Not That Kind of Doctor - Artificial Intelligence

TLTE Author

02/07/2024 Added

1 Plays

Description

Artificial Intelligence - Ph.D. discussions with Justin Olmanson, Nicholas Husbye, and Guy Trainin. www.youtube.com/@tltenotthatkindofdoctor and https://mediahub.unl.edu/channels/42980

Searchable Transcript

Search:

[00:00:00.629](upbeat music)
[00:00:09.900]Except I have nothing to grade until next Wednesday.
[00:00:12.570]I'm sure there's gonna be a nightmare about that
[00:00:14.010]at some point as well.
[00:00:17.910]So, new semester,
[00:00:20.310]new technology,
[00:00:23.400]and we're here on Not That Kind of Doctor today
[00:00:27.780]to talk about artificial intelligence, generative AI,
[00:00:32.376]and, particularly, to be thinking about
[00:00:34.620]some of the ways that
[00:00:36.780]we can use
[00:00:39.090]this generative AI in
[00:00:43.200]our teaching and our research.
[00:00:45.960]So, I'm not forgetting it this time.
[00:00:48.600]I'm Nick Husbye.
[00:00:49.470]I am an Associate Professor of Elementary Literacy Education
[00:00:52.230]here, at UNL.
[00:00:54.120]My co-host.
[00:00:55.020]I'm Guy Trainin.
[00:00:56.610]A professor here at UNL as well.
[00:00:58.830]And we are joined by-
[00:01:00.240]Justin Olmanson.
[00:01:02.100]Same department, same institution.
[00:01:05.550]Innovative Learning Technologies program.
[00:01:09.360]And so, we're gonna be talking about, like,
[00:01:11.820]artificial intelligence, and generative AI,
[00:01:14.640]and ways that we are both leveraging it,
[00:01:20.100]as well as being suspicious of it
[00:01:24.330]within the work that we're doing and how graduate students
[00:01:27.570]might be able to do that same thing.
[00:01:30.270]So,
[00:01:33.720]how are you feel, like, what are your,
[00:01:36.780]kind of, thoughts around generative AI?
[00:01:40.968]And how you're using it in your teaching, in your research?
[00:01:44.520]Just widespread.
[00:01:45.780]Just generally.
[00:01:47.610]Well, I think like a lot of people, I started out
[00:01:50.040]interested in how I could escape tasks I found,
[00:01:57.277]that I found unpleasant, tedious,
[00:02:00.300]or time consuming in a way
[00:02:02.040]where I'd prefer to spend my time, my work time,
[00:02:05.130]in a different way.
[00:02:06.600]And so, like most people, I think that's where I started.
[00:02:10.140]And then, because I do research in these areas,
[00:02:14.400]because I've been interested in these areas for a while,
[00:02:16.410]I started teaching a course in
[00:02:17.730]artificial intelligence in education in 2016.
[00:02:21.300]So, from then, I've been, sort of, in this area.
[00:02:25.500]And, now, there's an explosion of possibility,
[00:02:29.280]and uptake, and everyone's talking about it.
[00:02:33.030]And so, that's exciting.
[00:02:35.130]And it sort of feels like,
[00:02:38.250]I should be farther ahead, I should be farther along.
[00:02:42.360]Shouldn't we all?
[00:02:44.147]In what way?
[00:02:44.980]Tell me more.
[00:02:45.813]Like, how could you envision yourself being further along?
[00:02:49.140]Like, all the develop, like,
[00:02:51.630]I like to design.
[00:02:52.950]Right now, I've got several designs
[00:02:56.340]that leverage artificial intelligence,
[00:02:58.080]that leverage generative AI.
[00:03:00.240]And we're on a third round of data collection for
[00:03:05.220]an intelligent personal learning assistant
[00:03:08.220]for undergrads, and then one for grads.
[00:03:10.920]One that's based on them learning computational thinking,
[00:03:14.520]computational literacies, and programming.
[00:03:17.160]And then, at the undergrad side,
[00:03:18.870]how do I navigate a course that has,
[00:03:21.870]you know, 46 optional projects?
[00:03:25.920]You know, in terms, you know, if we're in a TLDR age,
[00:03:31.050]you know, how do you get students to, sort of,
[00:03:33.450]engage with 40 different optional projects?
