How New Computer Science Standards Prepare Students For An AI World

WEBVTT
00:00:05.280 --> 00:00:07.839
Welcome back to Change Ed.
00:00:08.160 --> 00:00:14.240
Changed Changes the number one rated education podcast in all of Pennsylvania.
00:00:14.640 --> 00:00:22.320
I am one of your beloved hosts, Andrew Kuhn, Education Consultant from Montgomery County Intermediate Unit.
00:00:22.559 --> 00:00:28.239
And here with me is Patrice Matek, also out of the Montgomery County Intermediate Unit.
00:00:28.559 --> 00:00:30.320
And everyone's favorite SDF.
00:00:30.719 --> 00:00:33.039
Tony Marabito, Carbon Lehigh Intermediate Unit.
00:00:33.359 --> 00:00:39.840
We have a guest on the show today who probably has a bunch of different titles that we don't have either.
00:00:40.479 --> 00:00:40.880
Yes.
00:00:41.280 --> 00:00:46.880
But also this person competed in a fantasy football league that we had.
00:00:46.960 --> 00:00:51.520
And there were four of us on here, and only three of us were in the top four.
00:00:51.600 --> 00:00:53.200
Isn't that crazy, Tony?
00:00:53.520 --> 00:00:55.280
And one of us had two teams.
00:00:55.600 --> 00:00:57.039
And was the commissioner.
00:00:57.280 --> 00:00:57.679
Yeah.
00:00:57.840 --> 00:00:59.359
And took home two trophies.
00:00:59.920 --> 00:01:01.520
Imagine and took home two trophies.
00:01:01.920 --> 00:01:03.439
Interesting how that happened.
00:01:03.520 --> 00:01:05.680
I wonder who that human could have been.
00:01:06.000 --> 00:01:08.560
That story always comes back to being about Andrew.
00:01:10.239 --> 00:01:15.359
This is why we said when we started that you could might want to talk about something, but it won't happen.
00:01:15.599 --> 00:01:18.879
We'd like to welcome Jigger Patel to the show.
00:01:19.040 --> 00:01:19.760
Welcome, sir.
00:01:20.480 --> 00:01:21.760
Thank you for having me.
00:01:22.159 --> 00:01:26.000
Glad to join the elite of the fantasy footballers.
00:01:26.239 --> 00:01:27.519
Yes, yes, yes.
00:01:27.760 --> 00:01:28.799
That would be three of us.
00:01:28.959 --> 00:01:30.079
Yes, we are so glad to have you.
00:01:30.239 --> 00:01:32.239
Tony, what was your position in the fantasy footballers?
00:01:32.400 --> 00:01:33.519
Jigger, tell us where you're from.
00:01:33.840 --> 00:01:35.840
No, that was completely by fluke, though.
00:01:35.920 --> 00:01:36.799
Let's be very real.
00:01:36.959 --> 00:01:40.799
My second place was only because someone someone lost that didn't need to live.
00:01:41.200 --> 00:01:45.920
Anytime you have a league where there's two teams and the same manager for both those teams, it doesn't work well.
00:01:46.079 --> 00:01:47.040
Thank you, Jigger.
00:01:47.359 --> 00:01:49.040
Especially when they're the one in charge.
00:01:49.280 --> 00:01:52.959
Our guest today, I would say, since starting at the IU five years ago.
00:01:53.120 --> 00:01:54.719
I don't think I've got only been five years.
00:01:54.879 --> 00:01:55.920
No, me, me, me, me, me.
00:01:56.000 --> 00:01:56.799
Oh, I was gonna say wait.
00:01:57.040 --> 00:01:57.519
Yeah, that tracks.
00:01:58.879 --> 00:01:59.519
Let me finish the slide.
00:01:59.680 --> 00:02:02.560
I've been for yes, I've been at the IU forever.
00:02:02.719 --> 00:02:05.120
My IU, Tesco Red Unit 11.
00:02:05.439 --> 00:02:10.639
This gentleman, it's it's the only person in the state where I can say his name, and everybody knows synonymously who it is.
00:02:11.039 --> 00:02:11.840
Yeah, first name only.
00:02:11.919 --> 00:02:12.560
He's like Cher.
00:02:12.800 --> 00:02:13.120
Yes.
00:02:13.280 --> 00:02:13.520
Yeah.
00:02:13.759 --> 00:02:15.520
So you're just have a unique name.
00:02:15.759 --> 00:02:16.560
Jigger title.
00:02:16.639 --> 00:02:17.199
And what do you do?
00:02:17.520 --> 00:02:19.199
No, it's because you're brilliant.
00:02:20.000 --> 00:02:22.800
I appreciate all the good things you're saying about me.
00:02:22.879 --> 00:02:25.039
I will send you your cash afterwards.
00:02:25.280 --> 00:02:25.520
Thank you.
00:02:25.759 --> 00:02:28.319
Yeah, as I was introduced, my name's Sudar Patel.
00:02:28.400 --> 00:02:35.120
I'm from near Hershey area, but I work for Tuscora Enemy Unit 11, uh, which is near the Racetown Lake area.
00:02:35.280 --> 00:02:42.639
My official title is Director of Innovation and Special Projects, but that just means I do a bunch of stuff or whatever is thrown at me.
00:02:42.800 --> 00:02:44.400
And so I like to say I do stuff.
