In this episode, Michael and Ben dive deep into the intersection of education and technology with their insightful guest, Daniel Hiterer.

Michael, a data engineering and machine learning expert, and Ben, an integrator of Gen AI tools, navigate through Danny's unique perspective on the impact of nurturing educational environments. Currently working at Cornell’s Studio entrepreneurship program, Danny brings a multidisciplinary background, combining history and instructional technology, and shares his vision for the future of learning.

This episode explores the transformative power of nurture in education, the evolving role of Gen AI in fostering curiosity, and the challenges and opportunities in integrating AI into the learning process. Danny provides thought-provoking insights on emotional access points, curiosity-driven learning, and the delicate balance between educational goals and productivity tools.

Listen in as they discuss personalized education, the promise of AI-assisted learning, and the future trajectory of superintelligence in education. Plus, hear personal anecdotes from Ben and Michael about their own learning journeys and the evolving landscape of curiosity and knowledge.

Socials

Michael Burke [00:00:05]:
Welcome back to another episode of Adventures in Machine Learning. I'm one of your hosts, Michael Burke, and I do data engineering and machine learning at Databricks. I'm joined by my amazing, beautiful, and wonderful co host.

Ben Wilson [00:00:16]:
Ben Wilson. I do integrations with Gen AI Tools at Databricks.

Michael Burke [00:00:22]:
Today we are speaking with Danny, who's actually a good friend of mine. We frequent live music together. He studied history at Queen Mary University of London, then Instructional Tech and Media at Columbia. But currently he works at Studio, a Cornell run entrepreneurship program that provides coursework and collaboration opportunities. Opportunities, but he might be

Daniel Hiterer [00:00:45]:
moving to a PhD program at

Michael Burke [00:00:45]:
some point in the near future. Maybe, maybe not. I don't know. So, Danny, starting it off, why do you care about education?

Daniel Hiterer [00:00:57]:
Hi, Michael. Hi, Ben. Good to be here. Why do I care about education? I think that people can be described no. Let me let me phrase this a different way. The best version of us is the sum of our positive influences or maybe even more. But either way, every person at their best are a product of their best influences. We don't just, you know, grow into who we are on our own.

Daniel Hiterer [00:01:33]:
We grow into who we are because of the help we get from others. Be it our parents when we're infants, older siblings, maybe teachers, and that's, you know, where we come into education. I just love to see people grow, and I think it's important. And I think we should keep growing and never, you know, get fixed into a zone of comfort intellectually. We should always seek out new information. And the people who can help us do that, I think they're instrumental to keep society rolling. So I just figured I would like to devote my time supporting people who are great in those ways.

Michael Burke [00:02:15]:
As an education professional, how do you think about nature versus nurture?

Daniel Hiterer [00:02:22]:
I go heavy on nurture. I think nature gives us the temperament and a few base qualities, but nurture is huge. I think I guess it's because, well, we could go into philosophy. I believe in free will. Like, there was a writer and a poet, Isaac Bashevisinger, Nobel Literature Prize Laureate, who said that we have to believe in free will. We have no choice. So just to, you know, just to be an optimist, I believe in free will, and therefore, I think that we can change, our predicaments and, change other people's predicaments. And so I think nurture and from what I've seen, good nurture can do a lot, and it basically shapes us.

Daniel Hiterer [00:03:08]:
And nature we work with the nature, but nurture can do a lot.

Michael Burke [00:03:15]:
Ben, what are your thoughts?

Ben Wilson [00:03:17]:
If that

Daniel Hiterer [00:03:17]:
if that's happening, what do you guys think?

Ben Wilson [00:03:19]:
I mean, I a 100% agree with everything you said. I think all 3 of us are peas in a pod on that topic, based on the years that I've known Michael so far. He's definitely in that camp, continuous learning, improving, through getting exposure to the right sort of information, the right sort of people, and yeah. And I'm the same way. I seek that out. And it's it's great to hear more people who are in that camp. But the big topic that that we were talking about before we started recording was was particularly Gen AI and its influence in learning and the the potentials that it possibly has. The first thing that popped on mind with talking about nurturing for me is at different developmental stages of a learning journey that somebody would have over their lifetime.

Ben Wilson [00:04:16]:
Do you see, if there's any what the what is the split between positive and negative influence that providing a very attentive, always on sort of nurturing force would have? So if we were to compare something like, okay, we have an AI that can help teach and mold somebody into that best version of themselves by stimulating curiosity and igniting that innate human ability that we have to explore our existence. Do you think that there's a difference between how that would impact somebody who's 4 years old versus somebody who's 34 years old? And do you think about that in, in your, your research?

Daniel Hiterer [00:05:09]:
Yeah. It's an interesting topic. I spent some time thinking about curiosity a few years ago. I was really focused on it. I think it comes down to a statement that, you know, people might disagree with. We don't know if it's true, but I feel that it's one of the truths that I live by is that people never lack motivation or people are never not motivated. They're always motivated, but the question is they're motivated to do what? Maybe someone's motivated to, you know, to to to complete a video game in a day, or maybe someone's motivated to read a book in a day. We assign values to all of these activities as a society, positive, negative, but I think there's always I don't wanna sound like I have, you know, rose colored glasses on at all times, but I think I do a little bit.

Daniel Hiterer [00:06:14]:
But I think it helps. I think that there's always a reference point that's familiar to a person. We all have reference points or topics that we're interested in that just connect with us. And the interesting puzzle for me in education is how can I find the right reference point to reach the person? Be it a 4 year old who, for some reason, like likes boats and pays attention to boats and you know? Or or toddlers, it's very difficult to get many signals from them just because they cannot communicate as well. So you have to observe them really closely to see what interests this person.

Ben Wilson [00:07:03]:
Mhmm.