[00:03:36.210]On one hand, they're excited 'cause they can find the things
[00:03:38.528]that they want to do, because there's something like that.
[00:03:41.033]And, on the other hand, you know,
[00:03:43.110]it just feels overwhelming.
[00:03:44.700]And so, logistically, our personal learning assistant
[00:03:48.540]helps them sort of parse that all,
[00:03:50.970]and gives them suggestions based on what their major is,
[00:03:55.020]what they're interested in.
[00:03:56.010]Are they interested in the maximum number of points
[00:03:58.320]for the least amount of work?
[00:03:59.880]Or are they interested in, you know,
[00:04:01.590]their speech path or a future speech path,
[00:04:04.140]and they want projects that
[00:04:06.660]are devoted along those lines.
[00:04:09.330]And then, we have another part of it,
[00:04:11.580]where it supports them in their microteaching.
[00:04:13.920]Where this is, for many students in my undergrad course,
[00:04:17.700]it's the first time they're thinking about teaching.
[00:04:20.580]And, a lot of times, they're like, "Well, this is it."
[00:04:22.747]"If this doesn't go well,
[00:04:24.780]I'm gonna have to change my major to Canadian Studies."
[00:04:28.470]And a wonderful PhD by the way.
[00:04:31.997]But, if that doesn't-
[00:04:33.750]So, if it works for them, then they feel great.
[00:04:36.240]And so, they put a lot of pressure on themselves.
[00:04:38.700]And, sometimes, that pressure comes at three in the morning.
[00:04:41.070]And so, having this, having
[00:04:44.460]an agent or a bot just sort of hanging out,
[00:04:46.650]and at three in the morning they can say,
[00:04:48.367]"This is what I've got,
[00:04:49.890]this is what I need to have, what am I missing?"
[00:04:52.290]Our bot, you know, has all the
[00:04:54.030]information about the project.
[00:04:55.860]And, you know, whatever Reddit has said about teaching.
[00:05:01.688]So, surprisingly useful, many students have found it
[00:05:04.320]in the research we just finished from the January term.
[00:05:07.980]And so, I wanna now take a step back.
[00:05:10.890]I know a lot of people do that,
[00:05:12.090]but I think it's important that we
[00:05:14.370]talk about AI and generative AI.
[00:05:17.640]So, let's start there just to kind of lay the ground,
[00:05:21.930]because AI has been with us for a long time, actually.
[00:05:25.590]And, what I'm often surprised with, is
[00:05:29.880]how many things in my life are AI driven
[00:05:33.060]without me having thinking about them as AI.
[00:05:36.660]For example, I've been using grammar check.
[00:05:40.410]If Grammarly wants to sponsor, we're here.
[00:05:43.200]But I've been using it for a really long time,
[00:05:46.200]and that's AI.
[00:05:47.610]And it's gotten really better over time in multiple ways.
[00:05:51.510]But there are lots of little pieces in my life
[00:05:53.670]that are not AI by definition,
[00:05:57.180]that are actually AI.
[00:05:58.800]So, we started thinking differently when
[00:06:00.840]generative AI showed up, when ChatGPT, really, shows up.
[00:06:03.630]So, can you talk a little bit about
[00:06:04.920]the difference from your perspective?
[00:06:07.080]Sure.
[00:06:09.690]Over 10 years ago,
[00:06:12.180]I went to a conference,
[00:06:13.830]an artificial intelligence and education conference,
[00:06:16.320]international conference,
[00:06:18.060]and I was amazed at what counted as AI.
[00:06:22.980]Okay, so, if you have a series of if-then statements,
[00:06:27.450]some would say, "Well, it's intelligent."
[00:06:29.287]"It's making choices."
[00:06:30.750]It meets the bar.
[00:06:32.310]What you mentioned, in terms of Grammarly,
[00:06:34.320]there's a long, rich, useful history of
[00:06:38.040]natural language processing,
[00:06:40.050]and it's coming from a linguistic side.
[00:06:42.060]That's how I started in AI,
[00:06:44.340]was thinking about things like WordNet.
[00:06:47.310]It was a computational linguistics course.
[00:06:50.070]And you could get quite a ways down the road,
[00:06:53.370]in terms of offering supports.