00:02:44.639 --> 00:02:46.639
That's a way better way of saying it we do.
00:02:46.879 --> 00:02:48.639
Does stuff well because anytime I'm in a jam.
00:02:48.879 --> 00:02:50.000
And super cool stuff, too.
00:02:50.240 --> 00:02:51.280
Yeah, I call Jigger for sure.
00:02:51.520 --> 00:02:56.000
So, Jigger, what do you do over there at Tuscarora IU as the director of innovation?
00:02:56.319 --> 00:03:02.080
So, in the doing of stuff at RIU, we have a distributor leadership model with our technology department.
00:03:02.159 --> 00:03:04.240
It's divided up into sort of three parts.
00:03:04.400 --> 00:03:09.919
One is infrastructure, one is ed tech, you know, that's technology integration, so on.
00:03:10.080 --> 00:03:15.439
And then the third is where I come in, where I sort of sit between the technical and the non-technical.
00:03:15.599 --> 00:03:18.159
We have over over a hundred different programs that we run.
00:03:18.240 --> 00:03:22.000
And then we have a few that are fairly large programs statewide that we run.
00:03:22.159 --> 00:03:31.120
And so part of my job is to support statewide programs that we run with anything technology related, but also it goes beyond technology.
00:03:31.280 --> 00:03:33.439
It goes into the instructional side of things.
00:03:33.599 --> 00:03:38.879
They run a lot of courses online, but they're all sort of people in the field with early intervention.
00:03:39.360 --> 00:03:49.599
And so we run a couple large instances of Moodle that has over, you know, close to 50,000 users and about you know close to 100 courses that run on them.
00:03:49.759 --> 00:03:50.080
Wow.
00:03:50.319 --> 00:03:57.199
So I help sort of support some of some of the instructional side of those things as well as some of the technical pieces of running those sites.
00:03:57.280 --> 00:04:01.199
And then the other piece is STEM and CS within RIU.
00:04:01.280 --> 00:04:09.520
So whenever our districts have any need with STEM or computer science, mostly with computer science, at you know, that's where sort of I come in and provide support.
00:04:09.840 --> 00:04:15.280
To support what you were saying, Tony, Jigger is so popular that I've been trying to get him on the show for over a year.
00:04:15.439 --> 00:04:17.920
I've heard his agents stopped returning my phone calls.
00:04:18.160 --> 00:04:20.480
I would stop returning my calls for me, too.
00:04:21.360 --> 00:04:22.240
Well, you have.
00:04:22.319 --> 00:04:24.240
That's why he don't even text me back.
00:04:24.319 --> 00:04:27.920
But it's that it's that trade that you declined that I tried to suggest.
00:04:34.879 --> 00:04:36.319
You're doing a lot of different things.
00:04:36.560 --> 00:04:46.399
And I'm I'm interested in in this because of of like all of our AI conversations, and and we we as an IU and all of the IUs are doing work in computer science.
00:04:46.560 --> 00:04:49.120
Because I know you're you're embedded in computer science.
00:04:49.199 --> 00:04:56.480
And I guess the question is what does computer science look like now that we have this large language model of AI that exists?
00:04:56.800 --> 00:05:00.480
And you know, have you been part of, I'm sure you have conversations of relevancy.
00:05:00.639 --> 00:05:03.680
Is it something that we are still teaching and doing and talking about?
00:05:03.839 --> 00:05:07.439
I mean, we are, but you know, these are things that we're hearing a lot with computer science.
00:05:07.759 --> 00:05:08.959
I'm interested in your perspective.
00:05:10.480 --> 00:05:16.000
With AI and computer science, I think the focus has been shifted.
00:05:16.240 --> 00:05:18.399
I would probably break it down into two separate areas.
00:05:18.560 --> 00:05:22.079
One is really understanding what AI is and how it actually works.
00:05:22.319 --> 00:05:22.480
Right.
00:05:22.639 --> 00:05:27.040
And then the other piece is has AI sort of becomes better and better.
00:05:27.199 --> 00:05:33.839
What's the role of computer scientists, specifically those that are doing the programming piece?
00:05:33.920 --> 00:05:34.240
Yeah.
00:05:34.399 --> 00:05:36.240
And software development piece, right?
00:05:36.399 --> 00:05:38.319
Because computer science is not just that piece.
00:05:38.399 --> 00:05:39.759
There's other parts of computer science.
00:05:40.079 --> 00:05:44.639
But that's a large portion of what we've been focusing on in education, is the coding part.
00:05:45.040 --> 00:05:47.839
We start teaching elementary school students how to block code, right?
00:05:47.920 --> 00:05:50.800
Does that is that something that will still be useful and relevant?
00:05:51.040 --> 00:05:52.480
Yeah, I think I think they will be.
00:05:52.639 --> 00:05:58.959
You know, part of what I've been working on or part of a larger group is with Computer Science Teachers Association.
00:05:59.040 --> 00:06:01.279
They are revising their standards.
00:06:01.680 --> 00:06:05.439
And so the new standards are going to come out in July of 2026.
00:06:05.680 --> 00:06:14.240
And I'm on the writing team with, I think there's 30, 35 of us on the writing team with people who are brilliant across the across the nation.
00:06:14.720 --> 00:06:20.240
We were tasked with the idea that we want the standards to be relevant for the next 10 years.
00:06:20.879 --> 00:06:21.360
Right.
