Greater Distribution
Our podcasts are also now available on Apple Podcasts and Spotify.
Today’s Podcast
In today’s podcast, which was recorded December 14th, we look the the following -
The significance of AI, as signified by Time choosing the “AI Architects” as the “person of the year.”
“Step level” increases in GPT5.2 to match Gemini3, and the consequential GDPval benchmark that shows AI doing 40 common office jobs as well or better than humans 70% of the time at 11X the speed and 1% of the costs.
ChatGPT comes up with a new idea, other AIs test it
The brewing backlash against AI and educational administrators caught in turbulent times
Forthcoming model releases, especially Grok 4.2 (soon) and Grok 5 (January)
Advances in multimodality and physical world modeling
AGI projections from the top researchers, including a 50% chance of “minimal AGI” by 2028 and a rapid take-off to AGI where AI builds AI starting in 2027.
Rapid social and economic change being triggered by current AI levels+ AGI.
Android real-time language translation for 70+ languages.
Robotics and robots growing food in shelves in major cities
The importance of skills relative to college degrees and the soft/durable skills
Concrete suggestions for schools, especially universities
Our predictions for 2026, including the difficulty of adapting school curriculum, “a real scramble in K-12 to prepare students for an AI world,” and the likely certification/accreditation of AI-taught courses and universities by the government.
Anand (00:00): Well, welcome back, everybody. This is our semester in review, and we’re looking back over the last semester. This has been, of course, the first few months that we’ve had the podcast—lots of guests, lots of updates. We wanted to give you some updates at the end of the semester, but really reflect on what we’ve uncovered, what we’ve learned, and tell you a little bit about where we think we’re going, getting ready for the next semester.
Stefan (00:21):
Yeah, it’s been a great semester, and doing this has kind of forced me, in a way, to keep up with everything. So, it’s been exciting and it’s been a great conversation.
You know, so one thing... you know, we’re really looking at a world of pretty dramatic change. And you can see that with the Time “Person of the Year.” Usually, it’s an individual; a couple of times, I think, it’s been more than one individual. But here we have all the architects of AI, and you can debate whether every... you know, all these people belong or they should have added some others.
But I think the major theme is when you look at kind of a world of change and the “Person of the Year” who’s bringing that change, they are all the architects of AI. So, I think this podcast is one of the most important things you can listen to because we’re definitely entering a world of change.
Now we see some really important quotes from some of the luminaries and leaders in the field. Ray Kurzweil is certainly one of the very top, and for decades he’s been talking about the singularity and where we’re going and how this is going to be developing. I mean, he said the next 10 years will bring more change than the previous 100 years.
And we’re starting to hear this from more and more leaders in the field, including from Ethan Mollick, who has certainly become more popular and famous with the development of AI since ChatGPT. And one thing that he has really underscored that I think I’d completely agree with, and that is: if we completely stop the development of new AI, then if we just explored how to roll out the use of the current AI models that are there, there would be massive disruption across society and the economy for easily the next 10 years. So that even if we didn’t develop any further, if they somehow stopped it, there would still be so much more to explore. And of course, that stopping is not happening. I mean, it’s certainly moving forward faster than ever before and full steam ahead.
Stefan Yeah, and you know, we can see these are the major models. Everybody’s familiar with ChatGPT. Of course, we’ve seen a lot with Gemini 3 popping out a couple of weeks ago. They said now that was the leader. ChatGPT-5.2 came out, and it seems pretty similar. Then there’s Claude, which you and I use a lot, which I think is awesome in many ways. It’s known for coding, but I also think it’s great for academic research and writing and some other things. And then of course there’s Grok. I think they’re up to 4.1. I think I heard 4.2 is due out very soon and I guess they’re going to have Grok 5 in January.
We really kind of see an escalation of the models. And to tie it back to what you were just saying about the capabilities, that’s where we really start to see—and we’re going to talk about these—some improved capabilities of the models. Because for some people, I think that the benchmarks are a little bit abstract, and they’re starting to get way beyond what almost anyone will use them for.
One of the capabilities... and this math, as you know, because you used to help me with my math, this math is kind of way beyond me. But the idea of what’s kind of being developed here, the concept, right, is not. And we saw somebody, a physicist, publish a new paper where they said, “Hey, was [it] ChatGPT-5?” Or not even like the 5.1 or the 5.2. They came up with this new idea. And then this physicist worked with Gemini and I think Claude as well. Then they published the paper here.
Right. So this is important because people were just saying, “Well, the AI is just based on the training data. It’s not generating anything new.” And now this person who’s really qualified and who wrote and published this paper about how, in fact, it did come up with something new. You don’t have to understand the math or the physics behind it to just get that this is a new idea.
Anand: Then there are also developments in image and video. You see this picture was making its rounds. And it certainly looked very realistic of all these AI tech leaders together. And of course, it was generated by AI. And along with that, there’s been the recent development that Disney has signed a deal with OpenAI that allowed for a billion-dollar investment in OpenAI by Disney. And that it also allows for the intellectual property of Disney—not the voice actors or any live action, but for a lot of the other characters you know, from the animated characters, Star Wars characters, Marvel, Pixar—to be incorporated into AI videos that OpenAI has on Sora, and that’ll probably be developed in some other ways, too.
This was pretty groundbreaking because, at the same time they were making this deal, Disney was also leveling a complaint against Google for IP infringement. We’re gonna start to see this carve out a little bit further. We’re at the stage where major industry leaders and some of the studios are starting to figure out: how can they use AI, and what agreements are they gonna be making? As they see that the development of the technology is becoming so good. And there’s just so much competition in terms of what they’re going to be able to do in the market.
Stefan And of course, another big kind of application—which also kind of works with the video, but really of kind of multimodality, right? Originally, the language models were just trained on text. And then they’re kind of trained on some images and they’re trained on some video and they kind of artificially, so to speak, combine these things together. But now you have training that’s really just multimodal when it’s training on everything at once the same way.
You know, we essentially train on something, right? We can see a written representation, a visual representation. Maybe we’re interacting with something all at the same time. And, you know, I put this kind of image in before, but it’s really profound and I’ve been using it a lot more. I write a lot on Substack and I’ll just take my Substack posts—which are probably too long—but I’ll just kind of put them into Gemini and I’ll just say, “Hey, make me an image that represents like what it is that I’m saying.”
It really, in a way, does a better job at articulating what I’m saying than I often am. It just kind of really gets at the heart of the issue, and it just kind of lays it out. And I noticed other people... I saw somebody make a little video for a program they were promoting, and they were just blown away about how the images really just match the text, even though they weren’t doing any of that matching themselves. So this is another place that we’re starting to get into the applications.
