Edtech Insiders
Edtech Insiders
Week in Edtech 02/21/2024: OpenAI's Sora Rivals in the Works, 2U's Financial Struggles, Reach Capital's 2023 US Edtech Funding Report and More! With Special Guest, Matt Sherman of LSAT Lab
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
In this episode of Week in Edtech, Ben and Alex discuss:
1. AI
2. Edtech
3. PreK-12
4. Reports
- Reach Capital’s 2023 US EdTech Funding Report
- Oppenheimer Report
Special Guest:
- Matt Sherman, Founder and CEO at LSAT Lab
Welcome to Season Eight of Edtech Insiders where we speak to educators, founders, investors, thought leaders and the industry experts who are shaping the global education technology industry. Every week we bring you the Week in Edtech. important updates from the Edtech field, including news about core technologies and issues we know will influence the sector like artificial intelligence, extended reality, education, politics, and more. We also conduct in depth interviews with a wide variety of Edtech thought leaders and bring you insights and conversations from ed tech conferences all around the world. Remember to subscribe, follow and tell your ed tech friends about the podcast and to check out the Edtech Insiders substack newsletter. Thanks for being part of the Edtech Insiders community. Enjoy the show.
Ben Kornell:Hello, Edtech Insiders. We are wrapping up February with an awesome episode bring you everything in Edtech. I'm Ben Kornell, along with co founder CEO extraordinaire Alex Sarlin. We love you, Alex. So glad to have you on the pod today. Let's dive in.
Alexander Sarlin:Let's dive in. There's some really exciting things coming up on the podcast next. Next week or so we just released our Hillary Clinton interview which we mentioned last week, incredible to be able to talk to the Hillary Clinton about education and mental health and technology. We also have some really exciting big names coming up. Next week, we're gonna have an interview with Robin cook from Audi School, which is doing sort of AI driven stem curricular development. So definitely some interesting stuff there. So with that, Ben, are there any events? Or should we jump into the news?
Ben Kornell:Well, we've got the South by Southwest Happy Hour on March 1, we're also you know, I'm doing a panel on generative AI what we've learned in the year, four o'clock on Monday, March 4, so we've got those two events at South by if you're going to be there and march 21, we have our happy hour. See you there. Let's dive in. On the AI side, the world is abuzz with Sora. It is open AI new generative video model. And it's fascinating because not only did it catch the world by surprise, but it also suggests new capabilities for LLM that are radically different than capabilities that we've been using for video generation before. Namely, that when we render video in gaming engines, often we're rendering object by object, in this case, the LLM is learning holistically from videos and then rendering complete scenes. And you know, you'll note that sometimes the feet are backwards or sometimes you know, XY and Z are wrong. But if anything we've learned from, you know, mid journey to today has been that the rate of improvement the rate of change will be exponential. It was exciting week to kind of glimpse into that video AI future. Alex, what are your main takeaways here?
