it accesses your camera as I understand it, and it streams. So yeah, you're on a video call with somebody on the other side, but the other side can actually start annotating things in your screen. And so it's like you're holding up your phone and somebody can can point out like, Oh, you know, this wire is loose over here. You just need to plug that back in. Or maybe, you know, you need to press this button over here that, you know, to test this can still work. And I thought it was really, really interesting because it feels like whenever I'm on a support call normally. Well, first of all, I feel like I'm usually talking to a machine. A pro am usually talking to a machine most of the time until finally somebody comes about. But even then, you know, the feedback doesn't tend to be very specific to the situation I'm in. It tends to be a very generic script, even when it's a person of like, OK, have you unplug? Can you plug? Have you done this? Have you done that as opposed to them just seeing the setup and being able to point out the things I'm supposed to do? And so I'm really curious as Streem rolls out this application of of the 3D AI technology. Are there some specific specific scenarios that are the most common use cases that you see a lot and that are really exciting to you?
Flora So so I love how you kind of explain what the experience is like. It's it's really I think it's magical, like anyone will use a stream for the first time, I was like, wow, because things are just so flawless. And so what happens is that, well, first stream vision, our mission is to make the world expertize accessible to everyone. And so understanding that you have these experts around the world that are not necessarily, you know, that can easily come into us based every single time. So how do you assume that expertize remotely? And so do you have me the consumer? I have my phone. I don't need to install anything. So first, that's that's that's amazing. I don't have this idea on my phone. I can call an expert. Let's see one of one of our because of based right. And they use that for their customers. And so I can call an expert and say, Well, I had this issue in my home. Can you help me figure out what's happening? And so you just have to turn on your camera? And so it comes back. So it says you just make a call like a normal call. There's a bit of money that happens that the video turns on. And then the expert can say, Well, I see this issue here. Can you go there in the button is going and what are you looking at? You are able to see actually able to say, Well, let me measure this for you, because maybe to ship in some basement platform for, let's say, a fridge or something else. And they. The idea is that then when if they ever need to come into space physically, they have all this data available to them do exactly what's wrong, how to fix it. And it's just, you know, it just means that it's less expensive for the expert because it don't have to come in physically and wait just more convenient, right? Because you can access that expertize at any point in time along that journey. And so we have a lot of customers besides one of them trigger if you use any barbecue grills. It is a lot to figure out what's wrong when it's not working. Some people to use it for retail has oven and things like that, but I really think that customer service and just troubleshooting is what is a big pain point for customers. And that's one of the big things that I think is really transforming is this interactive business.
Pieter: This is really impressive and exciting. Now when when you zoom out for a moment and you think about the future upstream, if you're able to share, I don't know what are some of the things that you are hoping to build in the near future?
Flora: So we are all about the whole so upstream. So we're focusing on the home, what's happening inside. And for now, you have you have things like being able to reconstruct a space we can understand and measure things we can help and figure out what is the right man for for some items to the home or maybe figure out, you know, how do have you with expert and expertize help them stick something? But for me, I think the next, the next, the big thing for us is basically what happens when you actually can do that is where teams that talking to someone necessarily. So the idea is that because we now have so I'm thinking of like what we are to us because now you have so much data around what is very wrong with this, this item is really fixing it. Then you now giving that information to have experts in a sense that can help you without having to connect you to someone, to someone ask, how do you figure out what's wrong with what you are looking at? And then someone on the journey connecting with an expert, I can get involved more help. And so this is something that is kind of in the works around expertize, but focus on the home, focus on how do we best support our customers with the that. And so that was just let's just do this.
Pieter: This would be a bit extreme, but I can just imagine, like the app looking at me as I'm trying to build some, let's say, IKEA furniture, and I'll just say, Oh, you're picking up the wrong part for the next step. It's, you know, you got the wrong part. Or I can even look in the app, highlight the part to pick up and things like that or how to install something. Mm hmm.