[00:06:56.580]I think what-
[00:06:57.540]And, for instance, OpenAI's GPT has been around,
[00:07:02.010]GPT-2, GPT-3, you know, it's been around for a while too.
[00:07:05.310]But, until November of '22,
[00:07:08.550]there was not really a sense that
[00:07:11.250]it could sort of be a sidekick.
[00:07:13.470]It could do some things, like I mentioned,
[00:07:15.150]do some things for you.
[00:07:16.650]Generate things that weren't trash
[00:07:19.920]and didn't seem random.
[00:07:22.140]That you could sort of, you know,
[00:07:23.700]without too much finessing, turn it into something
[00:07:26.940]that you could say, "Okay," you know,
[00:07:29.647]"I collaborated with this thing and it was useful."
[00:07:32.400]And I think, partially, that's because, you know,
[00:07:36.520]as PhDs, we have a reasonable amount of
[00:07:38.640]understanding about our fields.
[00:07:40.740]And we know when we're getting into areas
[00:07:42.930]where we don't know enough, right?
[00:07:44.790]And so, I think it's, we're uniquely positioned
[00:07:48.240]to sort of work with generative AI in our areas
[00:07:51.180]or adjacent to our areas, because we sort of,
[00:07:56.490]we have a sense as to how to navigate knowledge
[00:08:01.128]and sort of understand, "I don't think that's,
[00:08:03.480]I don't think that's going in the right direction."
[00:08:05.340]But I think that's one of the major differences.
[00:08:07.590]You can use AI
[00:08:09.870]to predict the next purchase someone's going to make.
[00:08:14.070]In the case of these two, probably Starbucks.
[00:08:17.820]Or you can use generative AI to, sort of, write a report
[00:08:23.850]as to, you know, based on some data,
[00:08:26.650]what type of purchase they're gonna make and why.
[00:08:29.370]And I think those,
[00:08:31.710]those might be, you know,
[00:08:33.780]in the ballpark of the differences.
[00:08:35.670]One is generating, synthesizing, and generating information.
[00:08:42.570]And the other is, sort of, just giving you information
[00:08:46.530]and you have to do things with it.
[00:08:48.930]Okay.
[00:08:50.670]So, let's shift about,
[00:08:52.650]talking about generative AI in teaching.
[00:08:54.450]You've touched a little bit about
[00:08:56.640]the use of generative AI in teaching.
[00:08:59.640]And I wanna start with the conversations
[00:09:01.860]I've had with my undergraduate students.
[00:09:03.510]And that's an important part of the job we do,
[00:09:06.480]whether it's graduate students or as faculty.
[00:09:09.030]And that is-
[00:09:11.880]I think that it's really important to give a very
[00:09:15.150]consistent message within each class.
[00:09:18.810]Because, right now, the sense I'm getting from
[00:09:23.190]our undergraduate students especially,
[00:09:24.840]but even from graduate students,
[00:09:26.340]is the boundaries are blurry.
[00:09:28.230]And that makes people very hesitant to use it effectively,
[00:09:32.160]because they're not sure what the expectations are.
[00:09:35.580]And they felt,
[00:09:37.620]a number of time, almost whiplashed.
[00:09:40.290]So, encouraged to use it in one course,
[00:09:42.960]completely banned in another course,
[00:09:45.450]and ambivalent in a third course.
[00:09:47.520]And so, that causes a student
[00:09:49.800]not wanting to get into trouble
[00:09:52.620]to say, "You know what?"
[00:09:53.797]"I'm not gonna touch this."
[00:09:55.027]"At least for a while until they can figure it out."
[00:09:58.770]Well, and I think, too, one of the things that's
[00:10:00.990]important to highlight is
[00:10:04.930]there's a variety of different entry points
[00:10:07.050]into generative AI.
[00:10:08.130]There's a lot of different ways to
[00:10:10.590]access what generative AI is able to do
[00:10:14.760]and utilize it well.
[00:10:17.070]So, when Justin's talking about his chat bot,
[00:10:19.770]that's very much from a coding perspective,
[00:10:22.380]which pulls on that AI
[00:10:25.680]in a slightly different way than
[00:10:28.920]I pull on generative AI.
[00:10:30.630]Like, utilizing generative AI as
[00:10:34.410]a thinking partner, as
[00:10:37.020]a way to help me
[00:10:40.440]identify where gaps might be
[00:10:42.990]as I'm thinking about, "This is my course."
