"AI Will Modify Our Interface For Information" A Conversation with Maven CEO Gagan Biyani

0:00Today on the AI Breakdown, I'm talking to Goggin Biyani, the CEO and co-founder of Maven. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Like, subscribe, and share, and go to breakdown.network for more information. Hello, friends. Once again, we are back with another AI Breakdown interview as I am out traveling. Today, I'm joined by another old friend, Gagin Biani. I got to know Gagin when he was building Udemy back a decade ago when I was an investor at Learn Capital, and I've seen him evolve through a number of different entrepreneurial pursuits into now what is a really interesting, exciting, cohort-based approach to online learning in Maven.

0:39In this conversation, we discuss how AI has changed what he does on multiple different fronts, in terms of not only how he runs the company, but also in terms of what people want to learn about and how they actually do that learning. To the extent that we're moving to the part of the cycle where people have bought into the hype, they understand that AI really is a transformative force. They're now looking for resources to actually apply it in their daily lives, in their professional lives. Maven is a resource for helping them do so. And so I think this conversation has a lot to offer. All right, Goggin, welcome to the AI Breakdown.

1:16How are you doing? Good, man. Thanks for having me. Yeah, it's been a long time. I've watched your career from afar because we've known each other for a decade and a half, a decade at this point. And I was really excited to see you come back to education in such a big way with Maven, and then really excited to see how you've kind of dug into AI. And I think as we dive into this conversation, what makes you in such an interesting vantage point is that you're kind of interacting with AI on almost three separate levels. The first being how it changes running a company, the second being how it changes everything else and so what type of education people need.

1:56And then the third is how it changes or could change learning as a product, as an experience. And so I think it's going to be really fun to chat with you. But for people who aren't familiar with you or what you've done, who are you and what are you doing now? I'm Gagan Viani. I'm co-founder and CEO at Maven, previously co-founded Udemy, and then a food delivery startup called Sprig. At Maven, we're trying to rebuild the university from the ground up. And the way we're doing that is by, instead of having professors who are, you know, who've had seven years studying and doing research papers, we have professors who've been seven to 15 years actually in industry, learning how to, you know, building products, designing products, you know, operating, writing code.

2:45And so we take experts who've been in their careers for 10 to 15 years, and we say, hey, why don't you teach the future generations? And so Maven has now hundreds of experts who teach what we call cohort-based courses. So Udemy video-based courses, you can watch them on your own time. It's convenient, but it's also really boring and kind of not sticky. And cohort-based courses are more live. There's asynchronous and synchronous components. There's other people in them. It's kind of like taking a college class on the internet. Yeah, so that's what I was going to ask is sort of like how you guys see the landscape of online education and where you're trying to fit that's differently?

3:22The main shift is just that it's getting more advanced and more deep. So, you know, when we started Udemy 15 years ago, nobody thought that online education was going to be a venture category. We got rejected by 200 venture capitalists, all of whom said, hey, we don't want to do education businesses. And, you know, fast forward to today, and now video-based courses, this format that we invented at Udemy, or we didn't really invent it, but we sort of popularized it, is one of the predominant ways people learn. You can go to Linda Masterclass, LinkedIn Learning, which was Linda, and Coursera or Udacity and Udemy.

4:00And you can buy a course or take a course for very low cost, where you can learn a subject by essentially like watching videos of someone teaching you it. Maven thinks that the future is actually going to be more advanced than that. People aren't just going to watch videos. They're going to interact with an instructor. There's going to be someone there answering questions. There's going to be peers that you're learning with at the same time. And so it's a much more social and interactive learning experience that's defined primarily by the fact that it goes over two weeks or three days or three weeks.

4:33So it's a cohort. You know, it happens, starts, it ends, and a group of people go through it at the same time. And so Maven's goal is to reinvent online learning through cohorts. And part of it, I imagine, is just what you can do differently when you have a more rigorous, comprehensive experience, right? It's sort of, you know, you're not sort of skipping rocks across the pond as you might be with just sort of, you know, buy a class. You can actually get into deeper learning, I would imagine. That's right. I mean, it's better on a number of fronts. First of all, it's far more engaging. And this is probably the number one thing, because the biggest challenge for professionals who are learning things is that they always give up.