Flora So I think, yeah, I think that's kind of like when of magical experiences, how do we yeah, how do we just make things easier for customers? It's for me, that's one of the key things that you said, do it because then it becomes more interactive. You don't necessarily have to wait for someone to be available. And it doesn't discount the fact that. You can get help if you need to, but you don't have to.
Pieter: Now if we zoom out even more, one of the biggest things that's happened not necessarily in AR but in VR, which is still a related space, is that Facebook has decided to prioritize VR with their metaverse and complete renaming. Right? I'm curious, is that influencing some of what you're thinking about doing next? And yeah, just generally, what's your take on that?
Flora So, yeah, so they did. [Unintelligible}
Pieter: Yeah, I mean, I personally prefer to be in the real world, but I don't. Yeah, I don't know. I mean, some people love playing video games, and it seems like people who love playing video games will likely are not unlikely to enjoy being in the virtual world.
Flora: I do think that one of the main applications of VR is gaming. And I've done gaming. The AR is just and competitive gaming idea versus gaming are the AR wins every single time. And so because gaming is one of these spaces where really you want to be immersive experience and the AI just makes sense for that. So I think the Metaverse is here to stay. We don't know what is going to be applicable to every single day of your life. I don't think I don't think about it.
Pieter: Yeah, we'll have to see. I'm not. I'm not expecting to spend too much time in VR myself, but you never know. I mean, things always change. It's hard to predict when actually on one of the things. If you go to an AI conference, I'm sure you experienced it during many times since. It's that it's pretty obvious diversity is lacking in many, many axes. Gender diversity, racial diversity. I mean, pretty much any, any access you can think of diversity is definitely lacking in AI. And one of the, I'd say, one of the exciting trends in AI that, you know, even though we haven't solved the problem yet, of course, at least there is more and more awareness of this problem and the missed opportunities. And you've played a big role in that through your work in “Black in AI” So can you maybe tell a bit more about, you know what, what made you decide to get involved and what is so exciting about what “Black in A.I.” has done and is continuing to do well?
Flora And I think one of the organizations that I kind of I'm so impressed with when of anyone they wonder have been more impactful this year is back in A.I.. And so I think I, if I'm not mistaken, it was founded in twenty eighteen by like amazing researchers, team needs Reddit and many others. We kind of saw that, well, there was where, where, where was a black community like they would go to a conference and they report them. And at a time. And if you remember, like I came in, I a bit later because I was coming from a treaty region background and equal in television or even in graphics. It was kind of the same thing where you go to a conference and it's just a very little diversity. And and so when I heard of the initiative that I was doing in 2018, I wanted to be part of it. So I think initially the first workshop, I just wanted to go and see Woo Woo, as was the like because it is all about, you know, bringing together black researchers in AI and encouraging them and supporting them. And so I think one of the big thing that they did was just to say that, oh, well, you're not alone. It's not just you. There are other people in this field. Let's let's come together and connect. And I think that was one of the big things that kind of brought me in at the start. And and then the following year, I wanted to be more involved. So I decided that I was going to be one of the core organizers of the flagship event, which which happens every two weeks every year. So that was that was a big I think for me was it was a lot of work that one of the most refreshing and one of the most impactful thing that I've seen that I've done because it was all about kind of bringing bringing all of this back. Researchers from around the world, like from every single corner, was amazing just where they were coming from and and just the richness of of of the body of work in that community. So I was. And I was glad to be part of that.