Daniel Hiterer [00:07:03]:
And then, you know, teenagers who might be not on their best behavior and rowdy, I always wonder what's going on there and what motivates them and where they going. And then looking how to integrate my goals with their goals and see if we can work together so that both of us achieve the outcomes that we can. And, adults as well, same thing. Just looking at when they're most interested, what they're most interested about, and seeing how we can just use these reference points or points of entry to push the conversation forward. Yeah. Including including how they communicate with AI. I don't know if if conversation agents, the large language model conversation agents currently have the capacity to understand reference points, but the fact that they can bring up examples for a situation that might be relevant for for a person who's seeking knowledge, I think that can help. Like, I some of the students that I work with are high school students, and, you know, coming up with examples that are relevant is the key thing.

Daniel Hiterer [00:08:24]:
And definitely, you know, large language models capability to to explain how a business works in terms of, you know, sneaker resale. That's big, and that's usable. Like, right now, like, tomorrow in a classroom. So, yeah, it's it's about entry points. I hope it answers the question. Let me know if it doesn't. No. I'm sorry.

Daniel Hiterer [00:08:46]:
I went on on a ramble there.

Ben Wilson [00:08:48]:
No. It was fantastic answer. And it it aligns to my own personal experience with, you know, using Gen AI daily. Like, yeah, I use ChatGPT quite often for tactical use cases, but I also have a different setup that I've done with that interface and some I also evaluate a bunch of these other providers to test their capabilities for work, of course. But there's also, I've always had a predilection from a young age. I remember how exciting it was to go into my town's public library and go like, I'd go there with a couple of friends and, and they go off to the comic book area. And I was always the nerd who would go over to the reference section. And I'd had something that I was curious about maybe on the walk over there or something I was thinking about the last week.

Ben Wilson [00:09:42]:
I'm just like, I know nothing about this, but, I wanna learn just at least a cursory amount of information about this. So I go and, like, try to find a book or find you know, go get the encyclopedia and go, like, read about this and and just feel satisfied that I'm like, hey, I now understand this a little bit better. I don't know really what is going on here, but at least I I know a little bit about this. And I think that general knowledge is is super useful so that you just not so much in the, oh, I now know trivia. It's more like, I now know what this is. So if I need to learn more about it later on, I have that base that I just built by going to get that reference material. And then fast forward, you know, from that point 20 years, to when, you know, I'm an adult and this website comes out that is so ubiquitous these days, and it's even used for the evaluation of Gen AI right now and used as a training set for Gen AI, Wikipedia, you know, everybody has interfaced with it. I used to spend free time just doing Wikipedia crawls where I'd pick a topic and I'd open every link on the page.

Ben Wilson [00:11:01]:
And then I wasn't done until I'd re I'd read one tree that's down so that I kind of understand that topic. And just, it was really fun for me to just kind of do that and be like, oh, now I know this, you know, something about this. Is this the academic truth of this topic? Probably not. It's not as depth as deep as you possibly can go. But at least I have this, this level of understanding that ignites curiosity in my own brain. And I've applied that throughout my career and throughout my life to teach myself new things. I've understood that pattern of this is how I learn or how I like to learn, and I like that reference material. With with the advent of advanced chat models, particularly when you're talking about, you know, GPT 4 o and its capabilities, I'll maintain, like, a different, you know, account with that.

Ben Wilson [00:11:59]:
That's just for my curiosity. And it's it's shortening the time that I would otherwise, which I don't have anymore, to do a Wikipedia crawl. I can get those that distillation in seconds and read through it in 3 or 4 minutes, and then start asking leading questions about additional topics. And I can scratch that itch, pretty easily. And I think that is for, like, for me, the application of this technology provided that, you know, there's appropriate guardrails for age appropriateness and topic appropriateness is incredibly powerful of, like, creating that, that satisfaction that that I think we all have about, you know, getting through that curiosity phase and feeling satisfied of of learning something new.

Daniel Hiterer [00:12:54]:
Yeah. That's wonderful. I mean, if you are a curious person and you have the resources you need to pursue new information, that's that's the best. Right? Since you were always curious and you had the library to go to and the attention span to browse so many books and still be interested, that's great. And the fact that right now it's easier and easier to get information you need, it's just fantastic. I think even I don't know how you both I'm actually curious to hear how you both deal with this, but when thinking about curiosity, I also think about not the impulse, but also a more consistent practice that you build being curious and exhibiting curious behaviors. No. Sorry.

Daniel Hiterer [00:13:46]:
That was not exhibit just acting on your curiosity. You know? Because, for example, one vivid example that I've had in my head for a few years from a conversation with someone is, when you're walking around a city and you see street names, you sometimes have the question in your head for a split second, like, oh, I wonder who this person is. But the question is whether we pull out our phone and we look it up. And we can't we can't be curious because the question pops up, but then my question is, how do we remember to act on the curiosity and actually do the thing and and seek new information? And then the another question is how much of that information stays in our head? You know? Because we might, you know, look something up, and then in 5 minutes, we don't remember it anymore. So it's it's interesting to think about distribution of attention across new info and how much of it we can or should intake because there's so much information. And in theory, everything is so interesting. But in practice, how much how much can you read or watch or listen to a day?

Michael Burke [00:15:03]:
I think I can give a concise answer. Yeah.

Daniel Hiterer [00:15:07]:
Yeah.

Michael Burke [00:15:10]:
I like the feeling of not knowing stuff, and I'm very directed with where I spend my energy. So for street names, I like, I've literally yesterday, I said I forget what it was, but, there was like, oh, I don't know what that is. We could Google it, or we could just never know. And I like that we could just never know line. It's actually kind of fun because I don't know. There there's a lot of beauty to not knowing stuff and having it just be magical. Like, for instance, when you're a chef, food tastes very different when you're not a chef. When you're when you play instruments, music sounds very different when you don't play instruments.

Michael Burke [00:15:47]:
When you're a ref in basketball, the games look very different. And having stuff just magically appear outside of your area of expertise, I think, is a really beautiful thing, and it's really fun to experience. And then I can save all of my energy for going really, really hard in a few areas. So that's my take.