5:11And so the average video-based course has like a 5 % to 10 % completion rate. The average Maven course has like a 75 % to 90 % completion rate. So people actually stick with this form of learning. The second thing is that it allows you to get unstuck. And there's a lot more nuance to what you learn. So if you're learning with a group of peers who are at the same spot as you, so let's say you're a mid-career marketing manager. You spent three to five years building up your career, and now you want to learn how to use AI in marketing. Well, we've got a course for you, and the people who are going to take this course alongside you are other people who are in the same position as you.

5:50So you learn a lot through that feedback loop. And then, of course, the instructor is someone who's been doing AI marketing for a decade or two, or ideally they've been doing marketing for a decade, AI marketing for a couple of years. And so that's the other thing. And then the third advantage, of course, of cohort-based learning is the fact that you actually do real projects. So you actually get things done and you submit them. And so not only will you passively watch content and learn from this instructor and from watching your peers, but you'll actually actively engage and have to put together a project, a final project usually, and sometimes projects in between that, you know, for example, if you're taking a mid-journey course, which we have a really great one from Nick St.

6:32Pierre, Nick will make you publish your mid-journey prompts and create actual assets, like image assets that you create through your mid-journey work. So you'll have real active learning. So we think cohort-based learning is really just a lot better than any previous form of online learning. And it's also just how learning has existed for centuries, right? Learning has always been cohort-based until the internet allowed you to learn on your own time. So let's talk about the AI dimension of this. I remember seeing a post from you on LinkedIn that said, basically, it sort of was about your personal journey as related to AI within the context of Maven, and that as recently as January, despite paying attention to it and checking out all these things, you were still skeptical of its impact for your business.

7:23So bring us into that and then kind of what happened from there. Yeah, I mean, ultimately, you know, I'm skeptical of hype cycles. I've been around now for 15 years in tech, and I saw many hype cycles, most of which did not materialize. You know, self-driving cars was really big because I was at Lyft for a while, and I remember seeing that happen. I remember seeing AR and VR has been talked about for about a decade now. You know, we just consistently in tech get really excited about things that don't end up becoming real until, you know, years or decades later. And so I had the same reaction to AI.

7:58I was like, okay, this is cool. I love watching people like chat with this thing. And I, you know, I played around with it myself, but how is it actually going to make a difference? Like, what am I actually going to use this for? And then one of my co-founders, he basically, our engineering co-founder Shrayons, he was playing around with AI. He's much more excited about new hype cycles. I'll say he's more susceptible to them, but in a good way. And so he got excited by AI and he was like, I want to build an AI recommendation system. And I was like, are you sure you should do that? you know is that worth your time and like he's like look i don't care i'm going to do it in my free time anyways so you know he went rogue and went and built this ai recommendation system and his first version okay working as a solo engineer on his off nights you know like whenever he was free was two to three x better uh than anything we were doing manually and i have worked on recommendations in the past at my last company, Sprig.

8:57And so I know how hard it is to do recommendations in the old school way. And he just plugged it into OpenAI and basically stitched it together and his recommendations were fantastic. And so that made it very obvious to me that, okay, there's real world implications. The other thing that happened is we started using AI in our blog and we started putting out content that was virtually indistinguishable from content that was written by a content writer. And so I started to see my team use AI. And the practical applications, I think, are still more limited than people would like to admit, but they are very real.

9:34And there are things that are completely transformed already by AI and will be certainly over the next three to five years. One of the things that I think is so interesting about this for you guys is that sort of step one was it being relevant for Maven as a company before you started to think about sort of all the, you know, all the different courseware that might come out of it, or at least they were happening kind of simultaneously. What then was sort of the decision to move into actually offering AI courses? I mean, I can't imagine that it was a particularly complex decision given how much interest and demand there is.

10:08It wasn't. Yeah. Yeah, it wasn't hard, but I'll give you some interesting colors. So we were using AI internally as a company because we're, you know, on Twitter, as many people are on LinkedIn. seeing it. But what happened also is that we ended up like some instructors on our platform have been AI enthusiasts and have been industry experts in AI for like over a decade, right? So like a lot of people think about AI as something that just happened in the last year or two. But as you and I both know, it's been around for a long, long time. And so there's an AI lead at Meta, there's AI leads at Google and other places, Amazon Alexa, that are on the Maven platform teaching AI courses.

