DR0000_0114
===

Sean: [00:00:00] Hello and welcome to Teaching Python. My name's Sean Tibor. I'm a coder who teaches, 

Kelly: and my name's Kelly Schuster Perez and I'm a teacher who codes. 

Sean: Yeah. And so this week we are at Picon US 2025 in Pittsburgh, Pennsylvania. We are actually sitting at sort of the end of the first day of talks and we've created an open space with a room full of.

Learners and educators and content creators all interested and curious about, uh, education and Python and how those come together. And today we thought for a change, Kelly and I would, uh, take a moment to listen and hear from the community, um, and participate in the conversation and stop talking for a moment.

So, um, we are going to ask each of the people here in the room to share a little bit about their experiences with AI and education. So the first question we're gonna ask them is. What's their personal use of ai or how are they using AI today? And then the second question is, how are they using AI or seeing [00:01:00] AI in education in whatever way that means to them?

So without further ado, we're gonna go to our first, uh, speaker, uh, tomorrow first for the second time. 

So once again, uh, my name is Luis. I, uh, I work for the University of Washington in Seattle as a Python instructor. I also do, uh, uh, Python coding as a freelancer. So AI in my job, especially as a freelancer, is becoming, again, very pervasive.

It's something that, like I work for a startup right now and, uh, it's one thing that allows me to produce a lot more code than I would be able to otherwise. It's, uh, it allows me to debug that code. Sometimes it allows me to just produce stuff that works right away. So it's becoming like, um, it's almost like being my own so development team where I am the, the project lead, uh, in teaching.

One of the things that I'm doing nowadays, because students ask me, uh, the first thing that I tell them at the beginning of a new class is that, uh, I am not a cop. I'm not a detective. I'm not gonna be running their code through, uh, like [00:02:00] any, any analyzer to, uh, figure out if they're like using any AI tools.

I believe, uh, AI is here to stay. Um, I cannot, I'm, I'm part of the school that thinks that it's becoming an extension of who we are as humans. So even the, beyond the topic of education, it's just, it's just. Slowly, but surely not, not that slow actually. But it is, it is changing everything. It is changing the way like we do our work.

It is even changing the definition of what a software developer is and what a software developer does. So trying to block it, trying to, uh, uh, like, like somehow police it, uh, it's just not gonna work. It's gonna, it's gonna take us over. Uh, what I do tell my students, uh, especially, I mean, we have like a three part certificate.

The first one is for people that are just getting into, into Python coding. Is that, uh, I mean, consider the fact of how, uh, using AI tools is affecting their education. If they give me an assignment and the assignment is working, if it's meeting the requirements, they're gonna get the full grade. Uh, they have to ask the questions, am I learning something from this, or did [00:03:00] I just like enter a prompt and I was lucky enough to get something that worked, that pass the test?

Uh, um, is that worth the money that I'm, that I'm paying in order to be here, uh, to possibly learning? I do think the, the concept, I mean the meaning of what, what learning is for us is just gonna continue to change. So yeah, something that somehow we have to embrace or at least like a brace for it.

Hi, uh, my name is Phil. Um, I, uh. I started out, uh, after I got my master's degree in Entertainment Technology, I started a company, uh, working primarily in Python making games and entertainment. Then I worked for a few startups and had an industry career, and for the last four or five years I've left that industry career and become a teacher.

Um, I helped to run a, an artificial intelligence camp for high schoolers through the Boys and Girls Clubs of Western Pennsylvania. And I've also been teaching, uh, at Carnegie, [00:04:00] Carnegie Mellon University and at Chatham University for the last couple of years. Um, and so, uh, the questions we were asked, asking here two questions.

One was, how do I use it myself? Uh, one is just sort of like in the invisible way that it permeates our modern lives that like Amazon recommendations and the next video on YouTube and stuff like that. Um, as well as for me personally, it'll be like the auto complete in pie charm. Um, that's just like, you have to probably take steps basically to turn it off.

'cause the minute you download it and start working with it, it's gonna want to start auto completing your, your code for you. Um, and for me as an experienced developer, that's pretty handy. I know what it's saying. I know how to read it and I know how to use it. I know whether or not that is in fact what I wanted to do.

And I know a fair bit about the consequences of what, like the, the, the performance impacts of what it's, what it's saying for my students. Um. It's much more problematic that it helps them in that way. Um, I [00:05:00] talked yesterday a quick lightning talk about one of the tactics that I've used in developing my college courses is to evaluate for everyone.

At the end, we do a, uh, industry standard job interview at a whiteboard where we just talk about the code and I get to, I get to see for real how much you actually understand the code that you've been handing in all semester. And in a couple of cases, it's quite illuminating that you, they don't have any idea.

How the code actually works. Which gets back to your question that you asked a moment ago before we started recording of like, is, uh, if, if you don't actually write the Python, do do, but there's good working Python there. Uh, have you learned anything? And it's an interesting question. I, when my students can't code on the whiteboard and can't think their way through a problem, I would be tend to say no, however.

I've written an entire career of stuff that has all worked perfectly fine on the co, on the computers, and I have outside of one college course, [00:06:00] never touched a line of assembly code. I've written an entire career's worth of stuff of code that I don't know how to read or write, and yet it works perfectly fine.

So to what extent is AI just basically turning Python into what has been assembly for my entire career? Um. Obviously debatable and there's a lot of people in the room, so I should stop hogging the floor. Uh, but yeah, that's me.