Ben Wilson [00:16:08]:
I like that take. I'm similar, but a little bit more broad. So I'm that nerd that will pull out the phone and look up the name. It's not so that I have hopes that I'm gonna remember that 20 years from now. I probably won't, if it's just, you know, general references that have no impact to the mental model that I have of things that I'm curious about. But I'll still do the lookup in the hopes that there'll be some reference point to something in my own past. For instance, like, yeah, like walking through a town around here or something like where I live. Like, man, I've seen that that name before, and I don't remember what it was or it seems familiar.

Ben Wilson [00:17:04]:
Look it up and be like, oh, that's why the the street that runs all the way through this entire county and actually goes to, you know, it's over 400 miles long. It goes all the way to this port city in the state. And I'm living in a state right now that I didn't grow up in, but I'm fascinated by history. And, you know, a road of that of that stature, like, oh, it's named after the original governor of the colony of North Carolina. And I wonder, like, I wonder I wanna know more about this person and, like, what their time was like. Like, were they a good person? Were they a bad person? Like, what is the the actual, you know, root history of that? I didn't have that curiosity about history until I read a single book in my freshman year of college, which was Howard Zinn's People's History of the United States, which blew my mind because it was so antithetical to everything that it, that I had learned in school up until that point. It's like, oh, these are essays and letters and references of people who actually live during these times. That is not the society narrative.

Ben Wilson [00:18:16]:
It's not the one that governments like to promote in education systems. It's the actual real story. And a lot of it's super dark, and it's full of a bunch of events that are I could understand why the government wouldn't wanna teach young children that it used to do things like this 200, 300 years ago. So now anytime I have something that's like a historical reference to something that there's probably a story here that isn't true, that has been crafted to to make make this be smoothed over. And then you go look it up and you're like, yeah, I was right. And I'll remember that for decades. And every time I go on that road, I'll think of that. And then start thinking about, oh, there's another road that's 20 miles down the road that, like, down the road that has a very British sounding name, or it sounds like it's somebody from Scotland.

Ben Wilson [00:19:13]:
Why is this name? And so it insights that curiosity. So I always do stuff like that just in the hopes of making a connection. But if anybody were were to look through my search history on any browser, they'd be like, what is wrong with this person? Like, why why do they search all this random junk? Because I'm constantly doing stuff like that. And with Gen AI, it's it's even worse. Like, in my account with chat gpt, I'm sorry, OpenAI, of all the ridiculous questions that I ask, because it it's just filled with random nonsense.

Michael Burke [00:19:47]:
I just bought the book. You were gonna say, Danielle?

Daniel Hiterer [00:19:52]:
No. But that's good. I feel like as long as it satiates the hunger that we have, either it's, you know, just hunger sometimes for new things or hunger to discover something every day, it's great that we have it.

Ben Wilson [00:20:08]:
But from your perspective, how do you ignite that in, like, through the education system? What's the what's your theory on the best way to get as many people activated in that way so that they are curious about not just about themselves and things that are very insular to just their own existence, but more to an industry that they may work in or the society, the neighborhood that they live in, the society that they live in, their fellow humans or their environment, you know, do you see that GenAI application in education would would provide a means to sort of spark that fire.

Michael Burke [00:20:59]:
And just sorry. Adding one layer, is that the fire that you're trying to spark? Because I feel like I did I wasn't curious about the SATs, but I did study. And so that's maybe a different angle, and that's not curiosity driven. That still gets results. So what is the fire that you're trying to spark?

Daniel Hiterer [00:21:20]:
It's interesting. The sparking the fire metaphor is is helpful as well. I think, first of all, we should say that it depends on the age and on the predicament of the learner because we have the traditional schooling system that is just, you know, the conveyor belt that we all go through when we're kids. We do 12 years of this and then most probably 4 years of that. And as we go through that kind of process, I never I don't know about you guys, but I think most people don't question why they're in school. They're just like, yep. We just gotta do this. We're growing up.

Daniel Hiterer [00:21:57]:
We need to learn all these things, and then, you know, then we'll be adults and we'll see. So in that case, I don't even know what kind of fire the students themselves are looking for to spark. I don't know how, you know, how often and maybe it depends on, where the students are in the world and what kind of school they're going to, but I don't know how much of school is just inertia for people, that this is just something that needs to happen. And I know that there's a fair amount of people who feel the same inertia in college as well. They're in college be just because that's what you do when after school. You know? A lot of people hold that belief. So in that case, I feel like it's more difficult to spark the fire, maybe because the learners don't have that. We call that metacognition is when students think about the way they think.

Ben Wilson [00:23:00]:
Mhmm. So

Daniel Hiterer [00:23:01]:
and and and think about their learning strategies and how they acquire knowledge. I don't know in these formal education systems whether you know, to what extent the students well, not all of them, but some, maybe a large portion of them, are are looking for that fire themselves or they're they're just to be there. But in other context, if we're thinking about, for example, museums where people go on their own accord, they're not required. And also adult learners, once people get to master's doctorate level education, professional education, executive education, we feel like adults can be responsible for their own decision, and they go into a learning experience already knowing kind of what they want to get out of it and what they want to learn and what they want to get better at. So I would say that it's easier. The extracurriculars and the continuing education is easier. Traditional education is more difficult. I've mostly been working with continuing education and extracurricular.

Daniel Hiterer [00:24:06]:
Most of the work that I've done, was either with graduate students, so MBA students, executive MBA students, right now at Cornell Tech, engineering masters and law masters, or with high school age, kids who are in experiences that are extra to their school. So it's summer fellowships, summer camps, summer school programs. And in all these cases, I think I've had an easier job sparking the fire than people in, teachers in traditional education systems and professors at universities just because the students were in the classroom knowing that they want to be there and choosing to spend their time in that classroom as opposed to somewhere else. And still yes. The problem is that some people might come to schools just to get a piece of paper to get a job, even for a master's, just to show a potential employer that they've made a pivot to a different field. So then it's more difficult to establish that ground, like, hey. What are we doing in this classroom? And, you know, for what reasons are you here? And why do why do you wanna be present? But then, again, we just come back to, you know, authentic context and and what the student cares about. That's why I really enjoy entrepreneurship as the discipline in which I do my work, entrepreneurship education.