Alexander Sarlin:Yeah, I mean, it reminds me of the moment probably around 1995 or so when the internet sort of was clearly making its move from being text based to you know, video and visuals becoming core to the sort of language of the internet, because, you know, we've chatted a bit he's only been around, we always have to remind ourselves, you know, less than a year and a half. And the first use case is everybody's looking at text to text, you know, ask a question, get a response in text, we are now quickly entering that Omni modal future that you've been anticipating where you can upload a video and get text out of it, you can upload a text and get a video out of it. The store is not available yet to users, we can't play with it outside of researchers, but the videos that they've released as sort of, you know, teasers about what it can do are pretty mind blowing. The one that really blew my mind was one of this sort of overhead shot of Tokyo of people walking through a Japanese, you know, TriGuard and it just is indistinguishable from reality. Really beautiful, obviously, could be used in advertising contexts. Really exciting. It's also worth noting that meta, you know, I mean, there's All right, so you're hearing right, Google released Gemini 1.5, it has incredibly big context window, which is a really big change and really matters in the AI space, you can upload 1000s of pages of text, or up to an hour of video or audio, or many hours of audio. So that is a big change. That means that you can get AI to respond to you know, a movie, or you know, a library of books. So that's a huge deal. And then, you know, back in the video space, we saw Mehta launch its new version of its text to video model, they're calling an emu video. And they're basically saying that now EMU is going to be integrated into Instagram, the number one sort of picture sharing on the globe with maybe the exception of Tik Tok and into Facebook. So, we are clearly seeing these big tech companies and Really going tooth and nail against each other to own the future. And I think Sora is going to be a really, if it's anywhere near what they showed in their samples, it's really going to blow minds in a big way. So I'm really excited about and of course, in education context, creating educational video is one of the most expensive things you can do. And many of the people creating online courses have had to use text and have had to use still images and have had to use, you know, static content on LMS is because it's just simply unaffordable to make educational video, especially if you're a single creator, that may not be true anymore, you might be able to, as a single person make an incredibly sophisticated video based online course. And that mean, you may be able to do that in a month. So please, you know, we should keep a big eye on that video space. Yeah,
Ben Kornell:I think it speaks to a couple big trends that are coming up one, video drives engagement, that's really the element of video that is truly profound. tiktoks got to be quaking in their boots, their whole platform is based on user generated videos, are they going to be able to have the stranglehold they've had on that audience, if you can make viral and addictive videos through an LLM and generator or maybe it explodes their business model hard to no second, the amount of compute that is going to be required for video could be astronomical. But what's interesting, the Gemini expansion of parameters actually suggest that we've had a 10x leap forward in what LLM can do, there must be some breakthrough behind the scenes that's allowing their context window to be that big gray paren still not charge an incredible premium. So we may be on that race to essentially zero cost compute over time, one by the airlines being more efficient, and to buy more compute power becoming available and online. Third, in the education context, what we really have hit on is that human extension is a really incredible tool for AI. Humans, I feel like every single thing that I see in the ED space, it's showing how AI helps us double down on the humanity. But if we can extend the human being if we can extend their reach, we can extend engagement, things like math practice, things like reading practice, where you don't need the human to be overseeing that part of it, they can teach the new concept. But then there's this extension, that future which seemed very science fictiony, yes, is now becoming almost an inevitability. And so then the question is, like, who's going to take advantage of that and scale it with access? My hope would be public schools, and our education systems that are strapped for talent would find ways to leverage these things. But if we're going to be in a direct to consumer market, then I think it's very, very concerning, because it's going to go to the highest bidder. Okay, last thing, before we move on, I will just say this is yet another moment, where VCs are actually putting on the brakes. Unless you're a VC that's putting 100 million to, you know, several billion dollars into one of these home run hitter LLM companies, on the kind of application layer side, you've got to be like quaking in your boots, because every time one of these new releases comes out, 1000 companies are crushed. And this is just another example where you may think you have a lead, and then boom, comes in one of the big tech players, and immediately washes out an entire segment of the industry. And
Alexander Sarlin:the big company under threat with this launch is probably runway runway has been sort of considered the leading text video platform in the space and they were sort of ahead of the game. But if Sora can, you know is going to love them, then that's going to be a big deal. One more quick, quick comment about this text, video technology and how it might affect education. You know, there's the concept of education providers and content creators can make video content at a lower cost than they ever have before. We've talked to a profit gem on this podcast. And they're leaning into that in a b2b way. That's a big one. Individuals can make incredible video courses. That's a big one individuals, even young people can create incredible courses. You may have high school students or even elementary school students creating YouTube courses without needing even a video camera. It's amazing. That said, the other thing that this can do is it can mean you can have personalized tutoring for individuals in video format. So if a student is struggling with a word problem in math, for example, they can say help me understand this and it will create a video that shows the problem shows how it goes together, splits the apples and cuts the pizzas and all the things we've always known. This is a game changer for education as well. That is not something we've Ever been able to do before? It's video on demand at the need of the student? So I'm just like, Yeah, my mind is really reeling with this. Yeah.