Anand: And a couple of implications for this. One of the first, of course, with the multimodality, you’re getting closer to the way that the models are working to model the way that humans interact with the world. And that’ll be really profound when we get into robotics and certainly when we think about world models. And there’s a lot more to explore there. The second thing that Stefan’s pointing to, though, is that it really changes our understanding of prompting. It’s not really about prompt engineering for the exact image. It’s about having the model develop—I don’t know if you would say it’s necessarily creative—but I think there is some creativity there in terms of it taking a lot of text, analyzing it, and being able to develop some visual that would work well with that text.
And it’s really pretty good. Notebook LM does this quite a bit with infographics built in and slides built in now. And it’s doing a more abstract reasoning about: how would you represent this data within infographics, within visuals, within slides? Another level of capability that we just haven’t been accustomed to until the last month or two that is now available. And I think we’re only scratching the surface on what this might mean for the way we interact with AI. And it’s not just a matter of text-based. Now we’re able to do that in multimodal ways across the board.
And we’ve also seen the models... they’ve been multimodal, as we’ve just been talking about, but there are also companies working to model the physical world, which a number of people argue is essential for AGI. And one of the biggest companies, or most important companies in this area is called World Labs, which was started by Fei-Fei Li, basically the inventor of ImageNet.
They’re trying to model the physical world and everything you see in this video is from their new model. It’s almost like a video that you would see, but this is entirely AI-generated. And the goal is to understand physics, to understand reality, to understand that objects that are in motion stay in motion, that the computer can’t also... like, half the screen can’t be in the ground, which is what we saw out of some of the initial models.
And here you can see this really impressive presentation, which combined, you know, if you just think about the video models we saw earlier and the images in terms of kind of what you can produce—not just to get to AGI, but in terms of maybe like a movie.
And the idea of a World Model really does change then the way that the LLM, the large model, is interacting with the world and understanding the world. And when it gets to the physics of the world, that’s when you’re gonna be able to do a lot more with other interactions with the rest of the world, certainly for robotics, but also to other types of representations.
Now moving on to Vibe Coding. This is certainly something we’ve talked quite a bit about. It is astonishing to me how quickly this is changing and how far it’s advancing. So that you get to the stage now... this is an example of something coming from AI Studio. We had a class that we taught this fall that was with students at UMW and they were using AI Studio to build apps that really were very advanced, incorporating lots of different tools on the Google platform. And of course, you have lots of other vibe coding platforms out there too.
Stefan: Yeah. And then, you know, we talked earlier about kind of just some of the benchmarks, right? And the companies are always battling for the benchmarks to figure out how they do on the benchmarks. Some people criticize the benchmarks. You know, the companies kind of game the benchmark. Maybe that was like, well, it can do really well on the benchmark, but what can it do kind of out in the real world?
And this is ChatGPT-5.2, which just came out, I think it was last Thursday. Yes, we’re recording this on Sunday, the 14th. And it beat Gemini Pro, maybe by a slight margin on most of the benchmarks, but what really did exceedingly well was on what’s called GPQA (or GDP Val), where they take like work tasks, jobs, and they say, “Well, can it perform this job? Like, perform it as well as a human?” And one of the things they did was they tested it on 40 different common office jobs. And they found that it could do it 11 times as fast as an individual at 1% of the cost. 1% of the cost! And 70% of the time, it did as well or better than a human.
And a lot of things that people aren’t considering is: okay, all right, well, that’s gonna keep going up, right? It’s gonna get to the point where it probably will do it as well as a human basically every single time. But also your human’s gonna have to be a lot better employee at doing the job at that task if you’re willing to spend 99 times as much to keep them employed. Right, and then wait longer for them to do it. So cost is a variable, speed is a variable, 100% accuracy is a variable, but even then, you know, the human’s not getting it 100% right. So a lot of people say this is kind of the most consequential advance just out of 5.2. And of course, we still have a long way to go. We’re just getting started.
Anand: And it really is reaching a tipping point where we start to find applications, uses in the workplace. That’s one of the reasons why OpenAI developed this benchmark to be able to measure how well it’ll perform in the real world with actual knowledge work tasks. And as companies start to see how this is working and what they will be able to do with it, there is some lag time to be able to implement it, to be able to figure out how to make use of it. But I think we’re at that tipping point where companies are going to start figuring out where this is useful, how they can make the best use of it. And of course, there’s going to be an incentive to do it because of the speed and the cost.
Stefan: You know, so a lot of the robotics advances we’re not seeing here in the U.S. We are seeing more. You can see, I’ve seen friends post on social media little delivery robots that they see. They’ve had some humanoid robots out kind of walking around a little bit in New York City. But really we have the most advanced robotics coming out of China. And here we see a founder... actually, I guess, you know, whether we want to call this boxing or whatever kind of sport this is, kind of doing like combat with their robot. You know, these robots are getting very sophisticated. I saw another one the other day that had like 11 arms, so it could do all these things that, you know, just a humanoid robot with two arms couldn’t do.
I saw a video of a robot... it was basically in kind of a greenhouse. They’re growing all these crops inside, like under optimal conditions with the exact water and everything, the nutrients that need to be monitored. They’re all grown on these shelves. And then of course the robots take them and the robots harvest them, so to speak, and they package them up for shipping. And this was happening in Shenzhen, which is a huge city—if you’re in the US, think of New York, but even probably bigger and even more modern—and the food is being grown right there. So they don’t have to transport the food from far away. So it’s a lot cheaper. It’s very efficiently produced. It’s grown under maximum conditions. So we’re going to see a lot more robotics over the next year coming to the U.S.
Anand (13:34): Now in terms of some of the developments in research and writing, there are some really interesting ones. Certainly, I want Stefan to speak to the competitive debate aspect. He’s been writing quite a bit about this. But it’s also worth noting that what Google DeepMind has been able to do... they’ve been able to expand a lot of the deep research that they’ve done before within Gemini. And now they have a Deep Research Agent and they’ve built it upon 3 Pro. And it allows for really more intricate research projects and developing reports in more advanced ways. And then also deep search QA. And then also not noted here, but something else that’s really fascinating is the way that they’ve incorporated AI into Google Scholar. So that you’ve been able to incorporate AI into some really more advanced research tools so that students and faculty and researchers are able to incorporate it, become that much more efficient, and certainly reducing a lot of the hallucinations we probably would have seen with chatbots before. But it’s also allowing us to access more information than we ever were able to before.
Stefan (14:30): Yeah, you mentioned the competitive debate is secondary reading instruction. It’s just been kind of an article/blog post I’ve been working on. The reason I put it here in this part of the slide deck is that I actually use Claude Sonnet 4.5 for this. And I can’t read exactly at the top, I’m a little far away, but it went through like a couple of hundred different sources. And I’ve been using the Claude Sonnet 4.5 for research. I find the research is more comprehensive.