Daniel Hiterer [00:25:30]:
It's because people who really want to do it, they craft their own journeys, and they select the topic of interest where to launch a business or where to do a case study or or what to analyze. They're almost naturally drawn to, to what they know. Not to what they know. Well, partly to what they know, but to what they're interested in. And then the question is how to not make it boring for them, how to keep them interested as opposed to make them because they're they are interested. But how do you keep them interested and don't overload with theory that doesn't connect to their context and maybe help them uncover a certain layer just as, you know, you see a familiar name as a street sign, and then you uncover that they have, like, a connection with this other person from history and so on and so forth. I feel like people like making connections like that. I really like explaining the when I see high school kids, who come to class in in all the sneakers, the the Nike SB Dunks and whatever.

Daniel Hiterer [00:26:30]:
I really like to bring up the, you know, supply chain of how sneakers are made. Just when we talk about supply chains, I feel like that all that always works for the students, and they're always excited just to know how much their Jordans really cost to make. So so, yeah, I feel like it's I I hope I'm not repeating myself too much, but I just look for access points and meeting students where they are and see where I could direct their attention next, and takes a lot of iteration. But at the end of the day, if if it's, if it's successful, it gives just outsized returns on the students' engagement and overall vibe in the classroom and and their achievement at the end of the day.

Ben Wilson [00:27:16]:
Do you see a system within the next, like, with the rate of advancement that we're seeing in GenAI and applications in particularly with very advanced systems like super advanced agent applications, Do you see something within the next 5 years that could be broadly accepted that would be an assistant that actually is personalized to target this? Something that's like, hey, you're in this course of study, you're doing this thing that's traditional structured, but taking and I'm sure I know I've had a couple of these classes. I'm sure both of you have as well. You're like, alright. This is a requirement for my graduation from this university program, but it's not really tied to my degree. It it's kind of adjacent to it. I remember feeling that about technical writing. When I took that, it was just core requirement at the college I went to. And I was like, yeah.

Ben Wilson [00:28:21]:
Not super excited about the concept of learning how to write the best white paper until years years later. And I was like, oh, I'm glad I took that class to understand, like, how these things are constructed and what you should focus on. But during the class, I just felt like it was a slog because I couldn't connect the dots. And the professor really was kind of checked out and didn't really provide that to to the instruct to the students. Do you think that focusing on applications or do you think that's gonna be something that that is widely available where it's like, hey, there's this thing that you give it a a syllabus, and then it already understands through contextual storage of your history of conversations that you've interacted with this tool. It knows how to present that information for this person. Like, hey. Yeah.

Ben Wilson [00:29:21]:
Here's the topic, but here's the things about this massive amount of information that is going to engage you the most or present it, craft it in such a way. Because that's not scalable from a professorship perspective. You know, any teacher anywhere with a a room full of students is not gonna be able to to craft, you know, a 120 different lesson plans.

Daniel Hiterer [00:29:46]:
Yeah. Widely available within the next 5 years, I would say yes to an extent. I think it can be by widely available right now even to a limited extent because we've already seen that, you know, with with, for example, chat GPTs or or OpenAI's custom GPTs, where you can build a persona that always has the same, like, additional context in mind, and you can chat with them on a given topic. And they will always reference not only what's in their training set, but also something that you upload, like a syllabus, for example, you know, workbooks, additional readings. It's doable. And I think with providing functionality, like providing you extra examples or explaining some concepts, probably not math, but maybe, you know, social sciences, distilling economics theories, or history narratives. It is possible that it performs, you know, something. It gives output that is something, and it's widely available.

Daniel Hiterer [00:31:01]:
Yes. My few questions here are on the quality of the output is number 1. And number 2, I guess, the the rate of adoption and how widely it's it's actually used. I guess, personally, starting with the nature of the of the output of LLMs. This summer, we tried to we did an experiment where we created AI study buddies or AI teaching assistants for students in an entrepreneurship class. Those were high school students. And, we basically gave them custom GPTs for every topic of the course to submit their work to. They can submit their work to the GPT and then get feedback according to the, worksheets, syllabus, rubrics, and all the material we have for the class before they submit it to the professor.

Daniel Hiterer [00:31:59]:
The class didn't have any grades because it was, like, an extracurricular, but still we did that. And to be honest, the output wasn't too good. Mhmm. Maybe there are things to do with the prong, but if we go kind of more to the technical side a little bit, what we think happened is that at least the GPT technology is having a hard time reading PDFs. Mhmm. So sometimes it reads them, sometimes it doesn't. So if a student uploads, like, a business slide deck, sometimes it just cannot access the PDF, and it gives an answer as though it did. Because as we know, the most difficult thing for Chad GPT or or another LLM chatbot to write is I don't know.

Daniel Hiterer [00:32:44]:
It just wants to please us more than give us the correct answer. So we've just seen very weird things happen. The LLMs giving feedback without having access work hallucinating content that wasn't in the student's presentation. On the other hand, I must say that a curious insight that we had is that the feedback would always be positive. Even the student did very poorly. The GPT would say, you did well. Maybe not great. It was like, good job on this and that.

Daniel Hiterer [00:33:19]:
And we had student students kind of write in their reflections. We collect their reflections just to see how they feel about, technology as well and just to get a glimpse into their experiences. 1 student said, oh, I I thought I did really poorly on this assignment, but then I put it into the GPT, and it told me that I did a good job. So, yay. And then and and it's a legit question. You know, some would say, hey. This is false confidence, and they think they know everything. Well, they don't know it.

Daniel Hiterer [00:33:48]:
And you know? But at the same time, if someone is telling you that you did a good job, maybe that helps you be more involved in a process. You know, this can go into the participation trophy discussion, of course.

Ben Wilson [00:34:00]:
Right.