Pieter: Yeah, I think I mean, it's been absolutely amazing and maybe as a personal anecdote I want to highlight here is that. The the blockchain, the AI workshop is not not just for black researchers in AI, everybody's invited, right? I mean, it's an opportunity for everybody to to come and see all the great work that's being done by black researchers and entrepreneurs and engineers and so forth. But it's it's mostly researchers, of course, all the amazing work that's that's being done. And I actually remember I think the first one was indeed in in 2017 or 2018 and was in Long Beach, California. And I remember coming to the workshop and the dinner after and it was just the event was just absolutely amazing. I felt like I've never been at an event, at an academic event with so much energy and excitement. It was really contagious, honestly. And it wasn't just me being there. I mean, a bit as an outsider, of course, but very welcome and, you know, really enjoying everything that was going on. Are there other people who are invited said the same thing, like they like this? This is the highlight of the conference. Everything I've been to this week, like this is where the energy is and this is where, you know, things are going to be happening. It was absolutely mind blowing going from, as you said. Maybe, you know, you can't even find another black person at the conference to having this event with. I think it was more than 100, maybe even 200, and it was quite the
Flora: I mean, the growth also of the organization has been amazing. It just shows that it's really it's something that was really needed in the field.
Pieter: I don't think anybody would have dared to predict that before that first workshop happened. It's absolutely stunning and hopefully it makes some people's journeys easier, right? I mean, your journey? I mean, it was a pretty long one from Cameroon to South Africa to the UK to finally get to work on the exact thing you wanted to be working on, right? And so hopefully other people's journeys now can be a large, abbreviated and more directly get involved in the things they want to do.
Flora: Yeah, I mean, they there bring a lot of success stories from people just being black and white. People have, you know, find a job job opportunities through that. Um, you know, that's provisos. I didn't like an event or dinner. So, yeah, for a lot of I mean, there there is you did that showing that the the initiative has really made great impact in in people's lives. So for me, it's an an as as an organizer to just kind of hear someone saying, Oh, this is the first time that I've presented, you know, in an AA event or this is, oh, you know, I had these opportunities because I was able to come to this event. It's yeah, it's just inspiring. And then when you kind of think of it in the struggle for people to where we're organizing like people to even come to the conference because of visa issues and and I think it's it's think it was a rollercoaster, but in the end result is so much worth it now.
Pieter: A lot of progress has been made, but I mean, obviously a lot more progress has to be made right. And so when you look at the future for blocking in AI and more generally diversity and AI, what are some of the things that. You think are really important for all of us to pay attention to and to put energy into?
Flora: If I zoom out a bit what I've seen around diversity efforts within organizations. So they're looking at looking at things like big corporate things like, you know, universities and all that diversity efforts. One of the questions that we often ask myself is like, is this is this moving the needle fast enough? And especially in the recent years, if you have a lot of companies saying, Oh, we are really getting this much so we never see. And then I think people, people who are kind of working in diversity really question this is this is is this actually working? And I think that one of. One of the. Will not be busy mistake. So the newborough is moving right now, have these places where people feel like, OK, I have a network I can rely on now within organizations. When people kind of Kudrjawizki earplugs, there's usually a bit of skepticism because. Because you look at sometimes day, the day you look at the team of the company and you sit and you see that the only black person is exactly that, it is again, I officer. You look at a university, look at all. What are they? Yeah. And you feel like, Oh, the only bad, all of them are in the same scene. Maybe it's external. Maybe it's a bias. And so there is a whole question that I still ask myself that now that we have realized, yes, we want more diversity in these spaces I'm in now just, you know, I mean, really talk about diversity everywhere or I we just see it as a numbers game that, oh, we can create these spaces and then allocated them to diverse people to fill in. And then that's how our numbers. And so and you see the same in in non-progressive, where. Companies talk about investors talk about we want more diverse, you want a more diverse portfolio. Bad day that does examine that issue as if you are black, they want you to be working on the black related problem. And so do so from what I'm seeing is that there that there is this. There is an issue that will happen where we want diversity, but somehow that diversity has to be associated with a specific area in your organization, in diversity or in diversity based on portfolio, which I don't think serves the community. And so that shows one one thing that I want to touch on that. When we talk about diversity is really diversity everywhere. No shows in in specific, uh, in the specific area that is, well, that is separate from all this.