10:52We had the storytelling course about AI that was around. And these courses predate, you know, ChatGPT 3.5. So these courses have been around and they've been doing well, but they started to take off. And so in January or February, as they were taking off, we added an AI topic category into Maven. And that became like 40 % of people who chose a topic would choose the AI topic. And so we were seeing all these signals and we were already selling AI courses. We already had great AI courses, right? And so we just said, okay, let's make a concerted effort over the next three months to really develop and deepen our understanding and our offering in AI.

11:34And now if you're, you know, curious about AI and you want to learn it, you can go on Maven and you can basically learn any subject you want. You can become a, you can get advanced AI for product managers, you know, prompt engineering for LLMs, mid-journey, et cetera. And like, we're continuing to add more. So that really came about because of multiple signals, not just the Twitter and LinkedIn signals, but also internal signals. One of the things that's interesting is that I think that in the explosive moment that happened around ChatGPT, and I think there were other things going on around the same time, but particularly with ChatGPT, is something like 100 million, 200 million people felt like they caught a glimpse of the future.

12:17And they very quickly started to think about what the implications are for me. And that has obviously led to a lot of exploration and people consuming new types of content. But one of the categories of content that doesn't really exist in a mature way yet is verticalized content around AI or the disruptive impacts of AI. So, for example, what AI is going to do within finance. Now, you could go find blogs about that or, you know, you'll occasionally find like, you know, on the real estate podcast, they'll talk about AI and stuff, but you don't have it as sort of a robust discipline. And one of the things that you notice if you go to Maven's AI course list is that a lot of it is verticalized, either in terms of, you know, here's ChatGPT for finance or advanced ChatGPT for finance, or it's sort of, you know, within a professional domain, like product management or product design or something like that.

13:13So maybe, you know, cuts across different industries, but it's still, again, kind of honed in. Is that something that you guys are intentionally doing, trying to like really kind of give people AI in the context of what they're actually experiencing? Yeah, I mean, Maven courses are in-depth, right? So you're going to spend three weeks or three days of intensive work sort of learning a subject. You know, it's going to be five to 10, potentially 20 to 30 hours of your time. So we need to provide you with value. They also cost money, right? It costs a few hundred dollars usually. And so you have to really justify that spend.

13:49And so that means that we are what we call it at the bottom of the content pyramid of bullshit, okay? So you can really bullshit people on Twitter because you barely put, you know, it's like five tweets and boom, you've got a viral threat, right? And then you can get on a podcast and you can talk about something for 20 minutes and it can be like pretty interesting and it's going to be more depth. Less bullshit, but still some. Then you go even deeper and you go into video, right? Advanced video, like maybe you're doing an hour long lecture at a conference or whatever, like less, even less bullshit, but still some.

14:23And then courses are the least bullshit of them all in the sense that they're so narrow and so granular. I have to teach you an actual skill. And as you get into it, that's where it's really interesting that AI is at that level, that it's actually still valuable there. because if AI is valuable on Twitter and on podcasts, that's already interesting. Like you are doing such a service to everyone to help us think about it, to consider it, what does it mean, et cetera. What courses do is tell you how to use it. And when you're at the point of using it, you have to get verticalized on some level. And that's actually what makes Maven so special and unique is that almost all of our categories are a cross section of a topic area and a vertical discipline or a function and a specific skill or something like that.

15:11So and that's important because that's how you get that's how you learn a real skills. You have to get that deep. With the people who teach courses on Maven, how much is that you guys going out and recruiting? How much is that people coming to you? It's not a fully open platform where anyone can come and teach a course or anything like that, right? Anyone can use the product to deliver a course, but we do not surface all instructors. We're fairly curated with who we surface. So if you're browsing on Maven, you won't see the average instructor. But yeah, anyone can use our software. We sort of split those two things up.