Daniel Hiterer [00:34:02]:
But but it's interesting. So the output we found that it's not really reliable at this point in a student facing way. Maybe it will be in 5 years. We don't know the rate of change of these, AI models. There was this, one person, I don't remember who it was, but he had a comment that, about the rate of growth of of LLMs. He said, my toddler is now 10 months old. If he grows at the same rate that he has been growing, you know, for the past 8 months, then in 10 years, he's gonna weigh a 1000000 tons, you know, or something like that. So it's it's, not necessarily linear.

Daniel Hiterer [00:34:42]:
But, again, this is a speculation. I I don't feel like that's my topic of expertise, but right now what we see is that output is not always reliable, and I don't know if it will be reliable. Maybe we're asking too much of it. Maybe we should just ask it to give examples and not give students feedback on on its way. Maybe that's we maybe we should limit ourselves to a more narrow use case to make it work. But but, yeah, the output is questionable. We have to work with that. And the other thing that we have to work with is the how much people actually use it, how much, professors and teachers employ it.

Daniel Hiterer [00:35:19]:
Because even if a technology is available, what I have seen is that it does take time and attention experimenting with it so that it's viable to integrate into the classroom. And, time and attention in a teacher's life are scarce resources, you know, especially if someone is overworked or, you know, maybe they're just someone who doesn't experiment with technology too much. Maybe maybe they're not looking into ways to improve. Maybe they're happy with how they teach and how they've taught for the past, you know, x years, and then they just don't use it. So the technology can be there for the taking. But if the teacher does not know how to adopt it or is not willing to adopt it, then, you know, it nothing happens. So just sorry. Again, I ramble on.

Daniel Hiterer [00:36:11]:
But I guess the short answer would be in 5 years, AI tools might be good enough for education focused use cases or some education focused use cases, not others. And they will likely be, implemented by a range of early adopters

Ben Wilson [00:36:34]:
Mhmm.

Daniel Hiterer [00:36:35]:
Among teachers. Among students, I'm actually more optimistic. I think more students will be using AI tools than teachers will just to help themselves. But that brings another question of, you know, how how good of teachers LLM, conversation agents are. I don't think I haven't seen a reliable benchmark that would evaluate a language model's performance in education, how good of a tutor it is. Because we have benchmarks that evaluate their kind of cognitive ability, so to speak. Mhmm. But that's a whole other rabbit hole that I guess I guess I'll just pause here.

Daniel Hiterer [00:37:15]:
Sorry. I've been talking way too much.

Ben Wilson [00:37:18]:
So I have a a purely theoretical question for you that you mentioned this, like, rate of advancement of everything that that's been happening. And it's truly staggering when you think about it, the difference between the true foundation of OpenAI's tech or the original GPT 2 model, basically, put up for free on transformers hugging face hub. You can download it now and interface with it and see how it performs. Spoiler alert, not so great. It it is not particularly sophisticated as compared to the latest state of 4, you know, 401 that that's in beta right now. So seeing that advancement and this this nonlinear trajectory of, you know, it's on logarithmic course now could be exponential in the next couple of years. Do you first part of the question is, do you feel that it's inevitable that we're gonna hit not this generalized, you know, highly capable intelligence, in AI that can do this one thing really well. Right? Right now, it can it can understand language really, really well.

Ben Wilson [00:38:39]:
At its core, it's what it does, and it can generate text that is related to whatever it's been trained on. But eventually, we're gonna move more into increasing capability of its core capability that it has now. So improved understanding of language and all of these other functionalities that are coming in. Anybody seen the speech demo that happened on stage back in May? Those APIs are coming. They're working on them right now, actually, the vision capabilities that are being baked in. At a certain point, you're gonna move away from human capability or, you know, this is equivalent to a human expert at this topic into a superintelligence. It's like you pick any human expert on the planet right now, they're not gonna be able to compete with the capabilities of something that in this one narrow field of, you know, an implementation of one of these models. When we start coming to greater and greater capabilities across the board that, that sort of understand processing as the human mind does without the whole, you know, experiential context that we, we have by just being alive and living on a planet in a society.

Ben Wilson [00:39:58]:
But all the other skills that we venerate, Do you feel that at a certain point, it's gonna be impossible for Luddites to ignore? So a teacher that says, I'm not interested. I I had professors years ago that they refused to have, like, stuff like an email address. They just wouldn't use it. They're like, I don't see the point. You need to contact me, come to my office hours. Or professors are like, I don't use phones, like cell phones. Here's my my landline. You can call my office.

Ben Wilson [00:40:35]:
If I don't pick up, I'm not there. No. You can't can't contact me out of outside of that. And there's still people, plenty of people on the planet that are just like, yeah. I don't want any of this stuff. But there's a point at which a technology becomes so compelling that it's almost impossible to live in modern society without adopting that. If we get to that point with these technologies, where do you where does that leave the education system? If the response from feeding in that syllabus to that course is a rewrite of the syllabus and a critique of the nature of the instruction itself. So it's almost like an attack on the instructor when you're interfacing with this thing, and it it is fully capable of creating a more effective lesson plan and a means of interacting than a human teacher ever could.

Ben Wilson [00:41:29]:
And then there's metric data that proves through cohort analysis of people that use this and their their teaching methodology, they are academically more proficient by a statistically significant amount than those who attended traditional. Where does that leave education?

Daniel Hiterer [00:41:47]:
What I could say without oh, being speculative and, you know, without being too speculative, because it's a very interesting hypothetical. We have no idea. But the thing that I noticed that I think we should pay more attention to is this tension between the world of productivity and the world of education when it comes to adopting AI tools and, you know, celebrating the the promises that they give. In the world of of productivity, we appreciate how technology like language models remove friction from sorry, remove friction from our work. They make things more seamless, more frictionless, faster, easier, and, you know, we we love it when we try to get things done. But in the world of education, and it's the topic that I guess interests me the most right now, this friction is necessary. We need to struggle to learn just cognitively. We need something that's difficult, not too difficult, but not too easy to grow.