Ben Kornell:I mean, the full disclaimer is today probably give you the wrong answer, and it probably wouldn't be that great. I think there's, there's a way in which part of what we do here at Tech insiders, as we're looking around the corner at what's coming in a year or two years, and this felt like that kind of pull the curtain back look around the corner. Yes, moment. Well, speaking of looking forward, things aren't looking great going forward for one of your alma mater companies. To you, Alex, I know there's some limits on what you can and can't say. But, you know, really, there was so much potential in the opium space. And then when we saw edX and two, you coming together, we both felt genuine excitement, that you are going to have marketplace, and, you know, custom courses for universities coming together in this great platform to create online learning opportunities at lower costs for millions of students. And yet, here we are, they are on the brink of bankruptcy. What's going on? Yeah,
Alexander Sarlin:no, it's a really crazy story, I mean to you, a few years ago, had a stock price of over $90. And now we're looking at, you know, under 50 cents for the stock price right now. It's in danger of being delisted from the NASDAQ. I mean, you know, I don't feel like I have a lot of super inside knowledge from my time at to you, all I would say is that the vision, the big vision of to you, since they sort of decided to expand out of the college OPM space was, we want to be the lifecycle of education, we want to be able to provide education for learners, at any phase, whether they're just dabbling with something whether they want to get a certificate, whether they want to go deeper and get a degree, whether they want to get a masters or ag yet and on and on, MBA on and on. So it's really supposed to be sort of the whole spectrum. And their purchase of edX was a way to move sort of all the way down to the course level of their purchase of get smarter was a way to move all the way down to the certificate level in the sort of executive education level, I wouldn't say down but move into the education level. And the trilogy, acquisition was about moving to the bootcamp level. And Boot Camps are yet another way to upskill. So their vision was very big, I say was, I mean, they're not gone. But their vision for what to you wants to be and wanted to be, it was very big. I think the big thing that really sunk them and this is my personal take is just not being able to put the pieces together fast enough in a way that actually convinced their own investors and their own shareholders that they had a real clear plan, but also not being able to fully execute on that sort of cross functional upsell, right, not being able to actually convince learners that you should come and do you know, a course on data, and then do a bootcamp on data science, and then get a master's in data science. That was the plan. I don't think they found it yet. And with the humongous amount of debt, I mean, it's over one and a half billion dollars in debt just for those two acquisitions over the last, you know, four years, they just had huge financial issues. Chip Pasek is now out he is starting a new company around athletics, because that is his personal passion. And the CFO is in charge. And you can expect more and more rounds of layoffs, potential bankruptcy, it is a real surprise. And I've mentioned this on the podcast before but when you know, when I was at Coursera, many years ago and Coursera first went into the degrees space from the individual course space, Chip reached out and was like, Okay, we're now real competitors. Because, you know, we've been owning this OPM space. And we've been sort of not owning, but you know, big player, very visible player in the OPM space, to now look and see them going from the, you know, I think, certainly visible leader and in many ways, the leader, you know, PMS, to really just not having the face of, you know, almost anybody in the space is very sad. And a lot of the people who I worked with a to you who are now not there anymore, are really struggling to sort of make sense of how a company with that much traction originally in the market sort of could stumble into this judo 35 cent a share space, it is I'm not sure I'm adding a huge amount of insight value on this other than it's really a shock. And I think it's not great for the public ed tech markets. Yeah,