15:45But our instructors are coming from two different places, inbound and outbound. So our existing instructors who already happen to know about AI and are telling their friends who are in AI are starting to bring out courses in the subject. And so we've been around for two and a half years now. And during that time, we've built up a large stable of experts from venture-backed startups to famous tech companies to professors at universities like Stanford and MIT and stuff. And so these instructors started to say, hey, I know something about AI. I'm going to teach on this subject. I already teach at this other institution, whatever.

16:24And then we also reached out to people. So we have a bunch of people who we've literally gone and recruited and said, hey, you know, you're one of the leading experts in this subject. And we also had a bunch of calls. I did a bunch of calls with people who I knew wouldn't teach, but who I thought would tell us who we should have teaching. And so actually, like I talked to a lot of people who are quite well known in the AI space. And I'm not going to like, you know, name drop them because it's not appropriate. But people you would recognize and you'd be like, wow, that person helped you with your curriculum.

16:55And yeah, we basically figured out like, okay, this is what people actually need to learn and what they need to use. And some of those recs resulted in people, recommendations resulted in people that are teaching on the platform today. Yeah, it's awesome. I mean, you've seen already, you know, some of the folks who started to build a following on Twitter or somewhere else jump over. I mean, you mentioned Nick St. Pierre. Undisputedly, I think, one of the best follows if you're interested in mid-journey and, you know, generative art in any sort of way. And I think he's now teaching a course over there.

17:26Yeah, we've also got Bilawal Sidhu, who's one of the, he's got like a million plus followers on YouTube and on other channels combined. And he's amazing with generative AI, more of the combination of AR, VR, text, video, etc. So more video content and Nick's more visual content and static and they're both fantastic. I also got Elvis Ravia, who started Dairy AI and built one of the most popular guides for prompt engineering for LLMs. And he's like a, you know, very popular instructor on Maven. And then there's people like you talked about the chat GPT for finance. You know, Nicholas Boucher, who teaches that course, like we didn't know him before he started teaching the course.

18:10It just sort of just showed up on Maven, started teaching it. It took off. And he's actually a very, very amazing instructor. has incredible reviews and has a great background. So that's the beauty of the internet, right? You get some people who have traditional backgrounds, Stanford, MIT, whatever. Then you have some people who are like famous because they built their audience. And then you have some people who are successful because they've worked in industry at places that we know. So Maven has all three types of instructors. Are there any themes or common threads that you're seeing in terms of where people are gravitating to create content within AI or where people are consuming, you know, any areas that, you know, any of those verticals that are distinctly more preferred than others or, you know, that might have surprised you?

18:56Until AI, we had trouble building courses that engineers really were attracted to. So we, you know, candidly, we just like didn't think that engineers wanted to do cohort-based courses for a while. And that was kind of a funny mistake. It wasn't that we, they didn't want to take cohort-based courses. They want to take cohort-based courses that were highly relevant to the things that they were working on. And I guess we just didn't have the right catalog for a while. And so prompt engineering has taken off. It's been really, really powerful. It's interesting to see that grow because again, we didn't have that audience before, but honestly, like, I think that it's, what's surprising is how many use cases there are for AI, but also how many use cases they're all already on Twitter.

19:40So like we do not have, We have not had a breakout use case yet. That is not something that you would see an article about or that you haven't done a podcast about. I bet that what it really is is if you hear about something and you want to dive deeper and really learn it and start executing on it in your job or in your work, that's when you take a course. So we're very down funnel, right? And what that means is that we don't have subjects that aren't popular. It's more like we have all the subjects that you would expect. Yeah, no, that makes sense. I imagine, too, there's going to be a natural progression of as industries sort of cumulatively get on board, you know, you're going to see a lot of pressure for people to kind of dive in in a deeper way.

20:23Like, you know, I used to be in the marketing space and it's Can Lions this week and every event. I mean, just watching from afar. I think that someone someone wrote that there were 200 distinct AI related events. And, you know, there's hundreds and hundreds of events there all going on at the same time. but still, I mean, it's dominating, right? And you got to think that every marketer is now racing to figure out how's it actually going to impact me? What do I need to do? Because this is not an industry where it's particularly controversial. It's almost all upside in a lot of ways. I think the controversy is pretty overblown.