Um, hi, I'm Keith, uh, member of Python, discord and Education and Outreach work group with A PSF. I'm, don't use AI in my day-to-day, too frequent, uh, because I just don't know how to utilize it. And it just, it pausing to move to something else to move back takes time and I don't get it. But in the Python discord, we have seen that AI is being.

AI is likely a reason. We are seeing a drop off in many questions, which is probably a good sign because many of those questions were very basic intro questions. Great. And that means that [00:07:00] some people are able to get their questions answered without needing to wait for a person to be online to answer the question for them.

But there is that other question of like, do they, uh, learn how to problem solve or have they just removed a problem?

Hello, my name is KK Kwa. Apologize ahead if I have too much to say. Uh, I've been an educator now 12 years in a private school, but before that I actually worked for an AI company. I actually had people that worked with neural nets and genetic algorithms, which we don't talk about much these days, but we used it for stock picking.

So it was a quant company. I never really did any AI there. I was their data guy. Okay. But I do a lot of AI now. So my take is this, um, I saw, uh, pretty eye-opening [00:08:00] Ted talk from Dr. Uh Barat, a nun. And he basically had three sets of students he tested on, one that did no search engine whatsoever and solved the problem that way, and one that just used Google, and then one that used ai.

And then when he tested them, assessed them on knowledge later. The one that did the worst was the one that used ai. And the reason, uh, uh, he's, he talked about later was that. It effectively was just like finding out the answers and just copying the answers down. It really didn't, like, they didn't really learn anything from that.

So whereas the one that did the best was actually those that just did the Google search because we actually put some effort in, found a good answer, and then learned from that. So that's where I am at with having students use AI [00:09:00] to get answers. I think. They can do searches, but I'm not big on them just getting the answers.

Um, in terms of what I do with my students in ai, um, I usually start off with the IRIS data set. People familiar with the iris data set? Yeah, it's, it's been around. It's a great starting point. I think you can go to, uh, GitHub and just, uh, search up Iris dataset, Jupyter Notebooks. Which is also part of that, and it'll kind of walk you through how to take a data set, load it up, uh, run training, testing, all that good stuff, talk through various, uh, models, you know, k neighbors, all that stuff.

So it's, I think it's a, it's an excellent intro into the world of ai. For those who want to teach AI and have never done it, I would absolutely suggest that. I've used it [00:10:00] now for, I guess two years in my high school. Great success with it. It also has other data sets past the Iris. I think it's like a Boston housing, and then there's one on real estate that comes after that.

I forget what it's called. So that is to me the best starting point. Okay. Um, I hope I'm not boring you. I'm gonna quit right here. Um, pass it on. The next one. All right. Um, hello, my name is Daniel. I am a lecturer at the University of British Columbia, and I also work in developer relations as a data science educator at Poit.

And I've been teaching, I've been with the carpentry since about 20 12, 20 14. So teaching related for a really long time. Um. How do I use AI today? There, there's a couple of personal ways I use it and professional ways. Uh, personally, you know, I am, I'm fully on board with what Gen Z and [00:11:00] Alpha are doing.

It is not a full substitute, but a pretty good one for therapy. Uh, so if you just need a talk out through whatever you need help with, um, it's, it's actually really great. Um, but other things, you know, I'm a runner and you know, very recently I've had it. Put together a physical therapy plan and running training plan.

So there's actually a lot of uses outside of coding, which is something I think everyone can just think about. Um, when I do coding, uh, personal coding projects, mostly I treat it as a more direct way for Stack overflow searching. So there are things around a project that I've had where I want an Instagram recipe and get me a shopping list from it.

Um. You know, there's no way I'm putting together a code for a new library. Um, so that helps with a lot. Um, other things that I've done is also creating, uh, subtitle, translations for movies so I [00:12:00] can actually read 'em in English. Um, you know, open AI's whispers actually really good at that. So, um, those are some of the things along those ends for people to think about.

But in the classroom, uh, one of my undergraduate classes. I've fully just embraced it to like the max in saying, do use it for your homework. You can have it for the exam. Students kind of like, are you sure? Like they don't really believe me, but yeah, they can take it for their midterm or their final exam, and I.

The trick right now is I've been iterating on this for two semesters so far, and you just tweak the assessments in some way. So, um, a lot of it is trying to get them to write the code, that's fine, but maybe you do have to figure out clever ways to assess them properly. Um, one the analogies I'd like to think about is, you know, we teach mathematics in school and math isn't really math.

It's really supposed to be a way to teach you how to. Reason and problem solve. [00:13:00] And so AI is sort of the same thing. Um, we don't really teach you how to code for coding's sake. It's reasoning and problem solving and breaking down problems. And so, um, that's really, I wanna say that's really why we're teaching those things.

But AI is just a tool to get the work done. So that's my little bit of food for thought.

Hi, my name is Sophia. I'm a software engineer full-time, but I work in a couple of classrooms in a volunteer capacity. Um, so at work I'll often use Claude as a tool, kind of like a sounding board because I work fully remote. So it's kind of nice to have that extra feedback, um, when others are asynchronous.

Um, I also, similar to Daniel, use it kind of as stack overflow, but I've become very frustrated with using any sort of AI for code generation. I just find it's not [00:14:00] accurate enough currently. Um. And then as far as education, I work with high school students. We tell them not to use it at all. Of course they do, and it's very obvious.

We can tell when they do. Um, but I've found that they don't know how to fix the problems that the AI is generating. And so in that sense, they're not learning the critical thinking skills required for programming. Um. And so that's something I'd be interested in. How do we, how do we teach how they could use AI without trying to generate code that doesn't make sense to them, and it in fact, doesn't work.