00:06:21.600 --> 00:06:27.360
And so if you think about like what how fast AI has evolved, that's hard to predict because it looks different tomorrow.
00:06:27.920 --> 00:06:28.319
Exactly.
00:06:28.480 --> 00:06:32.160
So how do you write standards that cover the next 10 years?
00:06:32.399 --> 00:06:32.639
Right.
00:06:32.800 --> 00:06:42.560
So one of the things that you'll notice with the new standards, if you're familiar with the computer science standards right now that CSTA has, these standards are endorsed by pretty much, you know, quite a few different states.
00:06:42.720 --> 00:06:44.560
Pennsylvania has them endorsed.
00:06:44.720 --> 00:06:50.879
And so if you compare them to the new standards, there's going to be a lot more depth in the new standards compared to now.
00:06:51.040 --> 00:06:59.920
There's also going to be, you'll notice a shift that things that are covered in middle school in the old standards, snippets of it appear in the elementary.
00:07:00.240 --> 00:07:03.199
And then same with high school, that's sort of shifted over to middle school.
00:07:03.439 --> 00:07:05.920
That makes sense because the kids are more digital natives now.
00:07:06.160 --> 00:07:06.639
Exactly.
00:07:06.720 --> 00:07:06.959
Yeah.
00:07:07.120 --> 00:07:09.120
And then there's also different topics.
00:07:09.279 --> 00:07:13.920
The focus is more on things like machine learning and artificial intelligence.
00:07:14.079 --> 00:07:16.000
You know, podram is a bit piece of it.
00:07:16.160 --> 00:07:20.000
Data science, data handling, data analysis is a bit piece of it.
00:07:20.240 --> 00:07:23.439
Systems and security is also a piece of it that we don't think about, right?
00:07:23.600 --> 00:07:26.000
As AI gets better and better, there are more bots.
00:07:26.399 --> 00:07:32.079
There are autonomous bots on the web that are always targeting websites or just targeting people, right?
00:07:32.319 --> 00:07:33.519
For scams and so on.
00:07:33.680 --> 00:07:35.680
So network security is important.
00:07:35.839 --> 00:07:39.199
And then there's this new topic area called computing and society.
00:07:39.360 --> 00:07:49.120
As AI, machine learning, and the use of technology gets more and more embedded into what we do every day, especially with kids growing up only knowing AI, right?
00:07:49.360 --> 00:07:53.199
If you think about kindergartners in three years, they've always interacted with AI.
00:07:53.279 --> 00:07:55.759
They don't know what a world looks like before AI.
00:07:55.839 --> 00:07:57.360
And so what are the impacts of that?
00:07:57.519 --> 00:07:59.839
What is the environmental impacts of using AI?
00:08:00.000 --> 00:08:02.639
What are the what are the ethical impacts of using AI?
00:08:02.800 --> 00:08:05.199
What are some of the policy impacts of AI, right?
00:08:05.360 --> 00:08:12.000
So the standards are really meant to address how do you think about technology in general?
00:08:12.160 --> 00:08:23.519
We're still keeping that computational thinking piece still relevant and tying in some of those things to understand how technology works, but also going just beyond how it works.
00:08:23.759 --> 00:08:28.079
Jigger, if this doesn't apply, please just tell me because you're much more intelligent than I am.
00:08:28.160 --> 00:08:37.039
But if I had to sum up the teacher shift when it came to science, it was a lot less of lecturing and a lot more of student doing and then a facilitation by the teacher.
00:08:37.200 --> 00:08:41.600
Do you see a shift in the same way that CS will be taught with these new standards?
00:08:41.759 --> 00:08:44.960
Or like what's your take on that aspect and will it change a lot for teachers?
00:08:45.200 --> 00:08:48.000
I think it'll be a similar shift as you saw with Steels.
00:08:48.240 --> 00:08:50.080
I think with the idea of phenomena, right?
00:08:50.159 --> 00:08:52.320
Like you're sort of discovering while you're learning.
00:08:52.399 --> 00:08:54.720
I think that will the same thing will happen with computer science.
00:08:54.879 --> 00:08:58.399
What will be a little bit different, the standards they're gonna seem more rigorous.
00:08:58.559 --> 00:09:03.279
And so naturally, what's gonna happen is how do we prepare teachers?
00:09:03.360 --> 00:09:08.639
Yeah, yeah, especially if especially if you think about like who teaches computer science in elementary, right?
00:09:08.799 --> 00:09:12.080
Some of those teachers are just to be librarians, librarians, yeah.
00:09:12.480 --> 00:09:14.399
They're thrown into it, they have no background, right?
00:09:14.639 --> 00:09:15.120
Yep, yep, yep.
00:09:15.440 --> 00:09:23.039
And some of the concepts that are in the new standards are going to be things that those teachers aren't necessarily familiar with.
00:09:23.440 --> 00:09:32.320
And so it's gonna naturally shift to be more student focused where you're assigning students to to do things and you know learn as they go along.
00:09:32.960 --> 00:09:37.919
And so you're naturally gonna see less lecture from teacher, I think, at those levels.
00:09:38.159 --> 00:09:39.519
There's gonna be more hands-on activity.
00:09:39.600 --> 00:09:42.480
I think there's gonna be more opportunities for like unplugged activities.
00:09:42.639 --> 00:09:45.279
Again, the focus is in a shift more towards computational thinking.