They’re also, you know, I haven’t really found any hallucinations. I’m still getting some of those in ChatGPT-4.5, but it’s really gone down. And you know, as you turned me on to, you can even ask to produce it as a Word doc, right? So now I’m getting my drafts in a Word doc and then I upload them into my Word Online and I do some of the editing there, which is really great. And it’s also worth mentioning here that ChatGPT-5.2 claims a 30% reduction in hallucination, which is really big. We’re seeing a lot of hallucination reduction in source quality improvement, both in the Gemini and in the Claude Sonnet 4.5, as well as ChatGPT-5.2.
Anand (15:39) So this is one that, Stefan, let me have you speak to this because I don’t think I saw this study.
Stefan (15:45):
Yeah, there’s this kit just came out, you know, as I was kind of reviewing some of the final slides for the presentation. There’s a new study and this is just the Gemini 3.0 Pro. So, you know, you probably can think maybe ChatGPT-5.2 can do this. It’s able to pass the Chartered Financial Analyst (CFA) Certification, which I guess is a very kind of high-end, very, you know, what it takes to do well. It’s a three-stage exam that tests progressively complex skills ranging from fundamental knowledge, which of course all the AIs can do, but also to analysis and complex portfolio construction.
And it says in 2023, it could answer a few of these questions, but it didn’t do very well at all. But now it’s scoring 97.6%, which my guess is kind of... it doesn’t go into it here, at least at Level One is probably higher than even what most humans are scoring. Again, I just have this because I want to see that it’s not just the benchmarks, but it’s actually the applications and kind of applying this into the real world where we’re starting to see abilities of these models.
Anand (16:52): Now, another development that came out this last week is that ChatGPT announced with Adobe that they’ve been able to incorporate many of the tools from Photoshop, Acrobat, and Express right into ChatGPT’s interface. And this shows another collaboration so that we’re getting to probably a little more advanced uses, really interesting applications. I kind of put this on par with what they were able to do with Disney in terms of incorporating intellectual property. Here, they’re incorporating tools so that it’s not viewed as just something that you have to go into another platform to make use of. I think we’re going to obviously see more of this if they’re able to incorporate specific tools that have a wide audience that you’re able to access through a particular AI platform like ChatGPT. This is just the beginning.
It is interesting to see because Adobe, of course, is developing a lot of its own advanced AI tools within its products. It is also important to note, though, that for this, they’re limited to just some of the more basic functions within Photoshop, but certainly ones that other users wouldn’t have been able to access before. And being able to incorporate it within an AI workflow is gonna be really interesting for developing more advanced visuals.
Stefan (17:58): And here you can see just kind of how it’s getting added kind of right in.
Anand (18:03): This is another story that was really fascinating. And I think we talked a bit before earlier in the semester about AirPods incorporating translation. And if you had the right AirPod with your iPhone, you’d be able to listen to somebody speak in another language and it would translate for you. And then it could even translate back onto the screen. Google has now leapfrogged over what Apple was doing by integrating Gemini into Google Translate. And if you have an Android phone with this program built in, you’d be able to use this through any headphone that’s connected through that Android phone. And it has nearly 70 languages. I think we’re going to see more and more of these Google... these types of translation tools available and really speech-to-speech capabilities are just getting better and better.
Stefan (18:45): Yeah, it’s something I think that’s the only thing that’s going to hit society. Wow, you know, I can talk to anybody. And, you know, we’re obviously seeing a lot of very kind of significant AI advances. And of course, people wonder... a lot of times people on the media—even me, right—kind of like, we’ll talk about some of the challenges that are being created. But there are a lot of benefits, right? Like we’re looking at huge improvements in healthcare with all the expanded diagnostic capabilities and treatments of disease and healthcare access. The cost of living will probably go way down because the cost of producing things like food is gonna really kind of start to collapse. We have fusion; energy prices are gonna decline quite radically. It’s possible—there’s a debate about this, I’m a little skeptical—that we’ll have greater leisure time, I suppose, if people only make us work three days a week, but they’re still willing to pay our health benefits. I’ll be surprised, but we’ll see where that goes.
And if we do have fusion, we could really kind of solve the problems of climate change. There’s definitely energy problems being created by these technologies, but there’s also a lot of improvements in energy efficiency, reductions in natural gas leakage. So we’re gonna see a lot of positive applications of AI.
Anand (20:00): But that’ll be important to be able to counterbalance some of the backlash that we’re going to talk about a little bit later within the slide deck.
Now here’s some of the representation of some of those changes. Jamie Dimon talking about shortening the work week, maybe down to three and a half days a week. You see some of what’s going on in terms of developments of AI within replacing workers, some of the proposals coming out of places like Amazon, and then also some of the fights about data centers that are occurring across the country and really around the world. There are a lot of developments here. We’re gonna continue to cover them, certainly in the spring.
Stefan (20:32): You know, one thing that’s emerging is people see... I think one thing that hits people is the data centers, right? I was even just talking to a medical student the other day and he’s like, “I’m prepared for my electric bill to go up.” You know, they’re building a data center like near where I live. So people are seeing that they’re hearing about the environmental costs. They hear stories, you know, not so much personally, but in the news about people getting laid off with AI and they see these technologies advancing and maybe replacing workers. So I’ve just seen more and more online, you know, that’s kind of indicative of a backlash, right? Like somebody prompting Hitler, right? “Use your brain instead.” There was a woman who kind of went after somebody on the New York subway because they were recording things with their Meta glasses. And I think we’re going to see an explosion of kind of Meta glasses and Google just announced glasses and all these other glasses people wearing them. I’m not sure that attacking the individual here is the solution, but I included it more as kind of just evidence of growing resistance to AI.
Anand (21:33): And there’s going to be continuing resistance. So this is a story that Stefan shared. Somebody posted about the “Time Person of the Year” cover that we shared at the very beginning of this and said, “There’s a non-zero chance that we look back on this the same way we think about the 1938, 1942 issues.” And those were the years when Adolf Hitler and Joseph Stalin were selected as Person of the Year.
There is growing resistance and reticence about AI. And I think that’s obviously going to grow. We’ll talk a little bit more about that coming up when we get into predictions, but I think as we get into some of the political changes, when we’re looking at a lot of societal changes, and there are unknowns and lingering questions about what that’s gonna mean for somebody’s livelihood, what that means for their environment, there’s gonna be a natural backlash, and we’re already seeing that build.
Stefan (22:19): Yeah, the final instance is resistance. This just really hit me. I saw this professor connected to on LinkedIn. He had posted maybe about a week ago about how he was helping his son or daughter with their schoolwork, using AI, and he talked about it in a positive way, right? Not just letting the AI do the kid’s schoolwork, but kind of helping and coaching them through it with him involved, especially because the kid was younger. And then he posted this message about how he got a lot of emails from colleagues that were really kind of not very nice. He got a lot of DMs, right? Just kind of people expressing outrage that he would even portray AI in a positive light. And of course, you work at a university full-time, start at a center. I’ve taught some classes online, including for you. And I think that environment is gonna grow. It’s gonna get more challenging because I do think people see it kind of as a bit of a threat to their livelihood and against the natural order of things, for lack of a better word.