Pieter: Yeah. But yeah, that absolutely resonates. And I think that's so good, you know, to hear that from you, I think. I mean, if I look at at Berkeley, I look at and definitely I mean, a lot of progress to be made, I think. But I think people are trying. I think there is a notion. People are paying attention to it, are trying hard. And of course, at Berkeley, I mean, we've been really fortunate because we were able to recruit a baby to join us as a faculty is amazing. Yeah. And so her influence and her ability to open our eyes to things we wouldn't see before is just unbelievable. Right. And so I feel I feel I feel really lucky that, you know, that she's spending the time with us to, you know, she's not spending time just on her own research, but also, you know, helping us see, you know what, what we're not seeing yet. And I think it's really very quickly within one year made such tremendous impact on the community we're building at Berkeley, which has been, you know, amazing.
Flora So that's that's amazing. And I think that when you have people like her, like in these, not only people like her in this in like in the universities, for instance, but also universities like kind of within a gogerty that are willing to understand and mix and kind of understand and make space for change to be made. I think that can be a really powerful combination. So, yeah, I'm super excited about that.
Pieter: Yeah, one of the things I'm excited to hear when I listen to you is, I mean, obviously, I'm hearing a lot of work ahead, but I'm also hearing some optimism that, you know, it sounds like you're optimistic that you know, well, things are moving in a good direction.
Flora I mean, yeah, I think these things are definitely moving. And I just I just think that it's just a matter of not being complacent and the Oh, things are moving. It's fine now. If we can, we can relax. And even though things are moving, I think that there are also there's just because things are also moving. I mean, a lot of influence, as you see, right? A lot of people really wants to solve these problems, but it's also one of. And so days there is a potential of people of moving too fast because you want to solve it, move fast and then maybe miss miss some other blind spots that that we all have. So I think that it's it's just is an ongoing effort and we talk. And I think when it comes to and I think I see a lot actually a lot of that happening in the academia academic space. So I see a lot of change happening there. Or does it not? Not, not so much, but especially when it comes to corporate and entrepreneurship. I think in those spaces this is the law has to be done in terms of just people being willing to make changes within the organizations, which at the new are more willing to do right now.
Pieter: Yeah, I think I mean, a lot of work ahead and I hear what you're saying is that we can't get complacent. It's not just because the needle has started to move out and, you know, we can start relaxing about it. Now, one thing I'm curious about is if a new person is trying, you know, from an underrepresented group is interested in A.I., and maybe they feel a little, you know, that they're a little isolated if they don't feel isolated. Well, great. You know, they're good to go. But if they feel isolated, what are some suggestions you might have for them to, you know, get connected and, you know, be more part of the community?
Flora So really? So I would just say, please go to https://blackinai.github.io/#/. There's just a lot of information around how to get connected to the community, whether it's on Facebook, you have mailing list, actually even have a forum for job or job opportunities. So it's just a very powerful, a very, very powerful resource for anyone who identifies as black. So I will I will just invite them to go to the website and online system does is that speaking at every event, there is a black and a social gathering. So I organize one at a city like a few months ago, where it's just not possible for people to just come, come together and connect. And so you see about all of these opportunities, it's if you are. Connected to the black and white of Asian. And I think even if they don't feel it in this fight to feel like all I'm saying about the need to connect us to a city like we love to have them because we would love to connect and see what they're doing. All the amazing work that they are doing so and I would say I also invite anyone we don't identify as black to come to one of my friends, as you say. The energy is amazing. It's for everyone to come. And yeah, you never forget it.
Pieter: I couldn't recommend it any more highly. Actually, it's it's absolutely amazing. Yeah. Flora, I think that that's a really great kind of message to end on may be. So well, Flora. So nice to have you on. Yeah, this is a really stimulating conversation. Super interesting. Thanks for making the time.
Flora Thank you. Thank you all for the invitation and well into your whole team for making this happen.