Um, and then in another capacity, I work with adults anywhere from like 18 to 60 learning intro to Python. Um, and for them there is part of the course that integrates how to use an AI library so they can kind of. Feel [00:15:00] that they're using AI without promoting something like chat GBT to generate the code itself.

Um, part of that program also uses some like internal technology that you probably couldn't use chat GBT to figure out. And so I think that's also an important way to force people into learning programmatic thinking. Um, and then I guess just overall though, I do think. AI is important to teach in some capacity, um, because we're all being exposed to it now and I think there's a lot of misinformation.

Um, but we all need to have some base understanding of how AI operates.

Hi, I am George. I'm a, um. Professor of practice at Lehigh University in the computer Science and Engineering department. Uh, [00:16:00] professor of practice as a title basically means that I had a long career in industry and, uh, hopefully now I'm applying some of that, uh, knowledge and experience in the academic environment.

So, uh, I'm actually a chemical engineer by training, and so I, I was asked to teach the introductory programming class, which is taught to all the, uh, engineering majors in their freshman year. It's a very large class. Typically I have between 280 and 300 students. I, I do have an army of TAs, uh, about 30 of them actually.

But, uh, yeah, it's still, it's still a big deal. Uh, and, and the class has, has changed a lot since the advent of ai. Um, when I first taught the class a couple years ago, there were no exams. I gave them engineering problems to solve using Python. And, and you know, that's really the purpose of the class really, is to teach them engineering, problem solving python's, just a tool to do that.

Uh, but then [00:17:00] AI came along and, and everything started to change. Now I view AI as a very positive thing in the general sense, at least it has the potential to be positive. I've used it myself quite a bit. It's great for product recommendations. You know, if you just want ideas on something that you're thinking about buying and you ask it to give you 10 ideas, it'll generate 10 ideas.

And, um, and I've seen a big improvement over the last year or two, uh, that the number of hallucinations has gone down quite a bit. The quality of recommendations generally has gone up. So it's, it's getting, it's getting better and better. Um, and that's, that's good. At a personal level, I've, I've found it quite useful.

Uh, I use it in my coding a lot. Um, like, like others have already described of, you know, it's become my go-to coding assistant, uh, replacing a stack overflow in sites like that, you know, when you, when you get to a certain point when your, your head just gets full of a bunch of stuff, you, you can't remember which of those 10,000 mat plot lib options you're looking for, right?

So you just ask the [00:18:00] chat and it just comes back and tells you right away. So for things like that, it works great. Um, I did use it for something a little more, more ambitious recently. Um, I wanted a way to find out if the students were copying each other's assignments and, you know, there's a program out there called Moss that'll do that for you.

But I, I had some different tweaks I wanted to put in. I thought, well, uh, do I wanna download the source code and spend a couple days tweaking it? And I, I didn't wanna do that. So with, uh, with Claude's help, okay. Claude is my go-to, uh, GPT as well. Uh, I was able to build a simple similarity analysis tool in about two hours.

You know, the purpose of the tool is to see if the student's assignments are the same, and I'm gonna look at 'em to verify that. Anyway, so I was pretty comfortable that, that, that it worked pretty well and I, you know, I, I understood everything it did, and again, I, I could have done it myself, but it would've taken me a day or two, and it took me an hour, a couple hours, uh, with, with the assistance of the ai.

So, no question about it. It's a powerful tool and I, I [00:19:00] want my students to be able to use it and I want them to understand the, the value of it. The challenge though. Is that for brand new students when given a problem, um, my sense is that they're, they're not, they're not using it to learn. They're using it to avoid learning.

Yeah. So they'll, they'll get the problem and they'll just dump it into the chat, GPT or whatever, and then they'll just take the answer, copy and paste it, and in their mind it's done. And when I remind them that, you know, it's not enough just to copy and paste, you have to understand it. They will, in some cases, dutifully read the code.

Try to make sense of it and it comments its codes very, very thoroughly, right? So there's lots of comments there. So they will delude themselves into thinking that they understand it, and I've even quizzed some of them and said, well, you know, explain what the code does and they will read me the comments and I said, no, no, don't read me the comments.

Explain what it does. It's pretty clear they really haven't internalized it. They're just. [00:20:00] Uh, they're just using it to, you know, to avoid learning basically. So what, what have I done? Well, I've, I've initiated having exams now, which I hadn't done before. And, uh, they use chat sheet PT on the exams. So then I had to basically create in a computing environment where there was no connection to the internet.

And the, the exams are now done on those computers, and now I can tell very quickly if, if they've only copied and pasted the assignments, they do very poorly on the exams. So, uh, naturally, you know, when you give students an exam, they want a practice exam. And so I, I have these question banks of, you know, hundreds of questions and I didn't wanna be bothered with writing a practice exam.

So I use Chacha to write a practice exam. And, and you know, you have to be careful when you do that because you know there are hallucinations, right? And I'll say, why is it asking me about this thing that we never covered in class? So, you know, you have to. Look through the questions to make sure they're all legit, but still, it's a whole heck of a lot faster to do that than for me, for me [00:21:00] to, to generate a whole practice exam for, you know, these 200 questions or whatever.

Um, let's see, what else did I, yeah, so going forward, I would like to do, uh, other things to change my assessment as well. And one of the things I, I would love to do oral exams. The size of the class makes that problematic. I'll be training my TAs. I think with the help of my army of TAs, I may be able to do some oral exams.

