Foreign. Hello. Welcome to the Hilberg School Podcast. My name is Bill Selake. Him, I'm our director of technology and

I am here with Marisol Ornelas. How you doing?

I'm good, how are you? Who are you?

Well, who am I? I am the director of the Reach Beyond Scholars at Hillbrook. And I also teach advanced physics and I teach calculus, which is more in a phase of teaching applied calc.

Applied calc, yeah. That's amazing. So you are doing some amazing things with AI as a teacher. That's our theme this whole week. And right before we started recording, you're like, am I? Yes. Yes, you are. The way you think about it is like, is atypical and inspiring and like so incredible the things you're doing with students, the way you're using yourself, like, amazing. So let's just jump right into it because we could talk for like three hours about it and that's not what this podcast is. So this physics project, Safety, right, Like street safety. Walk us super briefly through like what that project is and then how you're like so brilliantly integrating AI into this.

Yeah, absolutely. I mean, my, my whole premise for the physics class was basically advanced physics and policy. It's like there's a little, there's obviously a little quote unquote mark. Right. And Vision Zero is an initiative in San Jose that basically is trying to reduce traffic related deaths to zero in San Jose. And because this area has been relatively in Sandown, San Jose, where we have our high school, has been a dead zone. A lot of the data here is not necessarily accurate. And we're trying to think about like, okay, how can we have students outside collecting data to see where intersections may have a higher collision risk? And so in reference to talking about AI and how that gets incorporated, one of the big things that I always tell the students is I want you to focus your social, emotional, intellectual labor into certain areas. And for example, there's a lot of ways that like, do we need to collect all these data points? And we had one group that was trying to do some analysis of collision risk off the 87 ramp because a lot of our kids from Los Gatos commute off of that 87 ramp to come off of the Julian exit.

Oh, sure. And that's only like half a mile west of us.

That's right. That's right. Yeah. And like, we want to analyze the ramp. I'm like, great. And I had one student like, okay, maybe we can go on the ramp and get out. And I'm like, no, no, no, no, no, no.

So we're not going on the on ramp or the off ramp of the 87 Freeway, right?

That's right. That's right.

And, you know, but the idea of that is actually really cool, is like, let's collect data with something that's meaningful to me as a student.

That's right. That's right. And it's like we use. I use this off ramp every day, and I'm like, yes. I'm like, are there tools that you can use to collect this data that will help you understand what's happening, the physics of that ramp? Right.

And so hang on, what does physics of the ramp mean? I don't understand that sentence.

Yeah. So in terms of braking distance and being able to understand, like, the distance in meters from the intersection, the width of the intersection. And so the physics of that was trying to figure out if a pedestrian walking at some speed could cross that crosswalk in time for before the. Before the distance and the time it takes for the car to reach that. That stop sign.

Okay. So the width of the road is actually how long the pedestrian needs to walk.

Right.

About how fast pedestrians walk.

Right.

And then breaking distance. There's all kinds of physics with that.

Right.

That's why you're talking in meters.

That's right. Yes, yes, yes.

As a physics teacher, meters per second squared.

Yes. If I use the Imperial system, it would be like physics blasphemy to do that.

Right, of course, there.

And so the data that they actually wanted to connect is like, okay, like, what's. Even though there's, like, speed limits, I always talk to them about, like, there's a speed limit and then what's actually happening. Right, right. Because if you're on 280 near low cell to, those people are not going 65. Often, people are going 80. So that's when the students said, oh, we should go there and, like, collect data to see, like, what the speed actually is when they're going off the off ramp. I'm like, yeah, that's probably not going to be a good. Like, that's. That's not a good idea.

You know, this is unsafe.

This is unsafe.

So what did. What did you guys do then?

Yeah. So students were able to then use AI to model what the speed and the distance of typical cars do on different off ramps and then collect an average miles per hour of how fast they come off the off ramp. And I think the biggest use of AI there was students were then able to use some source code, do a physics algorithm. They actually derived A physics algorithm. Use source code, plug in the physics algorithm. But I didn't want them to design a website. This isn't a class on website design. Right. And so I'm like, have AI help you with the source code to actually do the web interface. Because I don't want you spending 3 or 4, 5, 6 hours, 10 hours doing the website. I want you to spend time really collecting the data, making sense of the data, deriving an actual variable model so that you can plug in data points and then making sure embedding them that's legitimate, so that we can actually go through the derivation, make sure that's real.

This is so cool. This is like so cool. So the physics algorithm is actually like grabs real data, grabs ChatGPT created data, puts it into the thing the students are making.

That's correct.

