Shaping the Future Of Learning: Sam Altman on AI in Education

Edtech Insiders joined the first Common Sense Summit on America's Kids and Families hosted by Common Sense Media at Pier 27 in San Francisco on January 28–30.

The conference brought together advocates, researchers, youth leaders, policymakers, and other experts to take stock of America's kids and families and explore solutions to the most pressing issues across four core topic areas: kids and technology, youth mental health, early childhood education, and K–12 education. 

In this special episode, Alex and Ben interviews Sam Altman, CEO of OpenAI.

Transcript

Alexander Sarlin: 0:04

Welcome to Season 8 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: 0:55

Hi, everybody, it's Alex and Ben, and we have a special guest who needs no introduction. Sam Altman, CEO of OpenAI, welcome, Sam.

Sam Altman: 1:04

Thanks for having me.

Ben Kornell: 1:05

Sam, it was wonderful to have you at the Common Sense Summit, where you talked about learning and AI. And you talked about several use cases in education. Is there a reason why you think AI as a technology is particularly relevant to education and learning.

Sam Altman: 1:21

This has been one of our most important areas of adoption so far. And it's one of the areas that I'm personally most excited about. Today, we're already seeing AI help teachers, personalized lessons for students and things like that. It's it's doing a lot to reduce friction for non English speakers. But in the future, we hope and we really believe based off what we're seeing get built, that AI can become a helpful, personalized tutor that will help accelerate students learnings in a way that is only accessible to a small number of people today, you know, we know what impact one on one tutoring has, but most people can't get access to it. And if we can democratize that, and if we can make that something that's widely available to learners, I think that'd be tremendous.

Ben Kornell: 2:00

Is that mainly through API's and third party partners and developers or you imagine that open AI itself or, you know, chat DBT? Some people aren't using as a tutor as we speak? How do you imagine it playing out?

Sam Altman: 2:13

Some people do use chat to be tea that way, which I think, points to how powerful this can be. And it shows that the technology is capable of it. But I'm very encouraged by the energy of developers in what they're building here. Yeah.

Alexander Sarlin: 2:27

So the partnership you announced at the common sense summit includes curation of the GPT. Store. And you know, we've played with GPS, we've made several GPS, they're amazing. How do you think about how students, families and educators are going to use the GPT? Store in the future? Will they be creating GP T's? Will they be consuming them from professional organizations? Or from individuals? Or both?

Sam Altman: 2:50

How do you envision the future they're all of the above, one of the cool things about GPT is, is it's an early example of one of the things that is interesting in new about AI, which is you can program a computer just with language. So anybody can create one, you don't have to be developer enough to know computer science, you can just you can just make one or you can use other people's and you also see this thing happening when people inspire each other and then build new things. They build tools for themselves, and then they realize other people might want to use them. And you can imagine teachers, textbook makers, ed tech companies, building campaigns up Ts to help students learn, you can imagine students building GPS to help other students. We've heard all sorts of things from teachers about using chatbot T to build quizzes and lesson plans and curriculum materials with chatty PT, and they're now building GPS do that more easily and to let other people do it as well. And I think what we're going to see is this idea of people taking what they know about the world and what other people want, and tasks that are important. And making GPT is to help other people do that more easily. It's going to be quite promising.

Ben Kornell: 3:52

So from a edtech entrepreneurs standpoint, how should they be thinking about the TPT? Store? Is that the place to try and build your you know, the one? And then eventually you have your own app? Or is this the place to launch new features? Do you imagine the storefront is kind of the consumer facing side for entrepreneurs? Or are people essentially using their own site and their own product as the front door long term?

Sam Altman: 4:18

It's very early days for the platform. I think we don't know yet what the answer to those questions will be. It's still in there, like lots of rapid experimentation phase.

Ben Kornell: 4:28

Super. So you mentioned the common sense partnership, but you also have partnerships with Arizona State University with Khan Academy. I was recently at an event with Teach for All which is a global education organization where they're doing some interesting things with open AI. You've really leaned into this partnership approach with education and education organizations, including these nonprofits. How does partnerships play into your long term strategy for both access and outcomes with AI and education?

Sam Altman: 5:00

Yeah, as I mentioned earlier, education is of special importance to us. And we're delighted to see what's happening already and what people are building for the future. I remember once, Alan Kay said to me that if we could just solve the problem with giving every kid on Earth, a great education, we would solve all the problems. And that kind of was one of those little clips in a conversation that just always stuck with me. But I think it's quite profound. And so we have been very interested in how this tool can get used for education. And we know that we need partnerships for that we are not educators, you know, we value common senses expertise, we're thrilled to be partnering with you all, we value what as you mentioned, Khan Academy and Arizona State University, are doing to start to get to watch these tools really enhance learning to really help provide tutoring to students that otherwise may not get it, it's a tremendously rewarding part of what we do.