But another, another suggestion that someone else made, uh, was to have them, um, have the students record a video or have them answer essay questions. Now, answering essay questions about computer science assignments is a novel concept. I. Hadn't really thought of that, but it kind of makes sense. Right. I can, I can come up with a list of questions that, that will, they'll have to talk about to describe their assignment and the thought process they went into to writing it and that, you know, that hopefully will help.

So I'm still trying to [00:22:00] figure out this whole new assessment kind of strategy, but, but I really think that, uh, you know, if I can get the assessments down, uh, having them use chat, GPT actually can be a very positive thing. We just have to train them. So that they're using it to help them learn rather than, than to avoid learning.

George, I think there were a couple folks behind you at the table that may want to speak also and wanna give them an opportunity if they chose. Yeah, go. Go first.

Hi, my name is Phil and I teach at Mesa Community College and I do Intro to Python, more Advanced Python. We also have a foundations of data analytics course and personally. Like most, I use it as kind of a working version of Google because Google sucks. And as we heard this morning, apparently that's on purpose.

Um, don't worry, we can trust all the companies behind the ai. Yeah, yeah. Oh yeah. It'll be great. Um, but [00:23:00] in teaching, it has become a problem. Now, some of them are really obvious for the beginning, students. From some library that doesn't exist, import this and then never use it in their code. First line gives it away that this is baloney.

The the problem is, is in the advanced class, it's not such a big deal. They're there. They wanna solve problems. Unfortunately, for those folks who have never done programming before, it could be very intimidating and they just wanna get it done. And you know, I, it's like, no, the point is not. To create a lottery thing, it's to understand how loops work.

It's understand how lists work to give you the tools and, you know, chat, GPT, Claude, uh, Gemini, co-pilot, whatever. They're great once you know what you're doing. But for the intro students convincing them that they need to understand what it's doing. Otherwise, well, as I said, if chat GPT can do your work, you don't have a job, [00:24:00] somebody, they're gonna replace you.

The people who will do well, understand what they're doing and use the AI tools to make themselves faster and more efficient, but they have enough knowledge to know, oh, that's wrong, or that's not efficient. Uh, vibe coding, I'm telling a lot of my folks, if you love debugging code, wait about five or six years and there'll be more work for you to do to clean up all these vibe coded pieces of stuff that they put out there that nobody, including the AEIs can figure out what it's doing.

And they need to upgrade it. Um, so it, it's, it's interesting. There's some interesting suggestions here. Um, again, it's a time issue. Can I watch through a whole bunch of videos? Can, and in many cases I'm doing live online or in person so they can be remote. How do I have 'em in the classroom to do a test to prevent them from using ai?

I would love to let them use it if they would use it as a learning [00:25:00] tool. But they don't wanna learn to solve problems. They just wanna get the assignment done. And I understand for community college students especially, a lot of them have a couple of jobs. They may have new kids, they've got a lot of other issues, but they're not gonna get a job if they don't understand how it works.

I, it's, it's, there's a lot of different things that come up and I don't know how to answer that. Oh,

hi. My name is uh, Dana. I actually teach creative coding in an art department. What's funny is that like some of the problems y'all have, I don't have because of the luxury of teaching along painters, so I can have them in class because all the painting faculty can have them in class and I can have them be in a classroom for three hours.

'cause that's how long a painting studio class lasts. And so when we, when I teach coding, it's like, I'm gonna show you something and you're gonna do it now in front of me. And I've got all the computers reamed in a nice panopticon so I can see all of them. [00:26:00] Um, but then out of that, um, after that I make them use ai.

I use it every day personally. I have a lot of really dumb things I have to write for the university. I don't write any of them anymore. I'm just, it's a waste of time assessments that no one's gonna read. Right. Just so I do that, a lot of that, that way. Um, I use, I mean, it helps me code a little bit similar I think to a lot of people.

Um. But, um, but one thing I struggle with, um, because I mean, I used to be a, I was a web developer in the nineties. At one point I got paid a lot of money to compress video, right? That's kinda a joke now. But, um, right. So I've made, I always been made obsolete periodically. Certain skills that I had that were expensive are now stupidly easy.

So, um, I also have an ongoing conversation with my students. It's like, okay, this thing is coming. What good are you? You know, like, I'm a teacher. What good am I? Right? Like, we can have this conversation. It's ongoing. Um, and it's an interesting conversation for them. 'cause the first time I asked them, they kind of freak out.

Um, but since it's a [00:27:00] conversation, but I, again, I have the luxury of being, um, in person. The flip side of that is, is that I am taking some grad courses in data science to sort of like, but they're online and like, um, is my friend. 'cause I have no friends in my class, right? I'm, but it's like, I've got some plants and I got a desk and then I've got a robot and that's like, it like.

And I don't know who to ask. And it's like, you know, it's an online environment. So if I, I could email a TA or a professor, that could be four hours before I get back, so why wouldn't I ask? You know, and it's as reliable as the people I used to sit next to in class also, you know, that it's not that great, but it's, you know, it's not bad.

Um, so it's, and it makes, and it also makes it a little less weirdly, it makes it less lonely to like, take a class if you can engage with. The chat as like, what do you think that means? And I don't understand what this is and you know, otherwise I just end up on YouTube anyways. So, um, yeah. Where's your students' perspective on, uh, [00:28:00] LLMs with regards to image generations?