00:09:45.360 --> 00:09:49.200
And once kids understand how logic works, how abstraction works, right?
00:09:49.279 --> 00:09:52.960
How algorithms work, the rest of the things are just fitting the pieces together.
00:09:53.279 --> 00:09:55.039
Or going deeper with it, kind of what you're saying.
00:09:55.200 --> 00:09:58.080
Once you understand those major concepts, it's going deeper.
00:09:58.399 --> 00:10:02.559
So there's a certificate, a computer science certification that's newer in Pennsylvania.
00:10:02.960 --> 00:10:21.519
From your perspective, are these new standards kind of aligning with that to where when we're getting to that level, it it's necessary for you to have that certification, or is it kind of opening up the door for it to be integrated into whatever subject so that you know a a math teacher, science teacher, or ELA would be able to tie that in as well?
00:10:21.840 --> 00:10:22.720
All of the above.
00:10:22.799 --> 00:10:30.480
And I say that because one of my favorite parts about the new standards is providing a foundation for all students.
00:10:30.639 --> 00:10:36.240
So the focus is each of the different topic areas are supposed to be foundational.
00:10:36.320 --> 00:10:40.320
So the idea is that's what everyone needs to have a driver's license, right?
00:10:40.480 --> 00:10:45.279
You want to drive a car, you need a driver's license, you need to know minimum, right?
00:10:45.440 --> 00:10:50.080
That this is what your knowledge needs to be in order for you to pass the driver's license exam.
00:10:50.240 --> 00:10:54.639
That's what we want students to sort of have, that foundational knowledge with computer science.
00:10:54.799 --> 00:10:58.879
And then what the new standards have is they have two things.
00:10:58.960 --> 00:11:06.000
They have one, specialties, which are focused around going beyond the foundational once they get to high school.
00:11:06.240 --> 00:11:13.519
Those specialty topics are there are things like software, like going deeper into software development or cybersecurity.
00:11:13.840 --> 00:11:17.039
They're going deeper into like game, game design, right?
00:11:17.200 --> 00:11:19.519
Game development or AI.
00:11:19.679 --> 00:11:19.840
Yeah.
00:11:20.000 --> 00:11:23.039
You know, and so they become sort of a career pathways for them.
00:11:23.279 --> 00:11:28.639
In that context, those standards are written in such a way that you might need more than just your praxis.
00:11:28.879 --> 00:11:34.879
You might need to take additional professional development to be able to support teaching those types of courses.
00:11:35.200 --> 00:11:44.559
And then there's a new topic, full category called Explos CS, and which is meant to be integration points into other top other subject areas.
00:11:44.720 --> 00:11:44.960
Oh, cool.
00:11:45.120 --> 00:12:07.759
So things like history, art, music, you know, math, science, and explus are written in such a way that it's bringing in some of those computer science concepts into uh other content areas and letting the teacher explore that content area with technology deeper than they were able to before at school.
00:12:08.000 --> 00:12:08.399
That's real nice.
00:12:08.639 --> 00:12:17.519
And I say the deeper because now they actually have something that they can, I hate to say like assess, but actually they can see whether their students are able to understand their integrations or not.
00:12:17.679 --> 00:12:23.679
So, like if a biology teacher wanted to do simulations with how bacteria multiply, right?
00:12:23.840 --> 00:12:34.240
And in order to do the simulation, maybe the students need to come up with an algorithm to figure out what's one algorithm that the bacteria follow in the way that they multiply, right?
00:12:34.399 --> 00:12:37.519
And so right now, like there isn't really a standard that assesses that.
00:12:37.840 --> 00:12:42.000
Like, how do you make sure that you are covering the TED piece but also the biology piece?
00:12:42.240 --> 00:12:42.639
Right, right.
00:12:42.960 --> 00:12:51.039
And so these standards are sort of written in the way that it helps teachers say, oh, like I can kill two birds with one stone and cover both standards.
00:12:51.360 --> 00:12:55.679
Do they kind of interact or connect a little bit with the T and E standards?
00:12:55.840 --> 00:12:57.840
I know they're a part of the places they do.
00:12:58.000 --> 00:12:58.320
Yeah, yeah.
00:12:58.399 --> 00:12:58.559
Okay.
00:12:58.799 --> 00:12:59.519
In places they do.
00:12:59.759 --> 00:13:05.679
Because sometimes I feel like teachers are like, oh, I've got one more thing I've got to do, especially at like elementary levels where we're trying to integrate everything.
00:13:05.759 --> 00:13:07.279
So that's good that they intersect.
00:13:07.600 --> 00:13:07.679
Yeah.
00:13:08.159 --> 00:13:15.519
Like I know the tech and engineering standards are not just computer science because it's like a small subsection of tech and engineering, but at least being able to.
00:13:15.759 --> 00:13:17.440
I liked your example of the high school.
00:13:17.759 --> 00:13:25.919
That would be interesting to see how science teacher and comp sci teacher work together and come up with a way it might actually make some like bridges happen.
00:13:26.000 --> 00:13:26.480
That's pretty cool.
00:13:26.960 --> 00:13:30.320
This is from the current draft that's posted on the CSTA website.
00:13:30.559 --> 00:13:41.360
One of the X plus CS standard is assess how well an algorithm solves problems within a non-CS discipline by analyzing their accuracy, efficiency, and relevancy to the intended goal.