Anand (23:20): And as we see more crises develop, more societal problems, AI is becoming intertwined in really all of them. And so it’s kind of unavoidable. I think we’re certainly going to see more of this and it’s going to be hard to navigate that, I think. It’s going to be very difficult to kind of disentangle a lot of it. So even if it’s not something directly caused by AI, AI is certainly getting connected in a lot of people’s minds about some of the societal pressures and problems that we’re facing.
Now looking at governmental changes and there’s some really important stories that have just been developing over the last week or two that are worth noting. This one is from a week or two ago and it was a discussion about how Trump signed an executive order to start the Genesis Mission, a new AI initiative as you see here compared to the Manhattan Project. And this is one that’s really meant to be able to really kind of turbocharge a lot of the efforts and the work that’s being done and very much in line with a lot of other of this administration’s policies regarding AI.
Stefan (24:17): Yeah, because you know... and also like on that other slide is just like, hey, they have a military chatbot, like GenMil, and they’re working with kind of JP Morgan to advance other jobs. Like, what are the jobs going to be about in the future? We’ll talk about that. And I think most provocatively, Trump signed an executive order that’s trying to prevent the states from regulating AI, and there’ll be legal challenges and those legal challenges will get detailed as to where the government has the authority against particular regulations or where the state kind of authority gets maybe retained.
But his goal is to kind of not allow the states to regulate AI at all. I know they probably wasn’t slowing AI down that much given the advances we’ve seen, but they’re kind of really taking a strong public stand. We don’t want anything that’s going to limit the development of AI. They did exempt like harms to children, which I think is good. I think there is an exemption. The states are still allowed to regulate data centers, but that doesn’t seem to be slowing them down too much.
Anand (25:19): Now moving on to universities and a number of important stories. Obviously, we’re looking at higher ed on this podcast. And there are a number of topics here and I’ll let Stefan kick it off talking about some of the most important topics that we’re seeing at the end of the semester.
Stefan (25:33): You know, as I kind of mentioned earlier, we can add a slide on this, but there’s obviously some resistance within higher ed to AI. But then you also have universities creating new majors. There’s a New York Times article about this, just as it’s happening. It’s in different universities, and some of those are technical. Others are kind of more like a combination of technical and then the humanities broadly speaking. So that’s one thing.
And then two, some universities are adopting kind of like the chatbots, right? We saw this earlier with like Arizona State. They were like all OpenAI, same with the California University system. Recently we’ve seen Pitt, where we were undergraduates and where you got your PhD, they kind of really embraced Claude. And of course, Dartmouth, where your son attends, they’ve embraced Claude. So I think you’ve been highly influential. I know you’re a big Claude fan. So I think we can see your influence there. But we see a lot of incorporation of those kind of technologies kind of right into everything.
And then people just trying to do more to promote kind of AI literacy and AI fluency. Ohio State, as we mentioned earlier in one of the other podcast editions, they want everyone to at least have the basics of AI. So we see schools adopting a lot of different things and they’re trying a lot of different things out. We don’t know where things will land. I mean, you could say, well, all this technical skills, maybe that will eventually be done with the AI, but students want that now. Where’s the balance going to be with the humanities? I think it’s good that a lot of different things are happening because not only will people not ever have the answer, but this is going to be a continually changing ecosystem for which there’s not going to be one answer or one thing for everybody.
Anand (27:17): You know, one thing that we’ve been doing at our university... of course, we have the new center. We also had this course “AI in Society” that was offered during the fall. Stefan was able to teach and co-teach a section of that with our friend, Alan Coverstone, a guest of this podcast. John Hines, another guest of the podcast, taught a section of it. We just completed that course this week and really had a wonderful capstone project where the students were working on selecting social problems and helping to find a solution and to develop a solution for them.
Ranging from dealing with AI synthetic relationships—really great work in that area—looking at the way that social media has led to doom scrolling and that affects social effects and effects on individuals, to looking at things like making other materials more accessible for some people, or another group that worked on developing an app that would help them find allergens within food products on a restaurant’s menu. And so we had some of those presentations. We had the finals this last week and they were able to give their pitches with some guest judges, including the provost and the dean of the College of Business, a really great event. And I think this is something we want to continue to move into: really thinking about the applied use of AI in some really limited ways and some responsible ways and to have those discussions about how that might play out and what you might be able to do with it.
I also included here the reference to NeurIPS—the Neural Information Processing Systems Conference. As somebody that we follow in another podcast referenced, this is kind of the Woodstock of AI. It’s the biggest conference in the world. Some 20,000 people, all foundation labs are there. And one of the places where traditionally some of the biggest breakthroughs have been released. I attended this year; it was held in San Diego, and I was able to attend as part of the education program where I shared some of the materials that we used for our Intro to AI class over the summer.
There’s obviously growing interest in that area, and in particular within NeurIPS, a lot of people coming from computer science. And the purpose of that education program was to find a way and develop materials—and they had an academic showcase for this—to develop materials to better explain and educate a general public about AI. And that’s something they want to be able to incorporate and support in all institutions of higher education, even making a lot of those materials available for free for others to be able to share. So this fits in with a lot of the other discussion about what’s going on. And now we’re talking about how organizations and the AI industry is working on how do they do that outreach to higher ed and how they support the teaching of AI literacy and AI fluency across the board.
I’ll let you take this one Stefan, what do you want to say about it?
Stefan (29:52): Yeah, okay. And you know, I mean, the other thing... we’ve been doing the weekly updates and this has been a culmination, but we’ve really had some amazing guests on the podcast who covered a lot of different topics that we’ll get into. But I’d like to just take a minute to thank everyone for coming and sharing their perspectives. I’ve had a chance to learn a lot from them and I hope you have as well.
Anand (30:16): Yeah, some really great conversations. And if there are any of these that you’ve missed, I encourage you to go back and check them out because there’s a lot there that you can continue to mine. And having these longer interviews was really instructive and I learned a lot as well. Now we wanted to highlight some of the themes that were touched on in these interviews and there’s certainly still a lot more to talk about. But these are some of the topics that came out that we kind of reflected on and that we’ve been thinking a bit more about since we had those interviews.
Stefan (30:43):
You know, we’ve seen a lot of different focus on curriculum. Tina Austin talked about the “Reverse Bloom’s.” Adam Pacton kind of talked about maybe kind of reducing so much reliance on writing as a form of instruction and assessment. We have Patrick Dempsey talked about maybe some larger changes that we need to make in the university model itself. And Meredith Pratsky talked about how AI is really kind of an operating system. A lot of people still just keep viewing this through the lens of EdTech, another piece of technology. It’s not a smart board, right? It’s going to kind of change how society operates.