Since they're all painters, right? Like, well, some of 'em are and some of 'em are actually most of our students end question. 'cause Phil wasn't micd up. Can you just repeat the question you asked me? Oh, sorry. What is their, uh, what are their thoughts about LLMs and image creation? Yeah. Um. Most of them, first of off, most of 'em are gonna end up being graphic designers.

Or at least that's what they think they're gonna do. Um, I, they, they all are by and large don't like it. Um, but then I'm just like, well, so what? Like, um, especially when I do demo Photoshop, 'cause Photoshop used to be a three day demo, it's 10 minutes. Like I can rock out assets for video animation 'cause, which is also something I teach, I can knock out all the assets in front of them because why would I prep?

Because I can show them how to make assets in real time. 'cause it does it so fast, you know, and I'm rocking out all that stuff and then we throw it in after effects and I'm animating it. It's insanely fast. So that like, um, I just need them just to understand that it's just there [00:29:00] and they don't like it.

I don't, but also it all, it does my, the downside to it as an image creator is it, it has a thing that it does and a lot of things it doesn't do. So that's why I want them to use it is because. I know AI looking stuff, because I play with it every day. I look at it all the time and I'm like, Ugh. And I'm sure most people can, but it's because that means, 'cause it's doing a thing that means it's not doing other things.

And so I want them to understand that so that then there's value in doing the things that that thing doesn't do, because that thing's gonna do its things and the people who want it are just gonna buy the cheap AI version of the thing. I hope I'm making sense. I'm using the word the thing a lot. Um, but.

Um, but yeah, they're really resistant. I would say my students are incredibly resistant. That's why I make them use it, is because they, all of 'em are like, no, I don't wanna, and I understand why, but I'm like, no, no, no. If you're not using it at all, then you don't understand what it is that you're afraid of or what it is as a tool.

Um, 'cause if you use it and don't like it, then it, that's fine. But now you know [00:30:00] what, why or what it is. Yeah. You could, uh, sorry, I'm not, still not micd up, but like, uh, you could be learning to be a jazz drummer, but you still. Want to know like what a drum machine can do. Yes. Right. Yeah. You know, and it's like you just, just because the existence of drum machines doesn't take away the need for jazz drum drummers.

Right. And they have a different feel than a machine does. You know? So I, sorry, I'm hogging the floor again, but, uh, yes. And Can, can you hear me okay? Um, hi. I kind of just walked into this conversation about, um, AI and LLMs and, uh, it's like an ongoing field, so very complicated. Um, my name is Mason. I'm a, uh, I guess technically middle level software dev at Amazon. Um, I was promoted in a server, so, um, I guess my experience with these conversations is primarily from an industry perspective.

And how, um, [00:31:00] uh, companies are, you know, trying to figure out all this stuff. Even, even tech companies struggle with, uh, this, this new technology. It kind of came from a little left field for some companies. You know, different companies had different strategies for dealing with this stuff. Um, I've noticed, particularly from an interviewing standpoint, um.

What seems to be happening is we, uh, a candidate will receive some sort of online assessment, um, that will be online and, um, it'll be like two leak code questions. So like, find, find the biggest number in a list of numbers or something, you know, basic. Um, it ranges from complexity depending upon a lot of factors.

So, um. They like, that's just online. You know, that's a website link there. There might be some protection around it. Um, but largely speaking, you can take that copy, paste the text for that prompt and just, uh, plug it into chat. GBT. You know, like what, uh, what's stopping a candidate from doing that really ethics at the end of the [00:32:00] day.

Um, so they'll get through that and then they'll have a wonderful conversation with a, uh, non-tech recruiter. Um, so someone who's just like, what are your skill sets? What are the keywords? What are the buzzwords of the time? What, what is this role looking like? Um, all that stuff. Um, and then they get to me and, um, I'm.

There, there's no formal policy around, uh, technically l LMUs. I would heavily discourage it during an interview loop because there are a lot of people internal that do not like that type of, uh, technology, especially during an interview process. Um, but I just. Sometimes I, I like seeing what, what will happen.

Um, also it's very easy to tell that someone is using these technologies is, um, the interviews are, are, um, recorded and also like over a video. So, um, I. But I kind of just wanna see what happens. And what I found often is the people who probably used like the Cha cbts to answer [00:33:00] the, the assessment online, um, they don't know what they're talking about.

And it becomes very easy to find the holes in the logic. Um, because frankly, I can reason through what I'm talking about and trying to figure out a solution faster than they can input a prompt. Into Chatt BT to get an answer back. So really the time constraint kills them every single time. Uh, they only get an hour of my time and they have to prove to me that they meet a certain level of expectation and that they can perform the job that they're being asked to perform, um, as part of the interview loop.

So it's weird in the sense that I am. Like, it seems like I'm sort of gatekeeping in some ways, but also I don't know any other way of doing it other than just to like let it play out. Um, and I, I don't know. I, I've had a few conversations with people's different viewpoints. Like I [00:34:00] said, some people are very discouraging of utilizing these technologies at all.

Um, but most of the time I think it is more of an interesting exploration into, um, this person's ability to like. Kind of eek their way through this system that was developed to try and prevent people from eeking their way through into that, uh, situation. So, uh, I guess that's kind of just my experience.

Um, like as someone who's like a bit of a guardian from a. Trying, like preventing people from not, sorry, let me rephrase that. Um, there was a lot of words there. Um, basically what I wanted to say is, um, I, I sort of act as like, uh, a guardian against, um, people who don't know what they're doing and are relying on, uh, these AI LLM models to get through things that they have no idea what they're trying to get through or what they're talking about.