00:13:41.600 --> 00:13:42.080
Oh, nice.
00:13:42.320 --> 00:13:42.639
That's cool.
00:13:43.279 --> 00:13:43.919
That's really good.
00:13:44.480 --> 00:13:45.600
That can be used anywhere.
00:13:46.159 --> 00:13:46.639
Exactly.
00:13:46.799 --> 00:13:47.039
Right.
00:13:47.279 --> 00:13:53.679
But now that gives an an opportunity for computer science teacher to reach out to a biology or math teacher and vice versa.
00:13:53.840 --> 00:13:53.919
Right.
00:13:54.159 --> 00:13:59.039
Like any survey that's done in any of those courses, and then the AI, you know, looks at the information and yeah, yeah.
00:13:59.120 --> 00:14:00.080
That's pretty that's really cool.
00:14:00.240 --> 00:14:03.039
So since since we brought up the topic of AI, now I'm gonna I'm gonna take that route.
00:14:03.360 --> 00:14:04.080
Yeah, if you don't mind.
00:14:04.240 --> 00:14:12.559
So just when I think like I'm on the like on the edge of like being caught up and like really like showing my teachers like great things, I I go to a meeting and jiggers there and has 9,000 views.
00:14:12.960 --> 00:14:15.840
And just absolutely crushes my soul because of how much he knows.
00:14:16.080 --> 00:14:17.919
So I'm sure that you're working on something.
00:14:18.240 --> 00:14:20.799
You're giving me way more credit than I than I deserve.
00:14:22.080 --> 00:14:25.759
So I'm sure you're working with some kind of project, some awesome thing that I haven't even thought of yet.
00:14:25.840 --> 00:14:27.120
So what are you doing with AI?
00:14:27.279 --> 00:14:29.279
I can't wait to hear this because I know it's gonna blow my mind.
00:14:29.440 --> 00:14:30.879
There's tons of tools out there, right?
00:14:30.960 --> 00:14:50.000
My two big things that I like to do these days now, or nowadays I should say, is you know, for people for how much people hate vibe coding, like I'm I'm still a fan, not that there's a lot of hate for it, but you know, people think that it diminishes the idea of knowing computer science and learning about computer science and programming and things, but I think it actually makes you better.
00:14:50.080 --> 00:14:54.240
You focus more on debugging because the AI doesn't always get it right, and now you have to know.
00:14:54.399 --> 00:14:56.159
So you still have to know, yeah, sure, right?
00:14:56.240 --> 00:15:01.039
You still have to understand how computer programs work in order for you to sort of vibe code.
00:15:01.120 --> 00:15:02.159
Yeah, that's one piece.
00:15:02.240 --> 00:15:06.960
And the other piece, I'm more interested because I serve a lot of rural parts of Pennsylvania.
00:15:07.120 --> 00:15:10.399
We have nine districts in our region, they're mostly rural districts.
00:15:10.480 --> 00:15:14.799
You know, our smallest district is 350 some kids, it's one building.
00:15:14.879 --> 00:15:22.720
And you know, it's also the district where a lot of people still don't have internet access, which is hard to talk about computer science without internet access.
00:15:22.799 --> 00:15:23.120
Yeah.
00:15:23.360 --> 00:15:23.600
Right.
00:15:23.759 --> 00:15:26.159
And during COVID, that was it, it was still the area.
00:15:26.240 --> 00:15:31.600
There were hotspots available after a while, but once the funding ran out, the infrastructure still isn't built out.
00:15:31.840 --> 00:15:33.360
And some of them still don't have internet.
00:15:33.440 --> 00:15:37.279
So they go back to having all these devices like Chromebooks that they can go home.
00:15:37.360 --> 00:15:40.559
They're very much internet dependent, but can't do anything when they're at home.
00:15:40.720 --> 00:15:42.879
And with computer science, it's even worse, right?
00:15:42.960 --> 00:15:55.200
Because with computer science, your 11th grade student, the stool is able to offer computer science and they have a project for you, and you have to write this code in Python, let's say, and you're home, you're working on it, you get stuck.
00:15:55.360 --> 00:15:56.080
What do you do?
00:15:56.240 --> 00:16:00.720
Well, you can't go on YouTube and look at anything, or you can't doodle anything because you don't have internet access.
00:16:00.879 --> 00:16:00.960
Right.
00:16:01.360 --> 00:16:01.440
Right.
00:16:01.600 --> 00:16:05.840
And even if you did, like you don't have access to YouTube because it's not fast enough.
00:16:06.000 --> 00:16:08.480
You can get to you know, web pages and so on.
00:16:08.559 --> 00:16:18.159
But so it becomes uh a a question of accessibility and affordability and equity when it comes to teaching computer science in rural areas where there's no internet access.
00:16:18.240 --> 00:16:26.639
So one of the things that I've been playing around with is I do a lot with Raspberry Pi, which is a small$35 computer, do a lot with physical computing with it.
00:16:26.720 --> 00:16:36.320
And one of the things that I've been exploring is like, can you run a low-level larger language model that's specifically focusing on teaching Python?
00:16:36.559 --> 00:16:58.639
And can you run it so it acts as a tutor for that student who doesn't have access, who then can not only develop on the Pi, because Pi is a computer, so they can go and build their Python program that they want, but they get a little tutor on the side that if they get stuck or need help debugging, they can ask the AI to help them figure it out or watch them through how figuring it out.