Ben Kornell:you know, the financial side of the story, which is easier for me to speak to, you know, when people see a company that's on the verge of bankruptcy, they're like, how can it be bankrupt, it's earning so much money. Whenever you're looking at the value of your equity, you have to basically subtract the debt from it. And really, this is a story that is accelerated by a context where debt and loans were essentially free with no interest rate yet, all the way to this accelerated interest rate circumstance. And so when they were refinancing their debt, they really really had an incredible hurdle. So I think that's the story of bite. Use here. That's the story of two you hear. But I think the lasting takeaway might be one. Any business that has low LTV is very hard in education because customer acquisition is high. Regardless, they went from a high LTV business, which was selling to universities, to a low LTV business in selling individual courses to students. Coursera went from low LTV to high LTV, because they're selling to universities. And but of course, they're also sells a lot to corporate. And so the journey that Coursera went on from low high penetration, but low LTV to higher LTV is actually the path you want to go to, it does make one wonder, does this mean Coursera is going to have a free run at the market that could be very, very good for them. Or it could be this market is just crap. And so no one's going to be looking good. I think that it's somewhere in between. But I do think this will actually clear some of the space for Coursera to run. They're the only scale player. And if I'm over at MIT, I'm gonna think, can I acquire edX back out of bankruptcy, because it was an incredible nonprofit play. I mean, and by the way, when you go through bankruptcy doesn't mean the end of the company, you just end up selling the parts off, I think trilogy could be a very good standalone business in the bootcamp space. So it may not be that it just like an edtech, every business is not a multi multi billion dollar business, you might have a very good business that kind of tops out at 100 million a year. That's still a really good business. So I think, you know, Icarus here was, you know, there's an Icarus component, and all that you said about them converting
Alexander Sarlin:and acquisitions. Right, right. So you invite us both big acquisitions, because it was cheap to get capital. That's like, definitely a part of both their stories, please go ahead.
Ben Kornell:Well, last story, before we go to our interviews, we had a couple of big tech reports come out first, we had to reach capital ed tech report, which is always a must read love everything that reaches putting out and they also have this great focus on seeing an early stage, they do a lot in K 12, which you know, is close to my heart. And then we also had Oppenheimer give a report on the Oppenheimer one, what was really, you know, heartening was that basically, we're back on trend with 2020 investment levels and m&a activity. And so there was almost the sense of like, Hey, what did you expect, we're back to normal. The second was that VC activity is down, partly because of interest rates, partly because of public markets, partly because this new technology AI, which did has created bumps in valuation, but there's a wait and see approach there. But it's not so meaningfully down. And, you know, Europe is a little bit down. But you know, the report who came from Matt Johnson, when a friend of Ed Tech insiders, shows that it's durable, the growth in ed tech in Europe, in particular is durable. But still, the main show is North America, given the kind of amount of money available for on the consumer side, and on the investment side. So a really interesting set of reports out there, we encourage you to check them out. We can include them in the show notes. And you can also check them out on our LinkedIn, any big takeaways from either of those reports for you, Alex, I
Alexander Sarlin:know we're low on time, and we're getting to our interviews. But I think the thing that was encouraging to me about the Oppenheimer report, as you said, it sort of told the story in a little bit of a different way of like, you know, yes, the pandemic created this huge boom, and lots of complexity, good and bad in the sector. But now that we're sort of coming back down to earth, it's not that it's coming back down, you know, through the floor, which I think it felt like for a while it's coming back down to a little bit more of a, this is what would have happened without a pandemic type of future. And, you know, when they they list out, you know, all the tech funding rounds, and the big, you know, the big players and what struck me, I didn't realize this is that reach had done by far the most investing in 2023, followed by learn capital. And that owl, even though they have the biggest fund did fewer rounds, and a higher percentage of them were rehabs for companies they had already invested in, but reach and learn, had did actually a big number of investments into new companies. That was interesting to me. And I think that by itself, showed that there's still some enthusiasm, you know, in the investing sector, at least for the people who had already raised funds. So that was a, you know, heartening for me.