Daniel Hiterer [00:43:03]:
We need to engage with a complex problem. We need to, you know, think about a history essay question, which is, you know, to what extent factor a or factor b played into the rise of, you know, Napoleon in France. We need to tell ourselves we don't know. This is a problem that's difficult. Let's think through this. And when we start thinking through this and then we start writing about it, I feel like that's when we gain knowledge. And if that friction is removed and it's no longer there, then we don't grow as humans, which is you know, it might be a problem might be a problem. You know, I'm thinking of Google Maps.

Daniel Hiterer [00:43:49]:
We have a Google Maps so that, you know, we don't have to kind of know the map by heart and we can orient ourselves. So do we need to do we need orientation skills when we have Google Map in our pocket? We don't know. Maybe maybe maybe yes, maybe no. But there is at least one study that I remember from a while back that, that concerned the calculators and the multiplication table, and the question whether, you know, why just not use a calculator? And that's it. That study found that when people learn the multiplication table by heart before they go on to use the calculator, that helps them achieve better results. Because by learning the multiplication table and how, I guess, algebra works overall, we form these mental models in our head about how the world works, and then it's easier for for us to more deeply understand the the underlying currents, under, you know, algebra. And when we use the calculator, it's a useful tool that supports our mental models and helps us kind of achieve better results in terms of productivity, but, also, I guess, AI, it it doesn't concern productivity only at this point. It concerns the creativity as well, which is it adds an interesting layer to it.

Daniel Hiterer [00:45:12]:
But, overall yeah, I don't know if there is something that's smarter than us. Then what the question is, what do we need to know? Like, can we have just superintelligence take care of us, at all times? Can we just, you know, as people say, just lie on the beach and and have it take care of us? Well, I don't know. I feel like, you know, going back to what we talked about earlier about our curiosity, we always you know, some things we're okay not knowing, but we always want to find out more, period. We always want to find out more. We need to know what's going on, and it's it's really difficult to say how how things will go in the future. Maybe we see you know, we both see the positives and the negatives, but it, I guess, it depends what people create, honestly. It just depends what people just sit down and and create, and make and and release into the world and see what what implications there are. And the best we can do is just to have the educational consideration attached to it.

Daniel Hiterer [00:46:20]:
At everything that we create, you know, let's see how it impacts education and, you know, how how we can work with that new technology and integrate it in meaningful ways, maybe bypassing the shortcomings and and going straight to the benefits as much as we can. But, yeah, it's a tough one. We don't know. We want to I want to find that out. That's why I wanna do research. I want to just figure it out.

Ben Wilson [00:46:46]:
Yeah. That's admirable. I mean, I'm personally optimistic about the direction that everything is going in. And I I don't I get the the counter arguments that people have about, like, woah. It's just gonna be this tool that we can ask anything, and it's just gonna solve it for us. So we're gonna become dumber as a species. I don't buy that. And I don't think that that's a driving force biologically for us.

Ben Wilson [00:47:14]:
We've evolved to tackle problems and a person in isolation who doesn't have meaningful problems to tackle or new things to explore, new adventures to go on. You were always looking for that next horizon as a species. And in order to get to those next horizons faster, which I think is also a driving force for most of most humans, doing it in in a safe way and an expedient way to hit all of the things that we because without thinking through the impacts of things that we do that satisfy our inner nature of doing more, having more, building more. We as a species have done some things that we regret. Right? Everything from exploitation of the environment to, you know, wars being fought, because resources aren't as plentiful as as the people would like them to be. And having a super intelligence around that we can interact with that can weigh the consequences or potential consequences of certain actions and help solve problems that we biologically are incapable of having that knowledge base to go off of and do those calculations or estimations. That was something that brought this to mind for me. It was something we were talking about before we started recording about like sci fi influencing our minds or, you know, like, popular culture influencing our our minds with respect to entertainment.

Ben Wilson [00:48:56]:
I've been getting caught up on, For All Mankind, and I I'm watching the season now that they're on Mars. That's like season 4. And I remember, like, sitting back and thinking through the problems that they faced on their first initial landing on Mars in season 3. And, the nerd that I am, I'm just thinking, like, what are they doing for radiation protection? And I don't think that, like, that would be viable for them to to be spending so much time out on the surface and then having this this little hub that they built that's just sitting on the surface, like, they're gonna have so much DNA damage. And, of course, the people that are writing the show, they probably consulted a few people that are that, you know, know a little bit about this type of stuff, but they're not talking to every expert that's out there. They're not talking to nuclear physicists or, you know, or medical professionals being like, what is the effect on the human body of having a very thin atmosphere for an extended period of time? They thought they covered a little bit in, like, the moon, phase of what they were doing. And I just start thinking like, well, what if there was a superintelligence that you could be like, hey, here's our plan. Here's all 37,000 pages of our plan to go to Mars from, you know, NASA Ames Research Center and JPL.

Ben Wilson [00:50:24]:
Here's like the high level plan. Can you review this and come up with recommendations of how we can not kill people and not have these issues happen? And then having it generate an a modification to that plan that has safety considerations and environmental considerations, things that you could think through in more depth that has all of that additional context. They think about how powerful that sort of system would be to just give us an an additional layer of checking.

Daniel Hiterer [00:50:59]:
Yeah. It sounds to me like, you're describing an omnipresent parent for humanity. You know? It's like a guardian or just someone who could, you know, make sure we're doing things right and correct when we're doing things wrong. Just kind of watch over us, take care of us. Well, I don't know. Sounds like it. Maybe it omnipresent teacher, an omnipresent mentor. Yeah.

Daniel Hiterer [00:51:26]:
That would be, I mean, that would be good. If that happens, I would take it. If if we live to see it. Maybe we do. Maybe we don't. Michael, what do you think?

Michael Burke [00:51:39]:
I knew this would happen in this episode. I had, like I have literally, like, 18 questions that I've been, like, trying to reenter into the conversation. So that's what I think. But, yes, I agree. It'd be cool to have an omnipresent parent. Be helpful. There's one thing that I can't end the end the episode without, getting clear, though. So we talked about these access points.