[00:10:44.887]"This is the outline of what my course is doing."
[00:10:47.880]Help me think through, if these are my learning goals,
[00:10:50.520]what might I be missing out on
[00:10:52.290]given these characteristics of my students?
[00:10:55.710]And it's super good at that, right?
[00:10:58.410]But that's me kind of consuming
[00:11:01.050]versus there's this production kind of bit of it, right?
[00:11:04.410]Where you are creating chatbots that very much are
[00:11:10.560]leveraging AI in slightly different ways.
[00:11:14.280]And so, it feels like
[00:11:17.160]that notion of, there's a variety of
[00:11:19.560]different ways to come into it
[00:11:22.260]and to use it effectively.
[00:11:23.820]And, also understand that we don't quite know
[00:11:26.160]what effective use looks like yet, right?
[00:11:28.800]Like, there's some stuff that goes really, really well,
[00:11:33.560]but we're still trying to figure out
[00:11:35.970]how to best build queries
[00:11:38.550]to get it to do particular things.
[00:11:40.320]And I can only talk from like a consumption process.
[00:11:44.430]I'm not creating bots, I'm not creating sites, I'm not,
[00:11:47.190]I'm literally utilizing it to
[00:11:50.580]help me think through some stuff, right?
[00:11:54.690]But it's that notion of,
[00:11:57.030]I'm asking it to help me think through.
[00:11:59.100]Like, I'm using it as a tool to refine,
[00:12:03.090]give nuance to,
[00:12:06.901]and reposition my own thinking,
[00:12:09.300]versus I'm not asking it to think for me.
[00:12:11.940]Which,
[00:12:13.410]sometimes, feels like
[00:12:15.810]that's what the expectation of generative AI is,
[00:12:18.720]is it will do the thinking for me.
[00:12:20.550]I don't have to think, 'cause I'm gonna put it in,
[00:12:23.640]it's gonna do this work for me.
[00:12:27.850]Which, you know, I don't think it's there yet.
[00:12:29.400]I don't think that it could do that without us
[00:12:32.910]being able to
[00:12:35.160]play with it a little bit, right?
[00:12:36.900]Like, play with what it's giving us, in terms of output.
[00:12:40.320]So, as you're thinking about
[00:12:42.840]how you're using it right now, and that,
[00:12:45.360]I'm struck by Justin's idea of like
[00:12:48.750]how we interact with that knowledge.
[00:12:51.180]Like, what are some of the ways
[00:12:52.230]that you're interacting with it?
[00:12:53.490]What are some of the ways that you're,
[00:12:55.500]it's not just this passive thing, right?
[00:12:57.390]It's,
[00:12:59.309]we are doing particular things
[00:13:02.220]with generative AI and the products that it's giving us.
[00:13:05.610]How are you thinking about those heuristics,
[00:13:08.490]in terms of when you realize,
[00:13:10.747]"Oh, this is going in the wrong direction,"
[00:13:13.110]and then how do you redirect?
[00:13:14.610]Or do you?
[00:13:15.443]Or is that the point where generative AI
[00:13:16.890]just stops being productive?
[00:13:21.300]I think on a-
[00:13:22.410]So, to think about it,
[00:13:25.530]sort of, at a wide angle,
[00:13:28.890]some of the research I'm doing with
[00:13:30.870]Gretchen Larson, a PhD student here,
[00:13:33.600]and Aze Dehasani, also a PhD student in the ILT program,
[00:13:38.210]we've broken it down into four different categories.
[00:13:41.310]There's how do we think about generative AI
[00:13:44.130]within academia in general?
[00:13:47.100]And, you know, in society?
[00:13:49.230]And then, how do we use it, like you described Nick,
[00:13:52.410]like how is it as a,
[00:13:55.500]as a helper?
[00:13:56.340]As something that's generating information,
[00:13:59.550]because we have a task at hand
[00:14:00.900]and we wanna solve that problem.
[00:14:02.370]We wanna overcome or address something.
[00:14:04.770]And then, there's, how do we use generative AI to design?
[00:14:11.100]And what does that look like?
[00:14:12.270]Because it will help you write code.
[00:14:14.518]Good luck if you don't know code.
[00:14:18.030]I mean you will overdrive your headlights fast.