They, they lack the context. Um, and that [00:35:00] context becomes very. Critical in key moments throughout that job. And if I allow them through, then that is setting them ultimately up for failure. Um, here.

Hi, I'm Alex. I'm a data scientist at an AI services company. Um, in terms of how I use ai, it's mostly for like coding help for myself. In the sense of like, I need to write like a block of code. And I've done it several times before, but I haven't memorized it. So I might get some support from like a coding assistant.

But once that's done, I will verify it. And I usually like have my own, like, I like single parentheses first and then double for my own like pernickety reasons. But I'll go in like that and I'll make sure like it's written in like my own kind of like Python accent 'cause I learned R first. Um, but all that being said, like that kind of goes to this idea of.

Data scientist taught me when I started, which was like, you own the code. [00:36:00] Like it is your work, it is your name on it. Um, and that goes back to like the AI ethics, um, and also that episode of Parks and Rec where Ron Swanson does those analog Yelp reviews where like he takes a piece of paper and like says, your hamburgers were awesome and science is a name, like your name is on it and your name should mean something.

Um, and so, you know, when I'm using an AI tool like that for support, I make sure that I understand what it's doing and that. It is something that I would do and it reflects me. Um, I know that's, you know, easy for me to do as an adult who's, you know, seen a world of coding that didn't use AI and I didn't know the same temptation when I was learning how to code.

Um, but kind of related to that, um, in a previous life I was a middle school teacher, um, and I've taught Python, um, from all different grades, like third graders up through like. College students. Um, and one of the things that I put a lot of emphasis on is problem solving and like kind of embracing the struggle.

And [00:37:00] so, and sometimes that means like giving students a problem or a task that like they wouldn't necessarily know how to do, but they struggle with it and they wrestle with it first. And then we go back as a group and work through it and say, okay, how can we solve this? And. That's not a perfect antidote to, um, you know, teaching ai.

And I think there's not one out there yet. Um, but to me, you know, my, my instinct and inclination is kind of like, let's learn Python, or let's learn coding without ai. And that's not to pretend like, Hey, AI's not here. Like nobody look at like the man behind the curtain. But try it on your own. Struggle with it.

Wrestle with it. And then when you do get help from an AI tool. Like, you're gonna be able to correct it or fix it in the way that kind of makes it more representative of your work and you can do a better job of owning the code. So, not a solution, but just set a suggestion from my own experience with, with teaching AI and how I use it.

Um, and I also, uh, work for an AI services firm. So we do a lot of AI development for clients. [00:38:00] If they don't wanna do it in house, they hire us. And so there's a lot of teaching AI there as well in terms of, you know, what is the right tool for the right job, or, you know, how does this algorithm work and. You know, making sure that clients really understand what it's gonna do for their company.

'cause if they don't have a good understanding of it, like that's not good for them. It's not good for their business. So, um, getting in tons aligned like that is, is always really helpful as well.

Uh, I'm Quincy Tenon. I, uh, teach at Fern Creek High School in Louisville, Kentucky. And we have a, a four year computer science program where python's our primary language, um. I think, Kelly, your last question is really the most important one is like, is is what we're doing? Does that matter? Um, and I think, I think it matters a great deal because, uh, I hear a lot of folks talking about what they're doing personally with [00:39:00] ai and uh, here's sort of two things.

One is that AI is useful for trivial tasks. Um, and sometimes those tasks are not, or the outcomes are not trivial, but, um, ultimately it's using a series of trivial steps to arrive at that result. And, um, the other thing that comes to mind is, uh, this idea I use all the time when I'm, I'm talking with my students, particularly when we get guest speakers in, is this, uh, and I'm gonna call 'em the spirit of Daniel Kahneman here, but it's, uh, this curse of knowledge.

And it's, um, it's our tendency to forget how hard it was to earn the ability, uh, to do the things that we do. So trivially in the, in the present. And a lot of times when I talk about coding with my sophomores, who are my, that's my introductory class. Um, I, I remind them that at some point their kindergarten teacher was kind enough to sit down with them and teach them how to draw an A, a B, [00:40:00] a C, a D, and an e.

And then over the course of several years, they finally learned to form a sentence, um, that made sense. And we, I think that we take that for granted sometimes, and that all learning requires this, um, these steps up to solving real problems. And, um.

The benefit of being a high school teacher, and I, I'm sure you, you all probably saw this too, but I think one of the coolest things about being a high school teacher is watching young people who are uncomfortable with solving non-trivial problems actually become comfortable. 'cause that's a change of character, that's not a change of knowledge, that's actually a change of very internal, powerful things.

Um, we're talking about, um, major cognitive changes happening that we can't. Put a grade on, but we, we can, through daily [00:41:00] observation that an educator can see these things and I'm sort of, I'm sure George and the other, you know, long, sort of long-term educators know about this. And so, um, you know, in my experience and to the way I put it with my students is very frank, if we're using chat GPT as a way of getting free answers, we are becoming dumber.

If we're using it as a way to interrogate ideas or get it to act like a tutor, when we have tough things we're working through, we're becoming smarter. Um, and

the other part I would say, you know, and, and finally to the point of Kelly's last question is does it matter? Everyone enjoys learning. Everyone ultimately, well I say everyone, let's say 90% of people. 90% of people love the result of having put in effort and actually retain knowledge and have [00:42:00] deep, you know, Bloom's taxonomy.