00:16:58.879 --> 00:17:02.080
I think that would be so powerful, not just for CS, but for anything.
00:17:02.159 --> 00:17:11.119
Like we're constantly creating Google Gems with teachers or they're using school AI and they're putting these tutors up there so that kids have help at home if they don't have a help from a parent or or what have you.
00:17:11.279 --> 00:17:12.720
They did whatever whatever the case may be.
00:17:12.799 --> 00:17:14.079
I knew you were gonna say something awesome.
00:17:14.319 --> 00:17:17.920
We can able to provide something that's not necessarily Wi-Fi isn't needed, right?
00:17:18.000 --> 00:17:19.359
That I think that would that'd be awesome.
00:17:19.519 --> 00:17:22.880
The other big thing is I I talked about like it being an affordability issue.
00:17:22.960 --> 00:17:27.119
Yeah, the big problem with AI right now is it costs money.
00:17:27.279 --> 00:17:34.000
We're fortunate, and I say that we're fortunate as educators, we have access to it for free for a limited amount of time, right?
00:17:34.160 --> 00:17:42.319
But if you think about parents who don't work in education, for their kids to get access to an AI tool that keeps their data private, right?
00:17:42.400 --> 00:17:43.119
That's the key word.
00:17:43.680 --> 00:17:46.720
They can use the free version, but what happens with the free version?
00:17:46.880 --> 00:17:48.000
They're giving up their data.
00:17:48.160 --> 00:17:48.319
Right.
00:17:48.559 --> 00:17:52.319
And so for families to be able to afford AI, it's impossible.
00:17:52.480 --> 00:17:53.039
Yeah, right.
00:17:53.200 --> 00:17:57.599
Because Chat GPD costs 20 bucks, Claude costs 20 bucks, Gemini.
00:17:58.079 --> 00:17:58.960
Claude's kind of worth it.
00:17:59.039 --> 00:17:59.839
I love him.
00:18:00.079 --> 00:18:00.960
Oh, yeah, me too.
00:18:01.039 --> 00:18:02.079
That's what that's what I use.
00:18:02.480 --> 00:18:08.640
I shouldn't make it a human, I shouldn't make it a human, but I love I don't use Claude and now I feel like I'm behind because Jigger uses Claude.
00:18:09.599 --> 00:18:14.079
I feel so good about myself because I have been on the Claude train forever.
00:18:14.480 --> 00:18:15.200
So I feel good.
00:18:15.279 --> 00:18:16.960
If that's the one Jigger's using, I feel good.
00:18:17.119 --> 00:18:18.640
Claude is so much more efficient.
00:18:19.039 --> 00:18:21.119
Uh yeah, I call it my assistant.
00:18:21.279 --> 00:18:21.759
Oh my gosh.
00:18:21.839 --> 00:18:27.599
So like I'll put things into Claude and then I'll put it into Copilot and I'll be like, make the things that he like he said to make.
00:18:27.759 --> 00:18:28.799
Like, just go do it, please.
00:18:28.960 --> 00:18:34.559
Although I just learned the other day, Claude is now able to make PowerPoints, and I almost died.
00:18:34.720 --> 00:18:35.599
I was very excited.
00:18:35.839 --> 00:18:39.599
Oh, I'll have to show you some Claude generated Canva deck.
00:18:41.599 --> 00:18:43.279
So what?
00:18:43.599 --> 00:18:44.240
What Tony?
00:18:44.480 --> 00:18:48.640
Jigger basically has his own Jarvis from Iron Man, and I'm still back here like using it.
00:18:48.799 --> 00:18:55.519
I need to I need to learn from Jigger how to make Claude in my Jarvis because I use it for everything.
00:18:55.599 --> 00:18:56.559
I know it's so wasteful.
00:18:56.640 --> 00:19:00.079
It's so wasteful, and I need to remind myself like not to use it for everything.
00:19:00.160 --> 00:19:02.079
But man, it's such a good thought partner.
00:19:02.240 --> 00:19:03.920
It is, it's just it's so much better.
00:19:04.240 --> 00:19:04.640
Much better.
00:19:04.799 --> 00:19:05.680
Anyway, sidetracked.
00:19:06.000 --> 00:19:07.839
No, we can talk about Claude all the time.
00:19:07.920 --> 00:19:08.799
Look at Tony's face.
00:19:09.039 --> 00:19:09.759
Um but it does cloud.
00:19:09.839 --> 00:19:10.000
It does.
00:19:11.599 --> 00:19:11.920
I love it.
00:19:12.000 --> 00:19:12.240
I love it.
00:19:12.400 --> 00:19:13.519
I call it my boyfriend.
00:19:13.680 --> 00:19:15.119
It's worth the$20 if you have it.
00:19:15.359 --> 00:19:15.759
It is worth it.
00:19:15.839 --> 00:19:16.400
It is worth the twenty dollars.
00:19:16.559 --> 00:19:19.759
I will not feed my children for a week if I need to.
00:19:21.359 --> 00:19:23.359
But it becomes an affordability issue.
00:19:23.599 --> 00:19:24.319
It does, it really does.
00:19:24.720 --> 00:19:27.920
Not everyone can afford to spend 20 bucks a month.
00:19:28.400 --> 00:19:28.640
Right.
00:19:28.720 --> 00:19:31.359
Like for some families, like it's a big deal.