James Hudson really kind of talked about the importance of lifelong learning and moving away from the “4 in 40” model—Four years of school, 40-year career—to lifelong learning. We talked about improvements and students getting just some basic AI literacy and how transformative that is, right, as well as kind of these soft skills. And then Alan moved us in, really kind of talking about multiple intelligences and how in this world, education is really kind of focused on developing like a knowledge-based intelligence, but we really kind of need to look at it broadly.
John Heinz talked about the role of debate. We saw some initial efforts right out of Edinburgh University—close to where we grew up, very close, the next town over, so to speak, in Erie, Pennsylvania. And now it’s Penn West. And they’ve just done an enormous amount kind of starting from a shoestring budget, from kind of a very small center, kind of growing it out, growing awareness on not just their campus, but in their entire community, especially in rural areas. Tim Dasey, a lifelong AI engineer, talked about a roadmap for skills. Francis Hong at BoodleBox, who’s done a great job really making AI available in a lot of different ways to many universities and also in a way that many professors are supporting about deep engagement and just persistence explaining AI and working with people and understanding it.
Anand (33:07): Thinking about AI in 2026. And we want to start to make some predictions to look at some of the trends and some of the developments that we see are already underway. And starting with that are some of the investments occurring with autonomous agents, government actions, societal and economic disruption, university efforts, and also a concern for students who don’t attend university. And this is something that we really need to do a little bit more to grapple with for a couple of reasons. Obviously, not everybody attends university and that’s always been a concern and we don’t want to leave those students behind or leave them out of those opportunities. But we’re also facing some of the challenges in higher ed and questions about return on investment. And I think we’ll see more students not attend for a four-year degree. And so we need to figure out how do we make space for the kinds of AI literacy and fluency that maybe needs to occur in K-12 rather than just waiting until college level. And then also to think about other opportunities for education and for workforce development and for reskilling that should occur outside of a four-year degree.
Stefan (34:10): And we’re going to start seeing the AI in 2026.
We’re seeing a lot of advances. People talk about AGI (Artificial General Intelligence), human-level artificial intelligence. And there are some kind of technical advances that most people say that we’ll need. There’s a little bit of debate about this. But basically how it’ll happen... we need more memory in the systems. You can tell if you use these systems, they’re able to remember more about you. Like for me, they’re always like, “Do you want this as like a debate argument or do you want it to support your AI programs that you’re working on?” It kind of understands that, but we want even better memory of the context and longer over time. We want it to be able to learn continuously, right? Now it’s kind of like trained as a data set and then calls some tools from the internet to do some research, but we want continuous live learning, just like we learn continuously in life. And then, of course, as we mentioned earlier, we need to understand the physical world. But even once it gets to that, right, we’re going to have to have more energy to power all these models. And a lot of people say that energy is probably even more of a restraint right now than overcoming these, because a lot of people think we’re going to be able to overcome these within like five years or so, maybe a bit longer, but it’s going to take a much longer time to kind of scale up the energy sources that we’ll need.
Anand (35:33): The infrastructure is certainly something that we have talked about and it kind of spins your head when you think about the trillions of dollars being discussed for building out for energy support, for data centers, for other capabilities. But certainly, it’s going to be necessary if they’re going to be able to continue to build out these capabilities to be able to work toward AGI. And that’s something that we’re working on over the next year.
Now we want to talk about some recommendations and some directions that we think some institutions should be thinking about, and individuals should be considering as well.
Stefan (36:02): We have some recommendations, but like any recommendations, those are built on some assumptions, which I thought we would share with you to kick that off. The one is just we’re gonna see accelerating intelligence. I don’t think I’m personally qualified to offer my own prediction on whether we’re gonna achieve AGI that we’ve just been talking about, but these are what some of the leading thinkers are giving this timeframe, right, of around five to 10 years. And regardless of whether we achieve it in five to 10 years or it’s exactly AGI, it’s gonna just keep rapidly advancing, right? And the intelligence capabilities are gonna improve. But it’s possible that we could see breakthroughs in all those areas we discussed—memory, long-term planning, understanding of the physical world—kind of combined with all kind of all the models that we have now essentially to really get us somewhere close to what’s similar to human-level intelligence.
Anand (37:00): It’s also probably not going to be a eureka moment where we just announced, “Yes, we’ve hit AGI, we’ve hit ASI.” What’s been stunning to me is just the rapid developments, the changes in the models, how quickly and how accustomed we are now to expecting a new model to be coming out every so many weeks or every so many months. And it’s difficult just to keep up with all the developments. I think one of the concerns is that it makes us a little bit numb. It kind of desensitizes us to the changes. And it is astonishing to me when I look back at a project I worked on in August, let alone maybe a year ago, how much more advanced the models are now—the capabilities of the platforms, whether it’s vibe coding or researching or writing or doing other work with multimodal types of outputs.
It is just astonishing how much better they are now than they were just weeks or months ago. And if you start looking back at what it was a year ago, it’s really shocking. And we kind of become numb to that. We don’t really... we aren’t as aware of some of those changes. So I think what will happen is, as we’re getting closer to these benchmarks, we’re getting closer to these marks, we’re going to cross over that line at some point and look back on it and realize, “Yeah, that probably was where we would recognize AGI or ASI.” One of the benefits of all of these constraints is that it’s going to make it difficult to roll it out to everybody and all of the capabilities in all use cases at once, just because there won’t be enough compute to be able to support all of it at one time. But it’s still not going to be slow enough for us to be able to really think through what the implications are, how we’re going to be able to manage that and navigate that, because there’ll be some really profound societal changes as a result.
Stefan (38:32): And the second kind of assumption we’re operating is that there are some really important skills that students need to learn, which I think we’re almost... people are kind of in a way kind of coming to a consensus on. Is, you know, this first chart here is from the World Economic Forum, a brand new report that they released just in December, just a few days ago on the “Future of Jobs.” And they’re showing, if you look in the upper right-hand quadrant, like what are the skills that people think and companies think, industry leaders are thinking, thought leaders are thinking, are going to be really important in 2030, right?
And data literacy—everything’s going to come about data. That’s a top skill. Analytic thinking, creative thinking, communication skills, these are all going to be the fundamental skills. And we also see some skills that are declining. And I think that some of these, it’s hard for people to accept. One of them in the bottom left-hand corner is writing. I think writing will still be as important, but probably not as important as oral communication is going to become. We already see the power of writing declining. It even shows like a decline in kind of math skills. And I think obviously that there’s a little bit of tension there because, at the very top of AI now, right, you need these kind of extraordinary math skills to build out these models and types of things. But for the everyday person, I think you’re saying how much math a person is going to need to know is probably really collapsing just because of what the AI is going to do.