Michael Burke [00:52:09]:
I intuitively know what you're talking about because I experienced them, but can you go a little bit deeper into what the hell they are? Like, I am in a current place in knowledge. I'm in an emotional state. I have a set of interests, and I know these things. I want to attain this new set of information. And the way that I envision that in my head is, like, alright. I have this structure, and I'm, like, adding pieces on top of it. And maybe the structure isn't geometrical. Maybe it's like a blob or something, but I'm, like, appending into my existing knowledge base.

Michael Burke [00:52:44]:
So to that end, 1, does information have to be appended to the existing base, or can it sort of live in isolation and float by itself? And then 2, for your access points, are they emotional? Are they where's the learner? Where do they need to go? Like, what exactly is an access point?

Daniel Hiterer [00:53:05]:
That's a fantastic cognitive science question. I'll try to answer it to the best of my ability, partly by kind of mirroring what I've read, from other people. I think you're spot on that not always, but sometimes, maybe even often, we integrate a new piece of information into our existing, you know, frame of thinking. Or the cognitive scientist, they call it, cognitive schema or the educational scientist. I don't wanna step on cognitive scientist, toes because I only know the educational part of it. But, yeah, we either either integrate the new piece to go with what we know so far, or sometimes what happens is we see a new piece that is so different that we adapt what we know to fit the new piece. So we see we see something new, and we're like, wait. This doesn't fit into anything I know.

Daniel Hiterer [00:54:03]:
So I have to go back to what I know and kind of rearrange pieces to make sure that the new piece of information fits. And that happens kind of almost, subconsciously, I think, if I'm using the word correctly. But in terms of how these, access points you know, how how access happens, my best guess is that it's emotional first, because I believe that feeling is the beginning of thought and there are many people who would disagree, but feeling in any case, feeling is a good place to start. I know that it elicits us a strong response in in our minds, and it helps things stick around for longer. And it sounds like, Michael, you're just trying to see how to transform yourself into a learning machine. Like, how how can I best learn the new information? Maybe I'm let me know if I'm wrong. But it is, I mean, it is a good quest to have.

Michael Burke [00:55:13]:
You're not wrong.

Daniel Hiterer [00:55:14]:
Yeah. I think it's just connecting things to to what we know. I don't know if things can live in isolation. I think it just goes back to to our overall cognitive schema and and and to what we know. And we we we make those associations whether we know it or not, and maybe it goes back to something that's familiar to us from childhood. Maybe.

Michael Burke [00:55:43]:
So it sounds like information doesn't live in isolation. It has to fit with your existing cognitive schema.

Daniel Hiterer [00:55:49]:
Yeah. I think information never lives in isolation. We have to make sure that, well, we have to make sure it's not it's not a conscious effort, but we have to make sure that it it just fits with everything else that we know.

Michael Burke [00:56:02]:
Yeah. Yeah. It's like an unconscious, like, reorganizing of the brain as you said. That makes a lot

Daniel Hiterer [00:56:07]:
of sense. Yeah. So I guess the strategy would be, you know, just to ex expose yourself to as many things as you can and see what sticks and what doesn't, roughly speaking. But at the same time as you do, I think it's useful just to notice these, path dependencies. Like, being like, okay. I like music. Why do I like music? Oh, because I'm interested in how it's produced and, you know, what what instruments are used and how the digital instruments and synthesizers work and everything. Oh, okay.

Daniel Hiterer [00:56:41]:
I'm interested in synths. Let's see, you know, how color is being altered in in movies and how color production works. You know, one is sound, another is color. Same principles. Okay. I know about that. Let me see how, you know, data management works at a large company. It's also kind of being on this big synth, which is a database and and tweaking knobs just to see what works.

Daniel Hiterer [00:57:04]:
And, you know, the the thing that I would love for people to think more about is that things are more connected than than we think they are. And I think, you know, what what I think happens is that people, especially adults, they sometimes close close themselves in a silo, and they form an identity that's, pretty stable, and they just think some things are not for them, I think, you know, sometimes the new things are worth a try because you never know what connections will will form with your existing cognitive schema, and maybe they'll give you more ideas.

Michael Burke [00:57:48]:
Yeah. That makes a lot of sense. Just thinking through your explanation, I can see my, like, data science brain using analogies about, like, search algorithms and and, like, optimal efficiency from getting from point a to point b for this, like, building of new knowledge. Yeah. That makes a ton of sense. Ben, do you have thoughts?

Ben Wilson [00:58:10]:
Oh, yeah. Yeah. I mean, I think what you just distilled is the point of traditional formulaic education. Or I think that's the optimal goal of it is, like, I've I remember at when, like, when I had my first kid interacting with, like, I went to go to the kindergarten classroom, and I was just like, man, what do they do all day long? Like, okay. I see there's there's, like, all this arts and crafts stuff, and then there's, like, a reading nook, and then there's stuff that they're doing, like, basic arithmetic. And and then they have all these projects that they're doing that are in all of these different disciplines to try to, you know, spark curiosity. And then at a certain point, as as my eldest kid got older, I noticed that it becomes more structured and more, like, just formulaic to the point where a lot of those other things are not being cared for. I'm like, hey, you're not you're learning, you know, your your English class, your math class, your history class, some form of science, but then everything else becomes kind of electives type stuff.

Ben Wilson [00:59:32]:
It's like, oh, this is an after school program if you're if you wanna do it. And I remember thinking just like, why isn't all of school like the afterschool program where you start having specialization earlier in the time? Because the kid that I'm talking about, she just started NC State a couple of months ago in their digital design. Specifically, it's like 3 d art for, stuff like animation and video games. It's what she's really passionate about, and she's phenomenally good at it. And

Daniel Hiterer [01:00:16]:
That's fantastic. Congratulations, by the way.

Ben Wilson [01:00:19]:
Where did she learn all this stuff? I remember, you know, buying her stuff like, oh, here's your your laptop that can do this type of art that you're talking about. I have no idea. Like, I'll set your laptop up up for you. I'll buy the software that you need and, you know, this crazy tablet that's basically a screen with a pen. And I was like, this is the stuff. Like, I I looked this up, and this is what they use at Pixar. And she's like, oh my god. This is amazing.