Ben Kornell:Okay, so Alex, other news that we saw in the headlines was this. Unleashed brands acquisition of Sylvan Learning, Sylvan Learning feels like the OG of ad tech space, from physical centers to their online learning. What does this mean to you? Yeah, it's
Alexander Sarlin:really interesting. So So unleashed brands is a conglomerate of basically, children's physical spaces that serve families and children, they own little gym, they own I think a coding center they own At a number of different businesses in the space, they call themselves the world's first youth enrichment platform. And it's the sixth acquisition for Unleashed. And this makes actually a lot of sense to me. You know, Sylvan is a very old school brand, you know, they own a martial arts, but how many stones martial arts, they own a gaming platform. I think what they're really thinking here is, they in this very online world, where people spend a lot of time at home, I think there's beginning to be an interest in people getting back out into the world and actually doing things. And I think they're betting that there are, you know, these entertainment ways that they've already been doing. They own, you know, urban air Adventure Park, but I think they're betting that one of the things that's going to get people out of the house is education and sort of supplemental education, like Silvan. So I think it's a pretty decent bet. I mean, I have sort of poo pooed, this strategy, when it's been happening in India, when some of the tech companies sort of snap back to in person center. So I don't want to be a total hypocrite here. But for a company that is already successfully running lots of in person centers of different kinds. I think this makes a lot of sense. So it's interesting, and I think we'll start to see you on a college planning franchise. I mean, I think we'll start to see a different kind of, you know, ownership of Sylvan and maybe they'll connect it to some of the other properties they have. Yeah,
Ben Kornell:you know, it's interesting, there's a almost like this reversion back to digital physical. And what we've basically been seeing is, the ad tech industry really pushed into full digital. Yeah, ad tech industry also pushed into product led growth with relatively low customer supports, and you know, kind of chasing software margins, and consumer and SAS models. What we're seeing now is a reversion back to what's been tried and true, which is a physical base, with digital extension on top right, and more and more customer support. I will also say, I feel like in our space, we're seeing much more traction with b2b than b2c. So if you have a b2c, that's pure digital, it's the kind of user fatigue, the churn the customer acquisition, the LTV, it's really not working for you, well, if you don't have some sort of physical in person or labor component that people feel like they're getting value for. So I think in the consumer space, we're going to see more of the digital physical, and in the b2b space, we're seeing more and more heavy customer success, where your revenue line for professional development or for implementation or customer support might actually be as high as your software revenue line, if you were to really fully break it out. I think that that's actually a healthy place for us to be in. And generally speaking, in Tech Tech, I will say, this era of software margins, I feel like it's going to be going away, because the cost to deliver SAS Type products is going way, way down with AI. And so the willingness of a company to pay, you know, 100 bucks per seat, when they know that they could probably build it themselves for $8 per seat, it's really going to drive software margins, down, down, down. So I'm actually more I'm bullish on. This is why I love education, high LTV, we make a big difference, we make a big impact. And our customer value is very, very high. So if you can just figure out the customer acquisition side of it, you're gonna have a great business. So that's where I see Sylvan going, and it kind of takes an OG brand like this to shine a light on that.
Alexander Sarlin:The only thing I'd say in response to that is the there have been some bright spots in b2c in ad tech. And the number one, of course, is Duolingo, which, you know, we mentioned last week has like doubled its user base in the last couple of years. And it was already a pretty big user base that we are also seeing, you know, big Coursera has still gotten lots of new users year over year for their b2c side. So there is some exceptions. And I think, you know, what will be interesting is to see these b2c ad tech companies that do have higher churn that do have to sort of do more, you know, to renew that have to convince people to use them, and then convince them that they've used them enough to to keep it going month over month or year over year. You know, and try to build that lifetime value. I think it's gonna be interesting seeing how they adapt to this new world of AI especially the language learning apps, right? I mean, Duolingo has become I always say this, you know, it's the Oregon Trail of the 2020s in that it's the go to tool that people think of if you mentioned education technology to somebody who's not in the field, they go Oh, right. Do a link of course I I've tried that, or my friend has tried that or something like that, you know, Khan Academy is maybe another one. And I think that Duolingo has this huge advantage. They also are started by AI people, right? I mean, they have a lot of technical expertise in this. That said, We're entering a world where there's this sort of Omni modal, super smooth, translational type of world where you can sort of translate anything into different languages. That's true of personalized tutors. That's true of videos now, I mean, so there's so much you can do, I think we're gonna see a sort of revolution in language learning that may, it may speed up, frankly, or it may really cause and be a big hurdle for these language b2c apps that have been sort of the one of the main exceptions to the rule that you just blamed, which is like, you know, that b2b and B to school business University models have tended to do better over the last and business to get you know, in vita, ESL into corporations to schools to universities have tended to be higher lifetime value and do better. So little bit of a tangent there. But I think that there's there's something interesting in watching the few b2c companies that have really taken off what they do in this sort of new era.