00:19:31.599 --> 00:19:35.680
And you know, if you get Germany, that's another I have you know 20, 30 bucks.
00:19:35.759 --> 00:19:39.599
So you're looking at at least 60, 70 dollars just on AI tools.
00:19:39.839 --> 00:19:45.920
And if you want kids to sort of use it in a safe way, yeah, in a in a way that it keeps their information private.
00:19:46.079 --> 00:19:51.279
And so I'm more interested in the open source models, which are which are freely available.
00:19:51.519 --> 00:19:54.319
Like Meta has their llama, which is available.
00:19:54.400 --> 00:20:00.799
You know, if you have ever played with OLAMA and just downloaded that runs locally, you can download different models on it.
00:20:00.960 --> 00:20:11.359
I'm more interested in sort of how can we leverage some of those models because they're open source, they keep your data on device, so it's not going out to the web to sort of it's a little more prepared.
00:20:11.680 --> 00:20:13.920
It's not using your data to train on the cloud.
00:20:14.079 --> 00:20:18.720
Uh it's training it on the device, and there's less of a privacy issue.
00:20:18.880 --> 00:20:25.519
Now, all of these still have the same issues with the bias and so on, but at least it protects the information that you put in there.
00:20:25.599 --> 00:20:25.759
Yeah.
00:20:25.920 --> 00:20:26.079
Yeah.
00:20:26.240 --> 00:20:29.839
And so running some of these models on the Pi, it's challenging.
00:20:30.000 --> 00:20:32.960
The the other challenge is because large aundaries models are really big.
00:20:33.039 --> 00:20:33.200
Yeah.
00:20:33.440 --> 00:20:48.319
And when you start compressing them down, and this is a this this is someone else explained this to me who's actually like works in in this field of machine learning and AI, is that if you think about all of these models, they have numbers associated with them.
00:20:48.400 --> 00:20:52.480
So, like if you go to Olama, you can get a 48 billion model, whatever, right?
00:20:52.640 --> 00:20:55.680
72 billion, and it's huge in file size and so on.
00:20:55.920 --> 00:21:05.119
When you compress the the models down, the smaller you go, there's actually just a physical limit because of just how our physics works right now.
00:21:05.279 --> 00:21:06.319
Some things get lost.
00:21:06.400 --> 00:21:06.559
Yeah.
00:21:06.720 --> 00:21:06.960
Right?
00:21:07.039 --> 00:21:08.400
It's like compression of an image.
00:21:08.559 --> 00:21:08.640
Yeah.
00:21:08.799 --> 00:21:08.880
Right.
00:21:09.039 --> 00:21:13.039
If you had to take a JPEG, you compress it down so many times, what happens?
00:21:13.200 --> 00:21:14.000
It gets blurry.
00:21:14.160 --> 00:21:15.039
You lose pixels.
00:21:16.079 --> 00:21:21.039
Same thing happens with large language models when you start to make them smaller and smaller in size.
00:21:21.200 --> 00:21:23.359
You some data has to get lost.
00:21:23.519 --> 00:21:28.400
And so, how do you make sure that large language models are efficient and accurate when they're small?
00:21:28.640 --> 00:21:30.000
Well, that's what's going to be my question.
00:21:30.079 --> 00:21:33.759
I didn't think the Raspberry Pi had that much memory on it to be able to do.
00:21:34.000 --> 00:21:35.119
Maybe I'm not understanding.
00:21:35.519 --> 00:21:36.000
You can run it.
00:21:36.079 --> 00:21:36.960
It's going to be really slow.
00:21:37.039 --> 00:21:43.039
It can run limited amounts of small language models, but the Pi 5, there's an AI hat that you can get.
00:21:43.200 --> 00:21:48.079
So like it costs about a hundred bucks and you can have it so it runs pretty decent models on it.
00:21:48.240 --> 00:21:49.920
And it it's not gonna be fast, right?
00:21:50.000 --> 00:21:53.279
It's you can have to wait, but you know, patience is is part of learning.
00:21:54.160 --> 00:21:55.599
It's also better than not having anything at all.
00:21:55.680 --> 00:21:56.400
So exactly.
00:21:56.480 --> 00:21:57.680
Yeah, that's kind of where we are.
00:21:57.920 --> 00:21:59.759
The other challenge is like how do you train them to do stuff?
00:22:00.160 --> 00:22:00.720
Certain tasks.
00:22:00.880 --> 00:22:03.519
Like I don't need it to know the history of the world, right?
00:22:03.759 --> 00:22:04.079
Right, right.
00:22:04.160 --> 00:22:04.799
What can you define?
00:22:05.119 --> 00:22:06.160
I just want a Python tutor.
00:22:06.319 --> 00:22:08.880
I just want it to focus on teaching me how to learn Python.
00:22:08.960 --> 00:22:10.160
I don't care about all the other stuff.
00:22:10.400 --> 00:22:11.920
And so how do you just train it on that?
00:22:12.000 --> 00:22:13.279
And how do you make it more efficient?
00:22:13.440 --> 00:22:15.920
So that's the piece that that's the hard part about it.
00:22:16.079 --> 00:22:18.240
Because you can put a model that is a tutor.
00:22:18.400 --> 00:22:19.599
It's not going to be that great.