Anand (40:04): And certainly worth looking at the rest of the report because there’s a lot here to delve into as we start thinking about what we should be doing. But when you look at somebody like Demis Hassabis, who’s Google’s top AI scientist, one of the co-founders of DeepMind, says the future skill is really going to be learning how to learn. And this was an incredible point. And I remember having a guest speaker earlier this semester come to our class, a former student from UMW who made this point. He was asked, “What should we study as college students? What should we be ready to do?” And he said there are really two important skillsets that he could see. The first: strong communication skills, written and oral, be able to communicate your ideas well to others. He said the second was being able to learn how to learn.
Because he noted his experience. He was a poly-sci major. He now is working public affairs for a data center company. And he noted when he was in college, there were no classes on data center public affairs, or even data center development. He had to learn a lot of that. Some of the technical aspects, some of the developments, some of the public policy elements that really no one had written in a textbook before. He had to learn about that on his own. And Demis is exactly right about this. There is so much that’s going to change that our students need to have the flexibility—and not just the flexibility in terms of the application of the skills they have—but their flexibility of their cognitive abilities. They need to be able to learn how to learn about new topics, about how to apply those new ideas, and how to apply the skill sets that are gonna be important in that world to those new ideas. That’s something that we haven’t really assessed or worked on in higher ed before. You know, we do some of it, but we haven’t made that the forefront. And I think that’s something we really need to be able to do.
Stefan (41:39): And then, you know, the third thing is... this is from Shane Legg. You know, he had a great recent podcast. You can listen to it with Hannah Fry, who I believe its a physicist. And he talked about... earlier we had that reference to AGI and how fast the technology is advancing. And he’s actually, I believe, the first person or one of the first people to coin the term. But the second thing he brings up is that he said he was talking to some university professors and he was saying, every faculty in every department, they have to kind of really seriously consider: okay, what is this human level intelligence regardless of whether it’s jagged or exactly fits? What does it mean for education, for planning, for literature, for politics, for finance, for medicine? Right, and he said everyone needs to kind of take hold of their own fields and really start thinking through these things.
And I think that’s one thing... there’s kind of a lot of backlash to AI and some of it’s happening at the university level and it’s certainly understandable because it is going to change how things are done. But the one great thing about a university is there’s so many people who are so well educated, who are so intelligent, so engaging and want to learn things and teach students. And I think one positive way to think through this is it just opens up so many new opportunities for scholarship and engaging with colleagues and working with students to really think through what it’s going to mean for all these fields. And I really encourage people to do that.
Anand (43:13): This is another story that came out recently from Fei-Fei Li, who we referenced earlier in the podcast, known as the godmother of AI. She was talking in a recent podcast about how the value of your college diploma is really kind of losing some... is being reduced. The power of your college diploma is being lost. And she’s looking for some different skill sets. And you hear this from a number of different leaders that are saying that it’s not just about the traditional experience anymore. It’s really about how the student is able to learn, how they have some of these core skills. And we’re hearing this from the leaders. I think we’re going to also hear this from other business leaders coming forward too.
Stefan (43:52): Yeah, and you know, she’s not in any way saying there’s not value to higher education. You can learn a lot of the skills, potentially learn a lot of skills and the knowledge you need there. What they’re saying is that the degree is no longer like the automatic signifier. Like, “You have this degree, you know you can do X, Y or Z.” We want you to show us that you can do X, Y or Z. And if there’s somebody who can do X, Y and Z who doesn’t have a degree, and you can only do X and Y but you have a degree, we’re probably going to hire the person who can do X, Y and Z without the degree.
Anand (44:26): Now moving on to discussion of some content. And this is a really interesting quote that Stefan shared from Gabriel Langer talking about economics and different possibilities for AI. There’s so many new intellectual opportunities within a field like economics. And it’s not just about how AI is going to disrupt and lead to unemployment, but also longer-term questions about reorganizing society and some big sociological questions and political questions. We’ve really only scratched the surface. As we’ve even on this podcast been focusing more on what’s happening right now, I think we need to start to turn our attention, especially as scholars, as researchers, to consider what’s to come in the coming years.
And if we’re looking at something like post-AGI research and we’re considering what it will happen after you start to introduce very advanced AI tools—more advanced than what we have right now—how is that going to disrupt not just different sectors of society? What is it going to mean for the way that we need to reorganize those sectors of society? These are really important questions that very few people are grappling with. And I can understand why we’re probably shying away from it in some popular discussions, because it’s a little bit unnerving and it’s difficult to know how do you manage that? But we need to start that discussion soon.
Certainly within academia, I think this is the place where it has to be incorporated into our curriculum. We have to have these discussions with students because if we have a student that’s entering college right now, they’re going to be graduating into an economy that’s completely different than it was when they entered. And the only way we can really prepare them for that will be to develop some of the skills we’re talking about and to have the honest discussions about what that economy is likely to look like, how those challenges can be faced and navigated. And not that we’re to have all the answers, but we need to equip our students with the right questions and with an understanding about how to develop the answers for themselves.
Stefan (46:15): And this is just one other quote that just really saying we really need to think about what all this means for how we are going to kind of navigate this new world when the entire social contract is changing.
Anand (46:34): Now, one thing that we’ve talked a bit about, and I think we’ve even teased out some ideas along these lines within the podcast, but Stefan and I have certainly talked quite a bit about: What would you design as part of a curriculum today to be able to prepare somebody for an AI world? And here’s a sketch of something that we’ve discussed, maybe something like an AI and Practice Certificate where you have a series of courses. Maybe this is something for returning adult students to come back and to be able to retool for an AI world. Maybe this is something that students could pursue in place of a four-year degree, at least initially, and they could come back and complete these courses. They would count toward a four-year degree and maybe later they could complete that four-year degree.
Including classes like, obviously, AI literacy and working on their fluency with AI technologies, thinking about small group communication. We’ve discussed it pretty much at length: the importance of collaboration. And that’s something that you have to consider not just in a traditional sense with other people—that’s incredibly important—but also with AI agents and with AI tools so that you can start to collaborate and connect with a number of different types of cognitive processes.
Then thinking about entrepreneurship and then building for AI in the workplace and then getting to something like a capstone project so that students are coming out with an item in their portfolio where they can show that they were able to apply a lot of this understanding. And this is something that would probably be a little faster. It would be a little more accessible for a lot of students. And it directly answers those concerns about ROI questions: “What’s the return on investment of this degree?” And perhaps allowing for a certificate like this in higher ed not to take away from the four-year degree, but to supplement it, to be able to build this in within the four-year degree for students that want to pursue the full bachelor’s degree, but then also making this available for students maybe that graduated already with the four-year degree and they need to get retooled to be ready for an AI world. There are a lot of possibilities like this and we want to explore more of these in the spring with some other guest speakers that we have and interviews that we have planned coming up in the coming months.