Ben Kornell:Totally. Well, with that, let's wrap our episode today. It was rapid fire. And let's head to our interview. Can you tell us it's AI meets LSAT, tell us a little bit about who's coming up.
Alexander Sarlin:This is so cool. So we had mentioned in an earlier newsletter that companies that have a lot of proprietary teaching data, have an interesting opportunity to use that data to train GPT is or to train fine tune AI models and sort of do something beyond what general purpose model can do. And one of the examples we gave was LSAT. If you're a major LSAT provider, you have 1000s of questions 1000s of lessons about LSAT, if you train an AI on that, it's going to be a better LSAT tutor than, you know, chat. UBT is on its own, or the Gemini is on its own. And a founder reached out and basically said, we've not only done what you just said, where we've trained an AI model to be an incredible LSAT tutor expert, but we have also done it with real time tutoring. So they've made a real time video based tutor that basically looks and sounds exactly like the founder of this company that you can ask questions to in real time in either voice or text at any point while you're studying. And it'll respond within a second. In speaking, it'll respond by speaking and explaining to you what you asked it in a second. It is nuts and it's live. So by the time you're hearing this, this is something you can actually go try for free right now. I have a strong feeling. This is basically the future of what you know AI tutoring will look like. So it was really thrilling and he had a great story. So that's Matt Sherman from LSAT lab. You're about to hear our interview with him. In funding news this week. We saw ed tech startup next wave that's next wave with Noah II, an Indian startup that does certification upskilling and CCVP certification programs, software professional development, raise $33 million in fresh funding. We saw ignite reading which is SF based virtual tutoring company for first to eighth graders raise$10 million. We started midcard as we mentioned in our India segment raise $6 million, just about $6 million. Led by GSV ventures. We saw a company called House of math I love that name out of Oslo, Norway raised four and a half million dollars. That's all about great appropriate math and physical exercise activities, which is pretty interesting combined math and movement. And then lastly, we saw the Moldovan edtech company balloon coding get a 1 million euro funding push to expand their coding within virtual reality in the Central European region. interesting set of funding rounds this week for sure. For our deep dive this week, we're talking to Matt Sherman. He's the founder and CEO of LSAT Lab, which does obviously LSAT and standardized test, tutoring and training. He has done something extremely interesting with AI and we want to you know, the tech insiders listeners to be aware of it and to go try it out. So Matt, welcome to Edtech insiders. Thanks for having me. So tell us about LSAT lab and what you've been doing with real time AI
Matt Sherman:LSAT lab is an LSAT test preparation software tool that is focused on very personalized learning. And when opened a launched last November 2023, we saw a real opportunity for incorporating the AI into how we create content, both in terms of the way we write explanations, but then it started to become more apparent that we could go even further with real time video and when we saw that real time video was an option. We have implemented a real time video AI tutor
Alexander Sarlin:Yeah, so for people People who may not have seen anything like a real time video tutor, explain what that looks like. So it's video and it's actually a avatar that looks like you in this case, right? It's trained, sort of a clone of you as an LSAT tutor. And what does it mean to be real time in this context
Matt Sherman:It means that you can talk with it the same way you would talk with a real tutor. So it looks very much like a video call like this does, where you can simply talk to the other side, and it responds. And it's perfectly trained to know what you're working on and the domain space so that you have the help that you need whenever you're working on content.