00:22:19.839 --> 00:22:31.759
Well, but if that's the most intended use is to be the tutor, then then if that's what you know, what that's you're getting, then that is actually not a bad price to be able to do something like that in a non-internet friendly area.
00:22:32.000 --> 00:22:32.640
It's pretty cool.
00:22:32.799 --> 00:22:35.359
It's actually really cool that you're able to do something like that.
00:22:35.839 --> 00:22:38.000
Andrew's awfully quiet this episode.
00:22:38.640 --> 00:22:40.720
He doesn't he doesn't do technology.
00:22:40.960 --> 00:22:41.119
Yes.
00:22:41.279 --> 00:22:42.720
No, I do Andrew Intelligence.
00:22:42.799 --> 00:22:43.839
It's AI, right?
00:22:45.599 --> 00:22:46.799
Been holding on to that all day.
00:22:47.200 --> 00:22:47.519
I can tell.
00:22:48.960 --> 00:22:50.079
That one's come out a million times.
00:22:50.160 --> 00:22:50.799
You haven't heard it before?
00:22:50.880 --> 00:22:51.200
No, I'm not.
00:22:51.279 --> 00:22:55.039
Well, that's obviously Tony doesn't listen to the podcast that we create.
00:22:55.519 --> 00:22:57.039
He doesn't even listen when I talk.
00:22:57.119 --> 00:22:58.400
So there's a reason.
00:22:58.799 --> 00:22:59.839
You know what I really appreciate?
00:23:00.000 --> 00:23:05.519
With someone who's already knowledgeable about the content or about the topic, it's an amazing tool.
00:23:05.680 --> 00:23:09.039
And it's it's uh a thought partner and he can push things further.
00:23:09.119 --> 00:23:14.480
But for someone who's not knowledgeable, it could actually be really dangerous because you'd be like, Well, look here, here, it's right here.
00:23:14.720 --> 00:23:21.920
Well, yeah, especially those and then they just accept it at the way that it is, not recognizing the bias that we're the other issues that are in there, not proofing it.
00:23:22.319 --> 00:23:24.640
Correct, which I think is why verifying it, yeah.
00:23:24.960 --> 00:23:28.480
This mission of access and equitability is so important.
00:23:28.640 --> 00:23:43.279
So there's lots of us versus, well, you know, we all just kind of start believing what's put in front of us when we have the ability to question, to think about it, uh, you know, even using the skills of CS that you were talking about, the computational thinking, like how is all of this being able to be big thinkers?
00:23:43.519 --> 00:23:52.160
Well, not to go deeper to that, but one of the things that I always talk about in the professional development that I do is all of these tools are great, like magic school, school AI, right?
00:23:52.240 --> 00:23:56.960
Like Gemini, Gems, whatever your preferences, all of the sort of these third-party tools are great.
00:23:57.119 --> 00:24:03.200
But we should be concerned that there's four large language models that we all are dependent on.
00:24:03.359 --> 00:24:03.599
Yep.
00:24:03.920 --> 00:24:06.240
Like OpenAI's, Chat GPT, right?
00:24:06.400 --> 00:24:09.599
Anthropic Claude, Goodle's Gemini, Copilot.
00:24:09.759 --> 00:24:11.599
There's five, and then Drock from X.
00:24:11.839 --> 00:24:16.079
But every like Magic's tool uses a combination of OpenAI and Claude, right?
00:24:16.160 --> 00:24:17.200
So does School AI.
00:24:17.359 --> 00:24:19.200
Brist also uses OpenAI.
00:24:19.599 --> 00:24:23.759
So all of the core knowledge sits between four large corporations.
00:24:24.079 --> 00:24:24.400
That's a lot of fun.
00:24:24.559 --> 00:24:28.640
And let's not discount the fact that OpenAI calls them some of a nonprofit, but they're not, right?
00:24:28.720 --> 00:24:32.960
Because if they were, they wouldn't be charging 20 bucks to keep your data private.
00:24:33.359 --> 00:24:34.240
Yep, ads and things.
00:24:34.400 --> 00:24:34.640
Yeah.
00:24:34.799 --> 00:24:34.960
Right.
00:24:35.279 --> 00:24:35.599
Jigger.
00:24:36.160 --> 00:24:36.799
Thank you for this.
00:24:37.119 --> 00:24:50.079
On behalf of the entire Change Ed team and nation, but also as a two-time award winner for the one-time season of fantasy football, I want to say thank you so much for coming on and giving us your time and your knowledge.
00:24:50.160 --> 00:24:53.680
And I look forward to the rematch next in next year's fantasy football team.
00:24:53.920 --> 00:24:56.319
Where he only has one team and he's not the commissioner.
00:24:56.559 --> 00:24:59.839
Didn't you say you could like design something that I'll have Claude design it, I guess.
00:25:00.000 --> 00:25:00.160
Yeah.
00:25:00.559 --> 00:25:02.799
We did talk about that, having our own fantasy league.
00:25:03.119 --> 00:25:03.599
I think we should.
00:25:04.160 --> 00:25:04.720
Well, look at that.
00:25:04.799 --> 00:25:05.519
We're out of time.
00:25:05.920 --> 00:25:07.519
So we're gonna shut this down.
00:25:07.920 --> 00:25:08.559
Thanks, Sugar.
00:25:08.799 --> 00:25:09.200
Thanks, Jaden.
00:25:10.160 --> 00:25:11.119
Thank you for having me.