[00:14:20.580]Just like, you know, if you're asking for advice.
[00:14:23.520]You know, if your friend looks like they're gonna die
[00:14:27.060]because they've been poisoned,
[00:14:28.650]and you don't have access to anything else
[00:14:30.330]except for ChatGPT, I understand why
[00:14:32.430]you'd have access to other things, you know,
[00:14:34.980]I would trust it if you're not a toxicologist.
[00:14:38.130]But, you know, in many situations,
[00:14:40.800]you know, especially when it comes to coding,
[00:14:42.720]you can get in trouble quickly.
[00:14:44.520]But, if you know what you're doing,
[00:14:46.020]it can magnify, amplify what you're doing and,
[00:14:49.680]you know, really move you down the line.
[00:14:51.240]And then, the final way, the fourth way, is as a material.
[00:14:54.840]Because generative AI itself is an API, or can be an API,
[00:14:59.610]where you can hook it up to a flow and
[00:15:02.119]it'll have things come through.
[00:15:05.070]But, I think that, I think on all those levels,
[00:15:07.470]on the three last levels, is an iterative process.
[00:15:11.280]Can you tell me API and flow?
[00:15:12.990]Break that down for me like I'm four.
[00:15:17.697]An API, it's like, it's like if you went to-
[00:15:24.540]So, if it's 1990 and you're looking for an apartment
[00:15:28.380]in New York City-
[00:15:29.310]As a 9-year-old.
[00:15:31.170]Okay, 9-year-old, not four?
[00:15:33.270]Not four. Okay.
[00:15:35.400]If it's 1990 and you're looking for
[00:15:37.230]an apartment in New York City,
[00:15:38.730]you probably have to go to some office
[00:15:41.460]where they have a big list of all the available apartments.
[00:15:45.060]And then, you have to know
[00:15:47.400]how to get to the different addresses,
[00:15:49.680]and go there, or call, and find, and check it out.
[00:15:53.370]Whereas, an API, you know, now, you would just,
[00:15:56.790]you know, there'd be a website that would
[00:15:59.100]drop all the
[00:16:00.930]little pins on a map for you,
[00:16:03.150]and you would set your price range,
[00:16:05.370]and it would call out the ones
[00:16:06.990]that didn't fall into that price range,
[00:16:08.760]or nothing that didn't have a hot tub.
[00:16:10.620]Whatever it was that you needed in an apartment.
[00:16:13.530]And then, it would all be there.
[00:16:14.790]In New York City? Well, yeah.
[00:16:16.616]In New York City with a hot tub.
[00:16:18.073]I mean, hey.
[00:16:20.040]And so, an API is sort of a way to
[00:16:24.750]ask for things in the background.
[00:16:26.940]So, give me a map, and then I have a, you know,
[00:16:30.600]I have this database of all these apartment listings,
[00:16:33.270]or I have a website with all these apartment listings,
[00:16:35.490]and each apartment has an address.
[00:16:37.710]And so, you can use the API from the map
[00:16:41.100]and the API from the, well,
[00:16:43.440]let's say from the apartment listings,
[00:16:45.330]and combine them, and it'll show where they are.
[00:16:48.840]So, an API for generative AI would be OpenAI saying,
[00:16:53.407]"Okay, for this amount, you get an account,
[00:16:56.520]and you can ask it questions in the background."
[00:16:59.190]Because maybe you want,
[00:17:01.020]you want it to do something specific.
[00:17:02.880]Like, you have a website where you're
[00:17:04.830]talking about fantasy football,
[00:17:06.330]and you're making recommendations to people.
[00:17:09.240]And you only want the bot to come back
[00:17:11.220]with fantasy football
[00:17:13.680]responses and answers, and not answer anything else.
[00:17:17.070]So then, in the background,
[00:17:18.870]you're asking OpenAI the questions
[00:17:22.170]they have for your technology,
[00:17:24.750]but you're also putting something in saying,
[00:17:26.377]"By the way, use this information
[00:17:29.100]and don't answer the question
[00:17:30.750]if it's not about fantasy football."
[00:17:33.560]Hmm, okay.
[00:17:36.060]And flow?
[00:17:38.460]Is flow just the order of operations?
[00:17:42.300]In this case.