Ben Wilson [01:00:44]:
And then within a couple of hours seeing the result of the first thing that she made with it, I'm like, like, I'm not even remotely capable of doing anything like that. And then that sparked curiosity in me. I was like, I wonder if I could do some something with digital art, and I did a couple of like, did a like, drew, like, a landscape thing. And, and she starts critiquing. She's like, oh, maybe you do a little bit different here and, like, shading here a little bit different. And I was like, yeah. I can do that. I was like, am I like, did I think that I couldn't do any of this stuff because of the education system? Because I never took art classes.

Ben Wilson [01:01:25]:
I never really engaged in that. I'd always thought, like, I'm terrible at this. But I've I've seen stuff with other like minded adults who are, who are very curious and are in that mode. Like I've met a very small handful of them in my life where at the age of, you know, 67 years old, they pick up the guitar for the first time and they're terrible, but they don't care and they keep on working at it until 2 years later, you're like, woah, you can play that, you know, classical piece of music on that guitar, fingerpicking style, and that sounds amazing. And then they're just off onto the next thing. They, they don't wanna get stuck in some rut where they're just doing the same thing. Their minds feel stifled and frustrated like that. And I think at least for myself, emulating those people is my life goal.

Ben Wilson [01:02:24]:
It was like, I just, I want to be that guy in his eighties picking up a new skill and just trying to prove to myself that I can do it and enjoy the process of being terrible and then seeing myself get better over time. I love that feeling. And, I'm curious, like from a cognitive science perspective or educational science perspective, is that something that can be taught to people outside of just purely seeing, like, a men like, having a human mentor or somebody who's a somebody that you can witness and see. Oh, that seems cool. I wanna be like that guy when I'm I'm that age.

Daniel Hiterer [01:03:04]:
Yeah. I believe that it has to come from within. I think it's it's amazing what you describe. Basically, people showing you what's possible. Right? It you know, people around you showing you that graphic design is possible, and it's not something that, you know, that is just done like magic. People showing you that picking up a guitar at any age is possible. I think we humans do good at relating to other people, which is, I think, one of the reasons why humans need other humans in anything we do. Just see, oh, this is a being who is just like me with with hands and legs.

Daniel Hiterer [01:03:54]:
You know, they can do this thing. Oh, cool. Maybe so if I'm a person and they're a person, maybe I can do the same thing. Then the question is, yeah. Maybe some people put more hours into one thing than than others. But I think it's it's just fantastic. And I don't think that it can be taught no. Maybe I do think that it can be thought taught because it's a habit at the end of the day.

Daniel Hiterer [01:04:21]:
Right? The habit of trying new things and and habits, we can form them. If someone helps us form those habits in childhood, Yeah. That's good. But we can do it in in adulthood as well if we try. But that needs, you know, specific influences and just, yeah, other people showing us that it's possible.

Michael Burke [01:04:43]:
Yeah. I just wanna affirm that that side of the argument. Like, the single best learning tool for me is being around other people that know how to do something, and I always elevate my game to the level of the players. Like, I play a lot of basketball. If I'm playing in a bad league, I'll, like, jog around and, like, I don't know, shoot dumb shots. But if everybody's better than me, suddenly, I lock in and I play really well. And then if I go back to that bad league, I get reminded of, like, how much growth I can make in such a short period of time. So I think it's absolutely essential.

Michael Burke [01:05:18]:
Like, what you surround yourself with and what you input into your brain, it can't be discounted. It's so important.

Daniel Hiterer [01:05:27]:
Yeah. Fully agreed. It's the right environment. In terms of the people, in terms of information we consume as well. I mean, it's a whole topic, I guess, for another day, but the information diet that we're on is is really important. Just, you know, what what we see when we, you know, pick up our phone in the morning or or, you know, in the afternoon or evening, any time of day. But especially in the morning, what do we, you know, go scroll around, Which can be good as well in its own right. And and who do we follow and who do we listen to? What information we allow to reach our minds? It's it's fascinating.

Daniel Hiterer [01:06:07]:
And our body as well. Let's not forget our body just overall, just taking care of ourselves. But it's the environment is a great example of taking care of ourselves in a holistic way and being around, you know, the people who who inspire us to do new things. I think it's something to to aspire to.

Michael Burke [01:06:27]:
Yeah. I will over time, on that note, I will summarize. Once again, there's, like, 8 hours of topics that we didn't cover, and I'm mad about it. But, so things that stuck out to me at least, there are many ways to learn. It can be curiosity driven. It can be willpower. It can be via someone telling you to do it, and finding ways to sort of access whatever motivates that specific person is really important as a teacher. And we discussed this concept of an access point where you can sort of reach someone and start sparking that fire.

Michael Burke [01:07:02]:
I think we all know it intuitively, but, it's sort of hard to define. Information does not live in isolation in your brain. Your brains unconsciously rewires to make new things fit. And there are really important implications for that because you if you're trying to acquire something new, you need to go from point a to point b, and there's no sort of jumps. There has to be a continuous scale where you can incorporate that new thing into your existing world view. And then also education is super highly dependent on the age of the learner. So, Danny, if people wanna learn more about you, your research, where should they go?

Daniel Hiterer [01:07:40]:
They can reach out on LinkedIn, to be honest. I'm online once in a while. Just shoot me a message. I'm, Daniel Hitera. Last name is hitera. Just reach out. I'm here. Happy to talk.

Daniel Hiterer [01:07:55]:
Happy to tinker together. Cool. Yep. That's me.

Michael Burke [01:08:00]:
Awesome.

Daniel Hiterer [01:08:01]:
And also, Michael, well summarized.

Michael Burke [01:08:03]:
Oh, I appreciate it.

Daniel Hiterer [01:08:05]:
Mandible. Thank you. Thank

Michael Burke [01:08:07]:
you. Alright. Well, until next time, it's been Michael Burke and my cohost

Ben Wilson [01:08:12]:
Ben Wilson. And have a good day, everyone. We'll catch you next time.

Daniel Hiterer [01:08:15]:
Bye, everyone.