Stefan (48:28) Yeah, I think this spring, you know, it’s just going to be really incredible. You just included a few here things at the end, just to kind of, you know, that number one: that AI awareness, right, that AI literacy. You know, a lot of the world... I mean, it’s just changing around people, you know, they’re not even aware. There’s a new study that just came out about how persuasive AIs can be. You know, a lot of people are spending their days in digital environments and being influenced in ways they don’t even understand. And that may be difficult to kind of control or account for. But we got to at least take a chance.
I saw an article the other day and said, “Oh, society’s facing...” I think it was from the product manager at Anthropic, right, Claude, one of the founders. And he’s like, “By 2027, we’re going to have to make a decision.” He said as a society, as to whether or not we want to let AIs build AIs. And you know, when I was listening to the Moonshots podcast yesterday, Alex Wissner-Gross said, Hey, I think the current models just through scaling are going to get us to the point where we can have AIs that then... okay, that’s probably not going to be enough. Scaling is not going to be enough. But the AIs that we can build through scaling are going to be enough to kind of create new algorithms that are going to build the AIs that allow us to build AIs.”
And you know, this Anthropic person is saying, “We have to decide this as a society.” And I was thinking, as a society, most people don’t even know what’s going on. They don’t even have a basic understanding of AI beyond a chatbot. How are they going to weigh in as to whether or not we as a society want to let AI build AIs? But I do think that’s important, right? I don’t know how much hate we’re going to have in that. Awareness... awareness is a lot. There’s a lot of stuff going on in the world that people don’t have a lot of influence on. But we see through the resistance, some things are kind of being slowed down like data centers, those types of things. You might say, “Well, I want to make sure my child is educated in an appropriate way. I want to make sure my child learns about this.” So it’s not all or nothing, but I think we really need to start building some more awareness in AI literacy.
Anand (50:34): Now we’ve talked a little bit about this platform and some of the tools that have been developed. Poplelli, that’s part of the Google Labs suite. And you can try this out if you have Google Gemini access. Really interesting to think about for persuasive purposes, for marketing. And there are more tools like this that not only should our students know about the tools and know how to use them, they need to know that these tools are available for others to use for persuasive effect. And part of that is just media literacy. We need to update our understanding of media and digital literacy to incorporate true understanding of what these AI tools are capable of.
Stefan (51:06): And you know, entrepreneurship, right? As we talked about, we talk about the Moonshots podcast. A lot of the people on it are highly qualified, but you know, Peter Diamandis is saying the career of the future is entrepreneurship. We don’t know how many jobs there are going to be in the future in terms of what you can do for somebody else. They’re probably going to get that done very cheaply in many instances, if not most instances, with an AI. So you’re going to have to learn to be a creator. We just saw some of the tools that you’ll be able to use, but you’re going to have to understand some kind of basic ideas about how to launch a business and launch a business in this new environment, which is a great opportunity for business schools.
Anand (51:45): The other aspect of this that I would probably want to add to an understanding of it is not just that everybody needs to be an entrepreneur and start their own company, but I think all of our students do need to develop an entrepreneurial mindset where they’re able to look at market needs, they’re able to look at societal problems. And so even if they’re working for an employer, that 40-year career is just gone. It’s not going to happen any longer. It’s very, very rare that someone will have one job that they will do consistently the same way for a long period of time. Society is changing too rapidly and there are too many changes within the workplace for that to be the case. So for somebody to be flexible, learning to learn is important and also having the mindset where even in your job, you’re able to then identify new roles that you can take, new actions that you could take, different production processes, workflows, new needs within the workplace that you can then identify a way for that to be changed. That’s one way that we’ve talked to students quite a bit about how do you navigate some of this change?
You know, we’ve talked before about the World Economic Forum report that discusses the loss of so many jobs and then maybe 90 million jobs, 170 million jobs being created. And it’s a little scary to think that you might be fired from one of those jobs that’s disrupted and then you have to find a new one. There’s also going to be space for people that are in some of those more traditional jobs that while they’re in that job to help navigate and change the role of that job for many of them, not all of them, but for many of them. And that’s an entrepreneurial mindset that allows that to happen. They are best positioned to be able to say, “This job needs to change in this way. This is what our company needs to do to be able to meet these new challenges, these new needs, to be able to meet the demands of our customers.” And that entrepreneurial mindset is going to be really important coupled with all of these other skill sets that we’ve been talking about.
Stefan (53:32): You know, so I figured, well, we should end with some predictions. I thought I’d share mine and, I have a lot, but these are kind of some of the highlights. You know, I think we’ll see a move to certify AI courses and maybe even institutions. There was a little bit more that came out with this a few months ago with it. You know, there’s certification and accreditation bodies and, you know, the Trump administration is not taking a super favorable look at most universities, partially because they want more of these kind of skills that people are going to need for the future to be taught and developed in universities. But also, as we’ve seen, they’re very much behind AI. So I could see them kind of accrediting certain AI universities or courses, and those have become important because then students can use financial aid and access student loans as part of those accredited institutions. I think that’s definitely going to... I don’t know definitely, I think there’s a good chance that’ll happen. Even if it’s not the certification of the institution, which I think is possible, there definitely be some kind of course types and kind of giving credence to those types of things.
I think second, there’s a possibility—and more and more people are talking about this—of an event that really raises awareness of the power of AI. I don’t know what it could be. Maybe it’s a bunch of Waymos causing some huge traffic disaster, could just be a massive cyber attack. I think that’s one thing that people are talking about. So more and more people just say that something could really be triggered kind of by AI that will kind of really kind of shine the light on the importance of this.
I think the third thing, and I’ll take this from Demis Hassabis, who I think is a pretty trusted source. He’s the leader at the top of the field there is the CEO of Google DeepMind, but doesn’t have a reputation for hyping things. Elon Musk does a little bit and some others, and they’re usually right over the long term, but they say stuff much faster. He says that we’re gonna really see... now we’re seeing agents that when you work, they work well with a person, but you need kind of a significant amount of human interaction. He says by the end of the year, those are gonna be able to act kind of largely autonomously on their own and still do human-level work. We’ve kind of seen it a bit at the micro-level with 5.2. So I kind of agree with that. I’ll take it there.