Right. And then chatgpt or whatever generative AI is then building a website that's basically just the interface. Right. So they're not doing like hillbrook. Org trafficdata. Right. Like, it's not a publicly hosted thing. It's just like they're making something that HTML can open. So you open it in Chrome or Safari and then that gives you like a web interface that's better than like Mac OS terminal, where you have to like type stuff, which probably could work. Right. And then let's dig in a little bit with like, what does that mean? They make an algorithm. So I get the like real data of like laser gun or something. This is 50 miles an hour.

That's correct.

You guys had that?

Yeah. So no, they actually got. They got average data. They use different AI tools to get like average data of what usually happens to cars when they get off off ramp in terms of their speeds and the braking distance. They also use some GIS tools and other Google Earth tools to get the gradient of that off ramp.

Oh, cool.

Yeah.

This is what Judge is doing also with ArcGIS data, right?

That's correct. That's correct. Yeah. So it's a little bit of everything. They use some GIS tools, they use some Google Earth tools, they use some AI tools. And what I mean by physics algorithm is an actual variable derivation. So they actually work will use physics equations to derive an equation that says, oh, if the time it takes for a pedestrian to cross x distance is less or greater than the braking distance, then there will be a collision or not a collision. So they actually get to use physics equations. And when we talk about algorithms, that's the math behind what we're trying to do. And so then you have initial parameters. Okay, let's say car is initial speed of 50 miles per hour. We're going to put in meters per second. So 25 meters per second. And the distance is blank. And the distance of the pedestrian crosswalk is also blank. And the reason why I really value the algorithm or the derivation is because then that can be used on any single off ramp they want.

Oh, right, right. Because it's not just 87 Joules exactly.

They could find out the collision risk if they just know the width of the pedestrian crosswalk, the braking distance, all these different parameters. And so this is a model. And so a lot of what I teach is mathematical modeling. It's not perfect, but.

But it doesn't need to be the best.

No, I mean, is it ever perfect? I mean, science is never perfect. Everything is actually somewhat of a model. There's too many variables to be specific and exact. Yeah, yeah.

And if you did them all, then it's not useful for humans.

That's right. That's right.

So then you're grabbing all this data and I'm going to repeat this a bit. I'm like adding a little bit. Each step grabs real life data, grab AI generated data, which is like gets you in there. And then AI built web interface. And then you can basically say like, grab these data points or run like. Could you say like run all the data points? Yes, like run it like 500 times and see how many collisions there are with pedestrians.

Yeah. So basically you can then like the web page interface is really cool because you could actually go to any intersection and you could drag a little line that would actually measure how long the pedestrian crosswalk was and the braking distance. You could put the initial parameters of initial speed, the speed of the actual pedestrian crossing the crosswalk. And then you press enter. So you put all the initial parameters and it will say low, medium, high collision risk.

And part of what you're building then in that web interface is like, here's what low, like give me low, medium high, or give me A. Like 0 to 100. What percentage risk do we have of being hit?

That's right. That's right. And this was a student project. And you know what's interesting is student was actually a little bit like nervous. Like, hey, is it okay? Like, I can't get this data because we can't go measure generally what's happening. Like, you know what, use AI. Like use AI and find out like what are some average speeds here? What are some average distances, gradients, all these different aspects that are really hard to Sometimes just do a Google search with, because you got to go down like a rabbit hole. But when you use Chat or Gemini, like, okay, you do a prompt. It's really good at crowdsourcing. Right. And that's actually the biggest piece is where do we want the students to spend most of their cognitive load is like, I don't want them doing, spending three hours crowdsourcing. I just want something immediate so that they can then learn from that research. And that kind of like, that lends into like the project that I'm doing now. We're doing a big project on renewable sources and they're going to be designing their own energy system.

What class is this? Is this still physics?

This is still physics, yeah.

You're designing your own energy system.

Well, I'm not. The students will be. Yes, this is.

Your class is so cool. You are so cool. Keep going, keep going.

Yeah. No, so, and actually, believe it or not, it was actually about AI. Like it started talking about, we said like, okay, San Jose is bringing in a lot of AI companies or trying to get grants to small AI companies. And a lot of the residents are thinking about, okay, how are we building infrastructure and thinking about energy storage and energy distribution differently?

Well, because you grew up in this city, right?

That's right.

So like this isn't like what is let's bring AI to San Jose. You're like, no, no, this is my town, my community. And you're talking like about your people in your community.

Yeah, like I'm thinking, how is this going to impact like infrastructure for residential energy distribution? And when I, when I introduced the project to the kids, I was saying, I do not expect in two months of this big long term project that we're going to solve this big problem. But I do want us to scratch the surface and start thinking about energy storage and solutions. Because you know, we, the AI wave is here and it's going to be coming to big warehouses in down in San Jose or South San Jose. And how are we preparing to understand like the resource management of that, the storage and the distribution. And so that was the, the initial introduction of the project. And that's why we're thinking having the kids think about designing their own energy system. Obviously there's different phases to that. There's understanding power input, output and all other aspects before we get to that big prototyping phase. But right now we're in the research phase. And I was actually very upfront with them. I said, hey, I just created a huge research library for you all that has a Lot of scholarly articles. I did not spend four hours doing this because I am not going to spend all my cognitive labor doing that. But I would love for you all to then look at those scholarly articles or so called scholarly articles and then actually vet them.

