We have to really convince, I think, the rest of the world that there is viability in even products that do rely on, as you say, that, you know, traditional education system, whether that's K-12 or higher ed, it feels like the narrative has taken over that like, you know, this isn't really tech in the way that we've all wanted it to be. It's more education and education gets time. Education service, education services, exactly.

And I will say to try and be a little bit more bullish on this one potential. Salvation feels a little heady of a term, but one potential positive avenue for these companies is the emergence of smaller language models and the on-device ilk language models, where users are going to have a more personalized implementation of these language models. And so I know one of the topics you wanted to talk about today was Microsoft's release of, I don't know if they classify it as small, but an on-device language model. If you want to set some context there, we talk about how that tech

companies might be able to leverage those. Yeah, you're right on it. I mean, Microsoft launched something called Phi3. It's an open LLM, importantly, an open LLM, that is designed specifically to fit on smartphones with limited resources. It's low latency and it's meant to be actually, you know, work on a phone without needing to do, basically, just quickly and powerfully and live on phone. And then we also saw in Thoropic this week launched their mobile app for their cloud, you know, AI and these sort of business services as well. So, you know, it hasn't been that long. I always like to remind myself and everyone, but we're starting to see the LLMs follow us into our daily lives in a lot of different ways. And this feels like a jump for mobile. So what, yeah, what do you think it means for ed tech?

I think it could be positive. And, you know, it was a week or two ago. Also, Sam Altman said, large language models are dead. It's now about small stuff. And, you know, I think my hope and, you know, we'll see if this is actually the case is that the smaller and on device models will allow basically they'll do that personalization part for other applications and allow the applications to tie in nicely with them. So, you know, the language model on your phone will know about you and have your like characteristics that, you know, today you would be having to put into the context window when you do your prompting. I hope, again, this is sort of the start of a steelman, not a perfect argument that applications. So a cheque or a to you or a Coursera or, you know, startups will be able to leverage those on device characteristics to build education applications and, you know, every other type of application. Right. So hopefully this applies across the spectrum of on phone apps, but that should allow a much higher fidelity experience, both with a language model and with the actual educational application, while still allowing for other applications to thrive, you know, and not saying it's all going to go to the small on device. Language models, they're not going to handle every application use case, but they will provide wind in the sails of other applications.

I mean, a noneducation example that jumps to mind for me because I was just discussing this with my wife this week is that, you know, at this point, we all have so many photos that we just don't even know what to do with years and years and years, especially people like you and I who have young children have ridiculous numbers of photos on our devices and we carry them around with us. Some of them are in the cloud, whatnot. You could pretty easily imagine a very useful use case for LLM's and, you know, AI in general, but especially that sort of personalized LLM where you say, I want a cute picture of my kid to send to grandma and it goes, okay, I know what you mean by cute. I know all your pictures, you know, I know all the pictures on the cloud. I know what you've done in the past. I can sort of make sense of that using generative AI and, you know, just sort of data mining. And I can pull out the photo you actually want or you could even say, hey, I have too many photos. Can you suggest, you know, 500 that I could probably delete and it will go, yeah, I can. And I know what that means for you. I know the kind of photos you don't want. And of course, we can then transfer that to the education use case and there's a lot of different interesting ways to look at it, including just curiosity of knowing what to search or being able to sort of recommend based on where you are or time of day or behavior in apps, you know, things that you might want. To learn or that you might be missing in your understanding of what you're doing at any given moment. There's some really, really interesting educational use cases that sort of could put the phone, which is in our pockets all the time, into our educational lives in a different way than it has. It's hard to imagine right now, I think, but you know, it's getting there.

Well, and just to extend the analogy, like imagine a world where, you know, in WhatsApp and, you know, I know WhatsApp and Apple are both big tech companies. So we'll see if the analogy works. But in WhatsApp, you can ask your on device bot to, you know, send a photo to grandma once a week, right? And you know, have that be part of your ongoing conversation with that person, right? And for education, you know, there's been so many articles about like, do phones belong in the classroom? Can you bring your phone to school? Yadda yadda yadda. You know, you can start to imagine a world where your on device assistant bot, whatever you want to call it, is capable of discerning like, oh, yeah, Matt's at school. Probably he should only be able to access these apps. And so the guardrails become even more intelligent, right? And the schools have a say, the parents have a say, like, those types of intelligent interactions that like, you and I can come up with on a piece of paper, but could never implement on a device today. I think we're, you know, not

too far away from being able to. Yeah, the interactivity between apps between the apps and your regular daily life. There's a lot of potential here. You know, I could imagine your sounds a little scary, but you know, you're on a phone call and you say, oh, you know, I wonder what the weather is going to be like when we go travel to Tucson next week. And it goes, ah, got it. You're going to Tucson next week. You're wondering what the weather is going to be like. I'm going to figure out, you know, everything you need to know about this schedule accordingly. Maybe I'll help change your your hotel plans or whatever you make that restaurant reservation indoors.