Alexander Sarlin:Yeah, so clearly, the difference between you know, a slow moving generated answer from an AI system, or even a video answer through an avatar. And this real time is the latency, it happens very immediately and feels very conversational, can you tell us a little bit about where this real time system is sort of how it's fueled, and what you've sort of seen in the early usage.
Matt Sherman:So it's fueled through a real time connection, but essentially, renders a video in sub one second period of time. So that, you know, depending upon the length of the video, you're gonna get different length latency while you're waiting for the response. But as long as you're below paragraphs, the latency is going to be very quick in the sub one second, the opportunity here is this still not perfectly conversational? Because at the moment, you can't interrupt it. That's the next thing we need to do is we need to make sure that if it's starting to talk, and you know that it's not talking in the right direction, you can just interrupt and say, No, you misunderstood me. But that means we need to make it able to listen to you, while at the same time that is talking.
Alexander Sarlin:Yes. So this is really interesting. And you know, I've seen a little preview of what you're doing. And what's so interesting about it is, as you say, it's like a video call, but it's also contextualized in your LSAT learning management system. So literally, you can be sitting there reading an LSAT question and like critical reasoning, thinking through all your stuff. And in the corner is you it looks just like you is it you know, at AI tutor. And while you're working on a particular question, you can just ask questions about it, you can ask anything, and this real time tutor will respond and support you. Tell us about how you've integrated this.
Matt Sherman:So then you can also generalize this. So when you're working on a question, not only can you ask it to explain what the conclusion of the argument is, or what's the assumption of the argument, or why this answer choice is wrong, but you can also ask it to explain the patterns that underlie it so that you can generalize and extrapolate from that one individual question, and go find more questions that you can actually go grab those questions and bring them to you at varying difficulty levels, so that you can not just understand that question, but grasp the concept that's underlying it.
Alexander Sarlin:I mean, that's incredible. I would imagine that two things are going through my mind. And I imagine many of our listeners are thinking this as well. The first is the LSAT and sort of, you know, long written argumentative content is a really terrific use case for this type of real time tutor, because you have to sort of unravel language and dissect it and sort of put together cases, but it's also extremely generalizable, there's no reason why the same type of tutor trained on the right data set might not work in a math context or a chemistry context, tell us about and sort of how you envision the future of this type of real time tutoring,
Matt Sherman:I envision that everyone will have access to many different tutors that serve various needs that they may have, that those tutors will be trained on all sorts of learning disabilities and special challenges that students might face. So that students can get the help that they need when they want it. They don't have to schedule a meeting with a tutor, you know, once a week, or they can just whenever they are working on something they can get help real time and it can move from, as you said, concept to concept from math to science to test preparation.
Alexander Sarlin:I mean, it feels like it's really getting closer and closer to that science fiction sort of vision that people have had for you know, since the Diamond Age and all of these ideas of literally having a partner with you while you do your schoolwork. You can ask it anything it can support you it's not going to tell you the answer. It's going to help you learn how to think and dissect and make sense of things. That is thrilling for somebody like me, who is so excited about the potential AI in education. I feel like this is really a step change. Just the idea that it's constantly there, that it's accessible and that it's sort of ready to answer any questions, and it's affordable. And it's tell us about the affordability person.
Matt Sherman:That's the thing is like, if you were going to go to one of the premier or elite LSAT preparation programs, it's gonna cost you a couple $100 an hour for a tutor, right? But with this, the cost is cents for a minute of time. You can get that entire experience for a couple of dollars.
Alexander Sarlin:That's incredible. And you know, those of us who have sort of been through the video revolution in education, we're always excited about hey look, you can get a video tutor you can watch a great professors online and you don't have to spend you know, 50 grand a year to go to a college but this is such a different level because it's completely interactive. It's totally responsive and personalized to each learner, you can ask it anything you'd like. And it's just totally different level. So one thing I have to ask you, and I know you've thought a lot about this is because this is open to asking it, students can ask it anything, how do you make sure that it is appropriate? And that it, you know, students can't go to it into saying something crazy.