So that's their job.

That's their job. Yeah. And I'm like, and I'm not hiding this because you know what? There is nothing to hide. I'm like this is a really good way of using AI. I just did a ton of crowdsourcing. I want to actually spend the time with you helping you understand the physics of every renewable energy source that you are all working on to explain. And so now it's your turn. Tag your it. Here's a huge library. There's videos and there's articles. Vet them and then spend time learning from the research. Don't spend time doing the research. Spend time learning from that research. And then we can work together on understanding the physics of that. And that's. That to me is a better model and modeling to them how I use AI versus kind of like pretending and oh, I'm going to help, I'm going to do this. But then hahaha, they're not going to know.

I don't have to do all the work because I have chat GPT.

Yeah. And I feel like there's, it's just better if it's just this is how I use AI and then I like to learn things. But this is a really good way. So crowdsourcing and research I feel is a really, really good way to remove the hours of doing that. And that can then be like shifted into spending time really learning and understanding.

At least twice, maybe three times now. You talked about like the cognitive load and how you spend your time and your effort and your hours. I think there's really something there.

Yeah.

About like it's less, it's less. Do we or don't we use generative AI? It's less of like trying to put one past the kids. And it's more like how do I want to spend my energy?

That's right. Because we all have a limited amount of energy. Right. I mean students are taking five, six classes in high school. It's a lot. They're doing a lot of extracurriculars.

Yeah.

I mean we've been talking how long now in education about, you know, students feeling like they're overwhelmed and there's a lot like the mental health crisis. We talk. I mean I've been in education for 23 years and we've been Talking about this for at least over a decade, you know, and how can we make more meaningful activities and projects and learning and not have it be additive? Right. Because project based learning, unlike traditional learning, is really, really cool. And it's a different type of cognitive load to be doing like three or four research projects in three or four different classes is. Is different than me. Oh, I'm just going to, you know, do school and take a few quizzes and tests, you know.

Right, right. That's. Yeah. You don't have the mental labor of like monitoring all your projects.

That's right.

I'll just go in and that's. Actually, we've talked about this before. Like, I went through calculus four.

Yeah.

And not one single math teacher ever taught me why this matters or how this applies or like, whatever. Like, all I ever did was, this is the next page in the book. Why are we learning this? Because it was the next page in the book. What are we going to learn next? The next page in the book. Why? Because it's the next page in the book. Why do I need to learn this? Because it's the next page in the book. Right. Like there's. And eventually I switched to music for my undergrad because, like, I knew implicitly why I'm learning the thing. It was always applied classes.

That's right.

And like, it feels like that is non negotiable for you as a teacher.

Yeah, that I. It's non negotiable for me as a person because I myself am just bored out of my mind.

Fair.

You know, and right now in applied calculus, every student is doing a different project where they're, you know, understanding that, oh, calculus is just understanding the study of change. I actually have a poster in my classroom that's like, study change, find meaning. You know? Right.

Yeah, yeah.

Because I hear kids be like, I don't understand why I have to take calculus. I'm never going to use this of my life. Like, ooh, hold my drink for a second. Like, I'm actually like, yeah, hold my Lacroix for a second. Because I'm actually going to tell you that everything you actually study is probably. You're probably going to be studying some kind of change, some delta and something. Right. And so every kid right now in applied calculus has their own topic that they're really interested in. And we were using AI in class today. Like, today was a research data modeling day where they're trying to get data sets and then create a mathematical model, like trying to understand if their data is more quadratic or linear exponential. And, you know, when we were doing the graphing, one of the students were like, oh, like, is there a way that I can highlight, like, the inflection point, like this critical point? And I'm like, you know what? I don't know. I'm pretty sure sheets can do that. So I'm like, I just popped up chat. And I just said, like, chat. How do I highlight a certain section of. Of a line graph in Google sheets and be able to highlight a critical point or an inflection point, right? And then I was like, all right, sit here, let's look at these videos. And then student just said, all right, I'm going to go outside. I'm going to watch these videos that were crowdsourced through AI. And then we figured it out. And then I remember, like, we figured it out and I saw the student, like, I actually tagged them. Like, look, here it is. It works. You know, because then I tried it and it worked and it worked and looked really cool.

That's really cool.