[00:17:43.710]Sure, how you'd, like-
[00:17:46.410]How you would get-
[00:17:47.610]So, you have to have a sense of, okay,
[00:17:50.548]where's the irritant in the oyster coming from?
[00:17:54.480]It's probably, in terms of generative AI,
[00:17:56.640]it's usually coming from a person saying,
[00:17:59.917]"I have this need."
[00:18:01.320]Okay, then, you have this need, and then where does it go?
[00:18:04.140]This is much better with a flow chart.
[00:18:06.030]And so, that's where flow comes in.
[00:18:07.920]Where, how does your, how is your data
[00:18:09.780]going through the system,
[00:18:10.950]and how is it changing at each step?
[00:18:13.448]Okay.
[00:18:15.300]Based on APIs that could be generative AI.
[00:18:18.480]Based on natural language processing.
[00:18:20.250]What are you doing to the information that's going through?
[00:18:22.590]How are you making your decisions?
[00:18:24.570]So, you can think of it kinda like a flow chart
[00:18:26.160]or a decision tree, but with lots of data
[00:18:28.770]coming in at different points.
[00:18:30.180]Gotcha, okay, that's helpful.
[00:18:32.250]Cause, yeah, I have enough technical information
[00:18:33.750]to be dangerous, but not enough to, like, be articulate.
[00:18:38.400]So, we have bots, and, or API embedded into-
[00:18:43.050]You notice that we have bots.
[00:18:44.580]Nick, you don't have a bot.
[00:18:46.170]We have.
[00:18:47.250]The cheese stands alone, it's fine, whatever.
[00:18:51.570]What are some ways you use AI in teaching?
[00:18:56.430]So, I don't have a bot,
[00:18:59.460]'cause, apparently, I was not invited to that club.
[00:19:01.230]Thank you very much.
[00:19:04.320]Rude.
[00:19:05.910]So, I think,
[00:19:08.970]if I'm thinking about
[00:19:11.010]the ways that I've used generative AI,
[00:19:13.980]in terms of
[00:19:16.530]a thinking partner, again, helping me identify gaps,
[00:19:20.700]but have also, in the course of the last,
[00:19:23.640]last semester in particular,
[00:19:25.380]really tried to help my students understand
[00:19:28.770]the ways that they, themselves, could use generative AI
[00:19:33.030]to support them in their teaching.
[00:19:36.420]And so, one of the things that I have found
[00:19:39.180]really, really helpful is
[00:19:43.110]when they leverage a ChatGPT, or a Bard,
[00:19:46.710]or something of that nature,
[00:19:50.790]having them design a lesson plan, right?
[00:19:54.300]Like, this is the instructional routine
[00:19:57.030]that I think would work.
[00:19:58.320]This is the content that I'm trying to teach them.
[00:20:00.960]Whether it's letter-sound relationships,
[00:20:02.700]or a comprehension strategy, what have you.
[00:20:05.460]And then, feeding that lesson plan
[00:20:09.600]to the particular portal,
[00:20:13.740]giving them some characteristics for students.
[00:20:16.290]So, feeding the generative AI assessment data,
[00:20:21.030]and asking the generative AI to help them think through
[00:20:24.180]where might this lesson breakdown?
[00:20:26.160]Where might instruction get tricky for students
[00:20:29.580]based upon what you know about them?
[00:20:33.600]And that's been really interesting as
[00:20:39.240]it's forced my students,
[00:20:41.430]who are learning to be classroom teachers,
[00:20:44.850]to think through and about
[00:20:49.620]how to put that assessment data
[00:20:52.140]into conversation with their teaching.
[00:20:56.389]And the AI has helped them get a little bit better at
[00:21:01.200]figuring out, "Oh, this is how I planned it,
[00:21:03.450]this is what, how it's in my head."
[00:21:06.570]But it's given them, kind of, some advice on how to
[00:21:11.940]fix those spaces where,
[00:21:14.850]pardon, instruction might break apart a little bit.
[00:21:17.880]And so, that's been really interesting.
[00:21:19.500]And there's something about
[00:21:22.530]the feedback coming from
[00:21:25.980]ChatGPT or Bard that, like, they're getting the kind of
[00:21:29.400]feedback that I would want to give them,
[00:21:31.140]but they don't see it as a personal attack from me,
[00:21:34.050]as their instructor, which has been a nice,
[00:21:36.510]kind of, side benefit of it.