Alexander Sarlin: 5:54

You know, we have noticed from our vantage point as edtech observers that you've partnered with very innovative and very access oriented organizations, Arizona State, you know, famously, has been trying to do incredible high quality education at a much more affordable price. Khan Academy, of course, tell us about that sort of lens on these partnerships. How do you choose these particular organizations? And what do you value in their approaches?

Sam Altman: 6:18

You highlighted, I think the two key points, we need innovative companies, as early partners, as someone as you do in any field, you know, you want the people that are going to kind of like jump into the void with you and say, Hey, this is new. We think we'll get going. But people that are focused on access, that's a really big part of what we hope for here, which is then a big part of the story of AI and education is going to be about access.

Ben Kornell: 6:40

Yeah, I will just say, even one year into a little over one year into the explosion of chat, GBT, we're hearing stories where in Haiti 95% of the world's content was inaccessible due to translation. And just immediately being able to translate anything in the world into Haitian Creole is just an incredible unlock. So I'd say early signal is that AI is really unlocking access. And then on the assessment side, we're also just seeing incredible amounts of data that was unstructured data, and very hard to assess before that now can be structured. There's a company called Dewey, that is basically pulling together data streams from all these different silos, and putting into a dashboard where an educator just has to use a search bar and say, what math subjects do I need to reteach next week based on students scores, it's pretty incredible what we're seeing, given that landscape change, we're thinking about early stage entrepreneurs and before open AI, you lead Y Combinator, where a meaningful part of the startup portfolio was related to EdTech. And education, podcast friend, Dan Carroll was one of the Y Combinator startups, given the new capabilities offered by AI, what advice would you have for founders looking to start a new education company? And is it any different than the advice you would have given them from your Y Combinator days?

Sam Altman: 8:05

I mean, the first thing I'd say is, I think this is just the best time to start a company in a while, is the kind of platform shift that startup founders should wait for, and hope for, because it really is when the ground is shifting, and startup founders can do something amazing that, you know, the big companies have a harder time with. So this is a tremendously exciting time, the biggest change to the advice is just that this tool is so much more powerful, I think, than the tools of previous new waves, new revolutions, that what a small team can do quickly, is more than I've ever seen it before. And so letting the scope of your ambition grow early and really think about what you can do. That's what I would say.

Ben Kornell: 8:42

Just a follow up on that. Do you think that the VC path where you have to raise huge amounts of money, that that's no longer necessary, given that you could have a micro company with 10 people 10 million in revenue and, you know, a million customers?

Sam Altman: 8:58

Yeah, it's definitely less necessary than it's been in the past if you're monetizing early, but if you're not, if you're gonna sort of follow the traditional consumer thing of get big and then figure out a monetize, the compute costs can get pretty big for AI. So it can kind of couple of things.

Alexander Sarlin: 9:12

I wanted to double click on what you said earlier about students being a potential user of GP teas and creating these user generated agents. One thing that is really exciting about this new world is because coding is now accessible, you can create tools with using natural language. It allows kids and students to do incredible work. I'd love to hear you talk about what the you envision the future might look like in a world where you know, a high school student can create an incredible piece of software that basically can be used all over the world.

Sam Altman: 9:42

You know, when I was a high school student, I made a little piece of software that got a lot of use, and it was this incredibly gratifying thing, the feeling of getting to create something and watch other people use it and talk about liking it is like really like, that was like a formative experience for me. And I'm excited for more people to get to experience that like one of the great lessons in life is that you can, you can create novel things that are valuable for other people. And that that is kind of how the world goes round. And I'm very excited for more students to get to do that.

Ben Kornell: 10:12

When you were a high school student in St. Louis, not in the heart of Silicon Valley. And I think what's really incredible about the toolset is that you can literally be anywhere in the world and create a tool and a product that can be game changing for users all across the world.

Sam Altman: 10:29

Yep, totally.

Ben Kornell: 10:31

So let's talk a little bit about that compute power you were talking about. We recently published an article where we were pushing for 1% of the world's compute, to be donated to charity to charitable purposes, do you feel like compute it is going to be the rate limiter of our progress for AI, but also, especially around social impact in AI, because a lot of us in the space are just concerned about the spiraling costs, and also the competition for compute makes it really hard for edtech players.