20:58There is some, but for the most part, for most individuals, they see this as an additive and a bonus of something that's going to make them more productive and get more work done. And so the average person is pretty excited about what is possible. And I'm too. I mean, we have started to implement AI throughout many workflows at Maven. Summarizing qualitative data, it's super good at. I ran an event recently where we essentially decided who was going to be accepted into the event because it was oversubscribed just entirely through ChatGPT. We didn't have to manually go through it at all through the applications.

21:35We find that it's super good for creativity, as people know, brainstorming, things like that. We're using it for the text to video stuff has just gotten super good, video to text as well, or audio to text. So yeah, it's starting to pop its head up in many places. There are also many places where it still has some refinement to get there. But if you're someone who's new and you're excited by it and you learn it now, over the next six to 12 to 18 months, you'll be ahead of the curve and you'll be there when it starts to become super, super relevant. So I think it's a good time for people to start learning it, but people can also, some more established individuals in their careers will learn it in six months or a year from now.

22:22And I think that's fine as well. Yeah. It's going to be different for different people. I think it's fascinating to see how it's sort of shaping the platform, but also just, you know, you guys are kind of like a live test case and how a company can best take advantage of it at the same time. Yeah, it's interesting because we're not actually like, we are eventually going to deliver learning content through AI. We'll have an AI chatbot that helps with courses. I'm sure we'll do all those things. But the truth is what we're really doing is we're like, we're selling jeans during the gold rush, right?

22:51Like we're telling people, hey, we can teach you how to use AI at work. or in your personal life. We also have a course on storytelling. We have two actually on storytelling that are more for your personal life or for personal interests. And so we essentially are selling the picks and shovels or the genes or whatever of this. And I think it gives us an interesting vantage point because we're not necessarily building this really complex or interesting AI tool, although we'll add AI tooling to Maven for sure, what we're really doing is teaching people how to do it for themselves in their organizations.

23:31This is sort of where I wanted to maybe round out the conversation is, you know, obviously you've been around education and learning in various forms for a long time. You've seen lots of trends, lots of patterns, both, you know, sort of discrete, but then also, you know, big, big patterns. Outside of specific features, like you mentioned chatbots and things like that, do you have a sense yet of how you believe AI might impact how we learn? You know, obviously, like cohort-based learning is a bet that you've made to bring, you know, online learning more into scope of, you know, kind of normal human patterns.

24:03When you look at what AI could do for the way that people learn, is there anything you're particularly, you know, either convinced of or even just interested to see how it plays out? I mean, I think that the best examples are Google and YouTube and what they did for learning. So, you know, when Google came out all of a sudden and Wikipedia, I think you have to, you have to say those in the same breath. Like when those two things came out, like all of a sudden, so much more information than ever existed before is available at your fingertips. And so students and kids learned so much faster because they had access to this information quickly.

24:34And there was less questions about facts. Facts were facts now. Now, you know, for a while there, I'm sure you remember growing up, you know, you could have debates about whether or not, you know, a certain war happened and how many people died in that war or whatever. And now it's like, no, you can literally like there is no debate. And then YouTube also then provided the next major trend in education, which was like, hey, if you want to learn something slightly more in depth, you know, five minutes, 10 minutes, you know, you can get it on YouTube. And I think that's where AI is going to make the biggest impact is AI will modify our interface for the Internet and really our interface for information.

25:11And as that interface changes, our learning will also change. Our learning process will change. We'll have access to more insight. So one of the things that's a challenge with Google and YouTube is they're really bad with nuance and they're really bad with qualitative information and data. So if you want an opinion or an argument, Google sucks, actually. Like the blog posts that come out there that have actual opinions are so basic because of the syntax of search makes it impossible to evaluate whether or not an argument is actually well constructed or not. AI will slowly dramatically improve the quality of, forgive me for saying this, the quality of qualitative information and insights and questions.

25:59And so as you start to learn on your own, you can ask a question like, you know, what's a counter argument to my belief that rent control is good for society? Right. Or what is the best way for me to convince my parents to let me go to, you know, Brown University instead of Princeton University or whatever? Like these more qualitative questions are going to be answered so well by AI that people will start to have their opinions affected by the Internet in a way that I think is much better than what we have today. I think it will be a long time before that goes to the next level where we'll have AI actually help deep learning.