The third thing I talked about earlier is, you know, we’re probably into 2026 into 2027. It makes sense. It’s consistent with what other people are saying about a rapid takeoff in 2027. And I’ll finally, I’ll say, well, I guess two finalists. One is I think we’re going to see a greater shift in focus in K-12 curriculum, or at least I hope we are. Otherwise, I think a little bit of the relevancy of some of K-12. And I would say more 9 to 12. I don’t really know that we need to change K to 8 that much. But 9 to 12, especially 11 and 12, to prepare students. I think more like you say, fewer students are going to be going to college. Those that do may not stay as long. They may see opportunities that develop, may see a greater emphasis on kind of physical labor rather than cognitive labor. And K-12, you know, people keep saying, “Well, what’s going to happen to education?” I think if anything, K-12 is going to become stronger. But the responsibility of them to prepare students, not just for college, which kind of have you, there’s kind of the set way of doing this, right? Like, get all the kids these core classes, make sure they do well in their AP exams and their IB exams. They do well on the SAT and that’s... have stacked their portfolio with some extracurriculars and volunteer services. Kind of this kind of formula, right? For getting kids into the higher educational institutions that they aspire to. But I don’t think that’s gonna be... that’s still gonna be there, but it’s not gonna be as significant as it used to be.
And then the fourth thing, I didn’t put it on the left, but you know, a LinkedIn post that I shared with you that we’re going to really kind of see the first AI native university. What does that mean? It means essentially that the content of a lot of things is going to be taught with AI and the professor will kind of be there to kind of build the social experience, build the intellectual environment to help the students like learn with the AI. I shared an LA Times article by a professor at the USC Business School a couple of weeks ago with some friends that just saying like, this really is the future of the university, that students are going to basically learn the content from the learning management systems. And I’ve heard other people say this as well, and that we really kind of need this environment to develop the people. That’ll become the future of the university. And I don’t think a lot of people are prepared for that, but I think we’re going to see the start of that next year.
Anand (58:33): You know, one thing I would add to your point about the shift in focus for K-12 is that I think there is a real need for us to incorporate in K-8 some understanding of AI as it is permeating society. There’s so much, there’s such ubiquitous deployment of AI in different areas that while I would agree, we really don’t want and don’t need grade school age kids, even most middle school kids to be using a lot of AI or any AI, they are going to be encountering with it. And there’s a lot of it that’ll be used at home. We need to have some of those discussions in the classroom.
Now move on to some of my predictions and I added in a few that I think I agree with pretty much everything Stefan was outlining and a few things that we’ve already kind of said but I want to build on.
I think that this AI backlash is going to continue to build and part of it is going to be aligned with the midterm elections in the US. There are going to be some populist appeals, some concerns about economic concerns, certainly job disruption, and a lot of that is going to be tied to AI. And I think there’ll be a lot of, as I said earlier in the podcast, there’s so much of that is really inextricably intertwined with different areas of society that even if it’s not the cause of those particular items, I think that it is something that will be tied in and it’ll be very easy for people to kind of see that as a connection, something that they would certainly tie to.
That significant event, I think Stefan’s right and we’re gonna see something that raises awareness of AI and people have been predicting this for a while. Hopefully, it’s something that’s not too disruptive or dangerous or endangering any lives, but we do need to be prepared for that and it’s the response and the discussion afterwards that will be really important.
I also think the capabilities are going to continue to outgrow what users are experiencing and most won’t notice the capability growth. We’ve discussed before how free users of AI platforms rarely get to see what the full capabilities are. Now, ChatGPT-5 changed that; they’ve incorporated that into the free version. But even there if you use the free version you’ll see that it really feels like it’s throttled with their limited interactions. You’re not really getting the pro level that somebody is paying $20 or $200 a month is able to be able to see. And as we talked about how these capabilities are changing so rapidly, it’ll be difficult for the rest of users, let alone people that aren’t using it, to have an understanding about what that really means.
I think there’s going to be a real scramble in K-12 to prepare students for an AI world. It’s difficult to make curricular changes. And I’m really sympathetic to the administrators and faculty in K-12 settings, as well as in higher ed, for working with traditional processes where it takes six months or a year to be able to add a new course or to be able to get it approved to be able to incorporate and then working on professional development and be able to find staffing for it. We’re using a traditional timeline that would often take years to incorporate new developments and to be able to incorporate literacy programs, for instance, in our schools, and we really need to be able to do that in weeks and months to be able to address some of these changes. And so I think K-12 is going to be scrambling and it’s partly going to be driven by calls from parents to be able to get that to happen. And I’m afraid that’s probably going to lead to more of a gap between the schools that have those capabilities, that have that support, and those that simply won’t.
And then higher ed. I think that higher ed is going to face a number of cascading crises that will be not caused by AI, but certainly are going to be exacerbated by AI. We’re hitting the demographic cliff, you know, obviously a drop in the number of births in 2008, 2009, because of the Great Recession. We’re seeing cutbacks from governmental support, certainly the Trump administration making changes to the way that funding for research has occurred or it is not occurring now. A change in the number of international students that are coming in, questions about return on investment. I think all of these are going to have a coalescing effect. And I’m really concerned about a number of institutions of higher ed that are just teetering financially. They’re going to have to find a way to adapt. And I think honestly, many of them are probably going to be closing in the next couple of years because it’ll be difficult to make those kinds of changes and to be able to adapt that quickly. And we’re probably going to see some real changes to curriculum and to the way that we’re administering programs and developing new programs as we go forward. There’s going to be a lot to explore and to talk about in Spring ‘26 and I look forward to having those discussions with you, Stefan, and with the rest of the audience.
Stefan (1:02:52): Yeah, the final thing is I just add is, you know, there’s kind of tensions in a way with some of the predictions, right? I think they’re both true. Right. You mentioned, OK, you know, in K-12, there’s going to be parents calling the head of school and showing up at school board meetings demanding more. Right. They want their children to have more education. You have other parents contacting those same people and saying, “We don’t want to. We want our student to learn in an environment without AI. We think AI is going to destroy the environment.” Maybe they’re “doomers,” right? And, you know, a lot of people in education are going to get kind of caught in the middle. It almost reminds me of the pandemic, right, where they were like, some of the parents that wanted their kids to go back to school full time and the other people wanted the schools closed. So, you know, they had these like hybrid systems where they’d be like, monitor on the wall and your kid could listen in on the class if the students are live and it just kind of... it was kind of this hybrid, you know, it’s kind of a mess and we don’t know that the hybrid is the answer. But I definitely feel it for school administrators whether they’re K-12 or university administrators, right, they’re just going to get pulled in a lot of different directions and when you’re swimming in a world of massive change and disruption just like COVID.
That’s kind of inevitable. I guess it’s going to go with the job description, but I think people are underestimating kind of how much kind of turbulence is ahead.
Anand (1:04:20): I agree. We have a lot to look forward to. There are going to be some challenging times ahead. But for now, we want to sign off for the end of the fall semester and say, see you in spring 2026. We have a number of guests planned and interviews ready. And we’ll continue with regular updates as well. Stefan, thanks for everything. I’ve really enjoyed working on this with you. And I look forward to what’s coming in the next semester.
Stefan (1:04:43):Yes, thank you so much and I’ll see you at the beginning of the year.