Sam Altman: 11:04

You know, the rate limiter or for a significant impact? I don't know. But it's at least going to be hugely significant. I think the best thing to do is to create as much compute as possible to drive the price down, like we don't want this to be in its current world have a great deal of scarcity. I think the more abundant we can make it, the cheaper it'll be. And I am worried about it. I think it's important to get that right.

Ben Kornell: 11:25

Can you talk a little bit about synthetic data, I mean, there's a way in which we almost already eaten the internet, in terms of MLMs. And you've done a really good job of fostering partnerships in New York Times recently, with existing content creators. There's a lot of excitement around what synthetic data could do to train MLMs. But also concern, how do you imagine synthetic data playing into our future?

Sam Altman: 11:51

I haven't heard the concerns. But I'd be curious to hear from you what they are. I mean, I think synthetic data is a fairly overloaded term, it can mean a lot of things. But somehow this idea that we're going to have, LLM 's learn more from less data feels obviously correct and important. And I think that's just good.

Ben Kornell: 12:08

Yeah, that bridges me to talk a little bit about Sora, your most recent announcement. Now, to be clear, it's only available for researchers at this point. But I saw your tweet storm where he were soliciting inputs. And it was incredible. And you know, it just feels like we're in an era where every once in a while we pull back the curtain and get a little glimpse around the corner. It's felt the same when I saw the vision Pro, it wasn't like, it's here, but it's like I can see around the corner. Now. Can you talk a little bit about how video and also the way you rendered with sore how that might speak to where this space is heading.

Sam Altman: 12:47

I think multi modality is important, like humans do not only interact with text, we get videos input, we certainly like to consume video data. And there's a lot to learn from video. So as we think about the richness of experiences that you can imagine someone wanting for, say education, there are times when watching the video is the best thing to do. Or there are times when you need to, like learn from video. So we're interested in having these models not just be text only can we can provide a much richer experience. Ben

Alexander Sarlin: 13:19

likes to use the term Omni modal to sort of represent this multimodal future of AI inputs and outputs. And I mean, one one use case we've seen that's very interesting is that for students who are preliterate for any reason, you know, being able to use voice inputs, or pictorial inputs, or you know, all different types of ways to interact with an AI can really unlock usage because they don't know they can't do a prompt in text yet. I'd love to hear you talk about you know that and how you see the future of that type of interaction with really

Sam Altman: 13:51

The accessibility of multi modality or mobility is very important. Speech in general, I think is going to be a big surprise to the upside of what that but this idea that you can talk, like real time two way talk to a computer. I think that's gonna be quite powerful and open up all sorts of new things.

Ben Kornell: 14:09

We only have time for a couple more questions on the Omni modality side. Part of why we're excited about Omni modal is it's not text to speech, text to video, text image, it's actually image to image or image to video or speech, you know, it basically can flow any which way. But that also makes it harder to create guardrails and safety and tuning because once you move out of natural language, you're moving into more abstract concepts. How do you think about safety and tuning and red teaming? These kinds of products that go beyond just natural language?

Sam Altman: 14:45

It's definitely an additional surface but I'm confident we can address it. We've been approaching this slowly. We didn't images first and then we have to confront some of the things with audio. Now we're confronting challenges with video. And this is why we go slowly. It's why we do a release to read tumors first, because there is a lot to learn about how people want to use these what you need to focus on for safety issues. But you know, it's additional work, and additional if to engage sort of with more complex issues.

Alexander Sarlin: 15:12

One thing that's been so exciting to watch in open AI a story is that Chad GBT is the most popular product and arguably the history of the world, the fastest growing, and you and your team have really spent a lot of time making education partnerships and talking about the education use cases and the future of education. I'm sure there are people knocking on your door from every industry, from finance, to medicine to anything. I know you started here, but I'd love to hear you talk a little bit more about why education has really been such a, you know, a primary sector for the future of Gen AI and why you've spent time there as opposed to you know, any of the many different ways that we might go.

Sam Altman: 15:52

Great thing to close on two things. One, the technology in its current state works well for it. So it's like ready, you know, you can go build these valuable experiences these helpful experiences now. And to is it's so closely tied to our mission. Like what a great thing if we can actually make an impact here.

Ben Kornell: 16:09

Wonderful. Well, with that, we will close our interview Sam Altman, CEO of Open AI, thank you so much for joining edtech insiders today, we really appreciate it.

Alexander Sarlin: 16:18

Thank you very much. Thanks for listening to this episode of Edtech Insiders. If you liked 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.