Like have this real depth of knowledge that they can create something new, um, and helping students understand that. Chat, GPT is utterly useless pretty much for that. Um, as far as antidotes, I guess for me as a computer science teacher, I'm always looking for ways that will just prompt students to engage that side of their brain, the system to right the rational thinking of like, okay, I have this intuitive response that I'd like to give up.

Or I'd like to send this off to chat GPT or, or Google it immediately, whatever that is. And sure if, if the question is I need to know how to write a line this string that's inside of an f stringing, then sure Google it instantly. Don't waste a single cognitive moment on that. But if it's anything that, that you really gotta think about, um, getting them to recognize that it's [00:43:00] probably not something we want to rely on an external source for.

If we're still learning, we're still stenciling the A's, B's, and C's. Um, I think getting them to see that's really the most important thing as a educator.

Um, I'm a senior. I'm one of Mr. Tenons students. I never use AI because I like writing code, and I think I never would've been able to learn anything without actually making things and writing stuff with it.

We applaud you. Okay. Hi. I am the opposite. I'm also a student. Well, I'm in the face of like learning how to, uh, love writing code. So I use AI sometimes, but now I have stopped using it. 'cause if I say to TPT, this is the question. I put it in, write [00:44:00] this code for me. It's gonna write it, and if I ran it, it's not a good code.

Like there are errors in there and I'm like, what am I doing? I could just go back and try to think of what I'm gonna write, write it, and then put it into a GBT. What do you think I should do to make this code better? At that point, it gives me like a good code like information and how to make it better.

And that's how I kind of use chat TPT with education. Personally, I use it for everything. Like recently, I use it to build a thing called, uh, to make me a budget list on like I have a 50, a $50 budget list for my college items. Just give me an opinion on like what to buy in order to like be on that budget list.

And that's how I use it for my personal. Then and how and why students like use AI that way? I think it's because of the foundation [00:45:00] we started with. I, we got the privilege of him teaching us how to use AI in the first place. But most people miss the fact that we don't teach students how to use AI in a way of like we being the tutor and telling it, this is what I wanna understand.

Don't give me the answer straightforward. And that's like the misunderstanding about ai. We don't, we just have a misunderstanding that it makes us more dumb, which it does at some point. If you just keep consistently going back, Hey, tell me what this answer is, and I go to an exam room and there's no answer, just makes me more dumb.

But if I learn how to like interact with it and have a foundation of like, Hey. Um, I have this question and I don't understand. I need you to teach me step by step, ask me questions, be interactive with me, how to make this problem easier for me to understand and learn, and it is gonna do that for you.

I've tried that and it does that for me many times, so I can't, I do benefit from [00:46:00] ai, but I'm trying to stop using it for write to write code so that I can fall in love with writing code, and that's where I stand right now. Thank you.

Kelly: Well, uh, Mr. Tenon has a, a very special group of students, so I have to, we have to give them a clap for that. I'm sorry. I, I, I love all your answers. Um, if, if I can have my time to speak, I, I have so many notes and so many things to say. 'cause this is what I love about listening to people. Um, and where do I start?

I guess for me it's the level of the learner, right? Um. I teach sixth graders. So I teach 10 year olds who most of them come in using, um, Chromebooks and don't even know how to turn on their MacBooks or their, or their, uh, PC Dells or whatnot. And so that's a huge learning curve and I, I, I mirror what Quincy has said about the ABCs, right?

If they don't have that [00:47:00] foundational knowledge, then going in there and trying to learn how to code when you can't even write A, B and C, it's kind of silly to allow them to use, um. Any type of generative ai, so we don't use that. Um, the other thing that I heard was, you know, the trivial, trivial tasks we use, and I wrote this straight from like Quincy, if we use the AI for trivial tasks, why wouldn't we allow the kids to use them for trivial tasks as well?

So if we are assigning trivial coding tasks like still, um. The dumb ones like, oh, let's do, uh. I don't know, what are the ones that they do? All the short, this list of five numbers or something. Why do people do that? Right. I, I mean, I learned how to, I learned how to co code in an older age, right? Um, my brain hurt so bad.

I felt that cognitive switch, and that's what I'm trying to get from my students, is that ability to like think [00:48:00] differently. Um, seven years ago, I, you know, I started learning code and the way that I've changed my thinking is that problem solving, so. A changing that assessment, like you were saying is always, is always a, um, hard thing for teachers, but it it, we have to make sure they get that struggle right?

They have to have that cognitive, um, pain in order to do the learning. Right. We can't just go get on a bicycle like I was trying to say yesterday. We have to ride our, our tricycle, we have to have our training wheels. Um, we can't go get on an e-bike and hope that we're not gonna get killed. Um, so that's something that I'm always looking for is how do I, how do I switch that task to make the kids struggle?

And, and sometimes it's not necessarily about the coding project, it's about maybe I'm setting them up into a group task. Right. That's the struggle I'm trying to teach them. How do you code one thing together? So, [00:49:00] um. That's it. I'm getting the three minute things. So another thing is, is real quick is I've always given them like Python, the mats, the information sitting in front of them during tests.

The Board of Knowledge has all the code up there, so why wouldn't I give them the code to make something better? I don't have the answers, but those are just some of the things I've been thinking about. 

Sean: Yeah, I was gonna, the only thing I was gonna add, and, and I think this hopefully, you know, encapsulates a lot of the things that you're all saying and, and, and.

Um, the experiences that you're having having, uh, when we interviewed Barbara Oakley, um, I think one of the things she said, or maybe I'm, I'm, I don't wanna misattribute it. One of our guests said, some, um, said something that was powerful to me. Um, and this was during the time of COVID when there was a lot of disruption in change in education.