Matt Sherman:I mean, so if you were to ask it questions that are unrelated to the outside, this is going to guide you back to the OSEP. That's simply prompting, and tuning. There's lots of ways of training these models, both deeply and superficially in order to make sure that they remember to do what you're asking them to do. To context is everything.
Alexander Sarlin:Would you say that there's also a little bit of a, like a stack issue there? Because like, because you're working on top of existing foundational models, some of the training and the foundations also protect against inappropriate responses, or, you know, giving away the answer to moms or hallucinations. Tell us about that. Because I know that edtech founders don't want to be responsible for making sure that their AI bot is completely failsafe they want to make sure that they're inheriting we want to make sure that we're saying, Oh, of course, but you don't have want to do it alone. You want the underlying technology to also do some of the lifting there, right?
Matt Sherman:Sure, sure. So we prioritize the most accurate base models that we can find. For speaking with students, there are other contexts where we might prioritize speed over accuracy, or clarity or whatever other factors might be important, depending upon the context. And this context is typically accuracy. We're going to focus on bass models, and then tune them with certain characteristics and personalities, and that we would like them to have where this is going, I believe, is every single gets their own model, and that that model is continuously learning. So the tuning the fine tuning process, we can tune simply every single time the student interacts with the models, constantly tuning the model means that that that that tutor is now built for that student.
Alexander Sarlin:We talked a number of months ago to Boto Hernan, who does a company called knowledge and he had made a digital twin of his kids, he has twins, and he has two kids, and each of his kids has a digital twin tutor that looks just like them and has learned and been trained to respond to them with their interests, and everything that you know, they how they learn what they care about. And that felt like a vision of the future, combining that with this real time always accessible, affordable model that you're pioneering here, it really feels like a new world, because you know, that was true, but you'd have to ask at something and then wait for it to generate, and then it would be, you know, text generation, and then the text would be spoken. So this is just a whole new level.
Matt Sherman:The thing is, I feel like this is just a sidestep into this artificial world where students take this as I want to give them the flexibility so that they can have the tutor that they would prefer, whether that's a real person, or whether that's a cartoon character, whether that's, I think that there's a lot of opportunities to explore with the presentation of your tutor, and the personality of the tutor. So that, you know, maybe we can, and the language actually, this is one of the amazing things. It speaks in 29 different languages. So if you're, let's say a Brazilian student, who's working on asset, you can actually ask it to explain things in Portuguese. And it does an excellent job,
Alexander Sarlin:guys, just saying the level of personalization from a language perspective, from an interest perspective, from a racial and gender perspective from a you know, what type of teacher you like, should they be funny or serious or, you know, personality perspective, it's just totally endless. So last question for you. This is live. And I think our ed tech insiders listeners would love to see if they can try it out. Where should they go? If they want to do a little of playing with your real time tutor,
Matt Sherman:you can try it for free. Just go to LSAT lab.com. create a free account. And on the homepage, you'll find dashboard where you can try the AI tutor in a setup where they can work on various exercises, essentially an arcade of various games with different activities and levels and the AIs are different or managing the process. Everybody
Alexander Sarlin:should check that out whether or not you're preparing for the LSAT, go see this technology in action. It's really thrilling. Thanks for being here with us. Matt Sherman, LSATlab.com really pioneering the future of what AI tutoring will look like.
Matt Sherman:Thanks, Alex. Great to be here.
Ben Kornell:And that's our show. Thank you so much for joining us this week. If it happens in edtech. You'll hear about it at ed tech insiders.
Alexander Sarlin:Thanks for listening to this episode of Edtech Insiders. If you like the podcast, remember to rate it and share it with others in the tech community. For those who want even more Edtech Insider subscribe to the free Edtech Insiders newsletter on substack.