And it was like a minor thing, right? But. But in terms, it sounds minor, but in terms of representing research and data, it's really cool when you have a line graph and part of it's selected, and then you can actually highlight certain points that mean something. In terms of policy for the student, it meant, like when strategic funding came in for unhoused, but it was intermittent housing, not permanent housing. So that was actually a really important part of the research for that student because they're researching homelessness and the rate of change of the unhoused folks in San Jose over the years, you know, and we're like, okay, so that actually matters in research. And just the visualization piece.

Oh, oh, yeah. It's not just. Well, that's where suddenly math stuff, like median versus other stuff mode, average, whatever,

or like R squared or. Or standard deviation.

Yeah, yeah. Where suddenly that has meaning to it.

That's right.

Right. I learned all that stuff and guess how much I remember that stuff.

That's right.

Almost done. Because there was no meaning behind it, right?

And now it's like, oh, wait. And like that student said, like, wait, why was there this weird, like, critical inflection point there? It doesn't even make sense with the model. And then we went down, like, a policy rabbit hole and figured out that during that year there was a big strategic funding for intermittent housing.

Oh, interesting.

And that's why it just dipped and then went right. Shot right back up a couple years later.

You can actually show, like, here's what Funding and policy looks like and the impact it has.

Right. And then the student's like, wait a minute, this was just intermittent housing. I'm like, hence why it didn't continue going down.

There it is.

Right?

There it is.

And I'm like, this would be a really. That's where the whole idea of using chat for that second is like, hey, let's, let's actually highlight this point because that point is actually part of the main meat of your research.

Yeah.

Like, this is where we're going to talk. Oh, policy funding. It was also a campaign year. Like, there's a lot of things behind that. Right. And now that students like emailing path and trying to get data on, like the rates of permanent housing and rehabilitation

versus intermittent housing, that's like meaningful work. Yeah, yeah, yeah.

And I don't want them to spend time trying to figure out and toggling things to get that point. Just go down, chat and figure out how to highlight that and we'll get it done. And let's keep moving.

Wow. So it's actually using that generative AI to help literally highlight the story.

Yeah. And I think, I think it's important to note that, like, there are times where AI is like helping us build a website. There is times where AI is just helping us crowdsource how to do some small detail on a line graph.

Right.

You know, and it has a wide spectrum of use, but it's still doing the thing that I'm not focused on teaching. Right. And it's just helpful. But it's not something I'm not teaching a web design class, I'm not teaching a Google Sheets class.

And yet you can do these really advanced things with websites with Google Sheets.

That's right.

Yeah. Superpowers when you need them.

Yeah. So it's. It's the end. Honestly, I have noticed that the more I make AI integrated into the way we operate in class, the. The less they're not as nervous, like, oh my gosh, can I use them? Like, yes. I'm like, this is not physics. You're using this to get data. Like, I've had students say, is it okay if I, if I used this AI to gather some data points? Because we could. Could not get reliable data from our motion sensor. Great. Go for it.

Yeah, I love that. And the thing that is inferred in what you're talking about, that I'm going to name right now is that you've completely bypassed the whole us versus them thing of, like, students trying to cheat using ChatGPT teachers saying, I gotcha with ChatGPT, right? Like, it just. That's not the issue because you're being transparent about how you're using it. You're using it in, like, really meaningful ways. Students are saying, hey, this data isn't working. And you're like, perfect. I've got a solution for you. Right. It becomes this, like, this tool that everybody is reaching for and having conversations around.

That's right.

That's so beautiful.

That's right. In fact, I've gotten called on a few students. Like, I'm like, yeah, just go ahead and use chat. Why are you always biased with chat? Can't we use Gemma? Yes, you can use whatever you want. I'm like, but you know what? Some of the stuff that we're using right now is not reliable. So just go and see some averages. Don't just use it one as one time as well. Like, use different. Different types of AI sources and see if you can get more of an average. Like what? Like, if you. Or even prompted a couple times differently to see if there's, like, some, you know, inconsistencies, you know, and we can note that as well.

Oh, that's so fun. Yeah, that's awesome. I'm so excited to see what you and your students are going to keep doing.

Yeah, thanks. Yeah, I'm excited, too. I'm actually really excited. And this type of. This type of, hey, I'm not trying to gatekeep it. I'm also not trying to police it in that way. It removes so much stress from me, having to focus always, like, hey, are you using chat or using chat? I'm like, you know, and I do understand that, like, the way that I'm teaching lends itself for them to use it in that way. Right. And the more I integrate and I'm just transparent about how we use it and how we should use it in class and we model it, like, with them, like, because then it doesn't become like, oh, my gosh, like, Ms. O has the Chat GPT website on. I'm like, no, let's actually see what it says, you know? Yeah. And so I'm just. And I notice initially everyone's like, why are we using this? It's like, we shouldn't be in this class. Like, no, it's okay. We're okay.

Awesome. Thank you for joining us on the podcast.

Yeah, absolutely. Thank you,

Sam.