And they said, people are always learning. It is in our nature. We are always learning something constantly. 

Kelly: Will Richardson, 

Sean: will Richardson said that? Yes, will [00:50:00] Richardson. Thank you. Barbara Oakley said a lot of smart things too, but this one was Will Richardson and he said Kids are Stu. People are always learning things.

The really interesting question is what are they learning at any given time, right? So when they are using chat GPT to get answers or, you know, regardless of the model, if they're using ai. To avoid learning. What are they learning? They're learning better ways to avoid learning. Mm-hmm. Right? They're always learning something.

And so I think as learners and as educators, if we keep that in mind about what are they learning at any given moment, and then do our best to inspire and influence the direction that that learning takes, I think it's, um, it's a powerful role that we have, especially in this time right now. And I think, um, you know, as we are.

Uh, going forward and learning more about AI and learning what it can do for us, what it can't do, um, I think that, you know, I'm, I am definitely encouraged and I'm definitely, um, [00:51:00] grateful for all the work that you're doing to make this happen and know that, you know, we're not alone and concerns that we have and the, the struggles that we're all facing, and more importantly, the desire to get it right and to do well in this environment where.

Things are changing rapidly and we're redefining a lot of the things that have been the bedrock of our learning and teaching experience for the past a hundred plus years. So I think we can, we can probably wrap here. Yeah. And not without saying thank you everyone. 

Kelly: Oh. Can I give two seconds? Sure. Yes, Sam.

Thank you. He's moved by your, by the, 

by the, the collective. Um, I'm, I'm not an educator. I'm just a software engineer, but I'm, um. I guess at the age where I was in school, somewhat recently, but also before LLMs were a thing that everyday people used, um, which now [00:52:00] seems like a long time ago. Um, so there, I don't know.

There are a couple of things. The, the first thing is in my school, in my. Cohort of computer science students. 'cause I, I studied computer science. Um, there were a lot of folks who were not interested in understanding how to code. They, they were interested in getting from point A to point B as quickly and as, as mindlessly as possible.

I shouldn't say mindlessly, that's a little mean, but, but as quickly as possible. Um, and so that's, that's not an especially new. Phenomenon Back then it was driven, I, I feel so weird saying back then, but it was driven by, um, just the, the insane scarcity of software engineers meant that people could command really high salaries, could command a ton of perks.

That world is [00:53:00] now, um, ending it seems like. And so those incentives maybe aren't there to the same extent anymore. Maybe. Maybe that's not true. Um, and so I don't know, it'll be interesting to see, um, what that means for software engineers and if, if it does mean that kind of, not to contradict what anyone has said, but that the folks who do decide to make a profession out of this field, um, will more likely be.

Folks who have a genuine interest in it, who, who enjoy learning in that particular field and, and trying to solve those sorts of complex problems. Um, yeah, so I mean, those are my 2 cents on AI and education. In terms of how I use it personally, I am very much a Luddite, I, technology scares me despite being a software engineer and despite, you know, um, [00:54:00] knowing a lot about how.

How LLMs work, how, you know, neural nets work. I, I did a lot of study on, on neural nets before they became these gigantic, um, insanely complex things. Um,

and I think I, I absolutely agree that they're useful for like, routine mundane tasks. Like some, somebody mentioned, um, like assessments, forms, um. I think writing unit tests, drafting unit tests, maybe I would use them for that. But my, one of my struggles is that whenever I run into a wall with a problem, it is a problem that requires a lot of context on the specific code base that I'm working in.

And so the AI agent needs to. It needs to have that context and also needs to have [00:55:00] context on the problem domain. There was, there was a talk earlier about domain driven design, um, and about the, the value in understanding and in, in clarifying in your code, the actual problem domain that the application is designed to address.

Like for me it's, it's astronomy and, and astronomical observations and, and you know, scans and frequencies and that sort of thing. Um, and. Eh, maybe this is a feature of my poorly commented, poorly documented code, but the AI can kind of seem like it struggles to help me. Um, like a human would help me a lot more because they, they have all of that context.

Um, so I, I don't know. I like, I know some of that reluctance is probably, um, to use it is probably misplaced, but, um, yeah, some of it is that for things that I. Actually want help with. It's just not, hasn't been that helpful.[00:56:00] 

Kelly: Very insightful and very impressive with lights and, and you're gonna have Sean following you because he loves everything astronomy. But, um, we, I, I wanna say like this, that whole buzzword with AI literacy, I feel like this collective was a very literate, literacy driven conversation. So I think everyone that was here.

Yeah. 

Sean: Alright, so we're gonna thank everyone for participating and contributing your thoughts and, um, it's just such a great experience to be able to do this live. Most of the time it's Kelly and I on a Zoom call recording this together. So it's really wonderful to be in the room with you. Um, so thank you for joining us for this and, and we'll be sharing this with, uh, through our normal podcast channels.

Um, I think, you know, and also Kelly, I just want to make the remark. We had a, a person who joined us to be able to listen and because of their job was not able to participate because they're not allowed to speak to the media. And I'd like to say this is the first time we've ever been considered the media.

No, that is true. Which is kind of cool. True, true. [00:57:00] Wow. That's cool. Yeah, I like that. Uh, so I've have a new title, note that, and that note from the media conglomerate of Teaching Python. So for Teaching Python, this is Sean and 

Kelly: this is Kelly signing off.