[00:21:40.020]So, trying to think through,
[00:21:41.010]I've been trying to think through ways to
[00:21:44.430]leverage AI,
[00:21:47.430]generative AI, as kind of a partner
[00:21:50.580]in the work that I do with teacher educators,
[00:21:54.750]or with pre-service teachers,
[00:21:57.750]in terms of this isn't going away, right?
[00:22:01.317]And there's the real efficiency driven,
[00:22:05.610]kind of, stuff around,
[00:22:08.820]okay, you can use this to write your lesson plans,
[00:22:11.220]but I also need you to know
[00:22:14.130]when that lesson plan sucks.
[00:22:17.400]When it doesn't actually reflect
[00:22:19.080]the content that you need it to reflect.
[00:22:22.290]And so, let's not start there.
[00:22:24.540]Let's draft our own lesson plans,
[00:22:27.780]but then let's feed it back into generative AI,
[00:22:31.350]and see how that works.
[00:22:32.760]So, that's been interesting to me,
[00:22:35.100]because, well, and I don't have a bot.
[00:22:37.890]I don't either.
[00:22:39.480]You just said you had a bot.
[00:22:40.710]No, I did not.
[00:22:42.160]Oh, mean.
[00:22:43.770]Okay, sorry, I did not mean to.
[00:22:45.600]It's been a while, Guy, why are you mean to me?
[00:22:49.230]It's been a while, so we gotta start somewhere.
[00:22:52.949]I don't like how this is starting.
[00:22:56.550]So, I wanna start-
[00:22:58.170]So, my take is a little bit different.
[00:22:59.820]And this is exactly the conversations
[00:23:01.560]I think we all need to have, because,
[00:23:04.650]if we are all looking at it a little bit different,
[00:23:07.260]and we're all exploring it, we need to share the information
[00:23:09.900]and we need to be very clear.
[00:23:11.190]So, the first thing I think is very important
[00:23:13.140]is to be very clear with our students
[00:23:14.550]about the expectations.
[00:23:16.230]How do we see this?
[00:23:17.400]What do we want them and don't want them to use?
[00:23:21.928]And include that in the syllabus,
[00:23:23.370]or wherever we're sharing information
[00:23:25.530]with all of our students, because
[00:23:27.360]they are unsure, and because it's really different.
[00:23:29.820]There are many of our colleagues who say,
[00:23:32.437]"Do not use any of these things."
[00:23:35.587]"I want you to be here, to be discussing,
[00:23:39.330]to be reading, and doing things
[00:23:41.310]just like we did in the 19th century."
[00:23:43.830]And that's gotta be fine.
[00:23:45.960]And, that's fine, that's a fine attitude,
[00:23:48.150]it's just gotta be very clear.
[00:23:49.500]And, if we are being
[00:23:52.680]supportive and, also, want them to play with it,
[00:23:56.460]and to experiment with it, and to do interesting with it,
[00:23:59.220]I think we need to be very clear.
[00:24:00.330]So, that's the starting point.
[00:24:01.920]The second thing is, I actually do use it for,
[00:24:06.120]to ask them to generate some lesson plans,
[00:24:08.790]and do some compare and contrast,
[00:24:11.070]because it can change their role as teachers
[00:24:14.880]from people who
[00:24:17.706]are doing the grunt work of writing the lesson plans
[00:24:20.280]to the ones who are, can be critical
[00:24:23.550]and detail oriented towards that,
[00:24:26.160]and ask the hard questions.
[00:24:27.840]Because, when you are writing the lesson,
[00:24:30.000]it is actually much harder to be critical about it.
[00:24:32.940]Because, going back to your point about,
[00:24:35.820]how do you respond to criticism, right?
[00:24:37.860]It's hard to be critical about what you just spent your
[00:24:40.980]blood, sweat, and tears at creating.
[00:24:43.800]Now, with enough time,
[00:24:45.360]you can be really critical about your own work.
[00:24:47.640]But, if you're doing this in real time
[00:24:49.680]for a lesson that needs to happen in two days,
[00:24:52.170]you're much less likely.
[00:24:53.160]So, actually, generating lesson plans
[00:24:55.710]has, also, an advantage of,
[00:24:58.530]now I'm the critical voice seeing,
[00:25:00.630]is this really what I need?
[00:25:01.860]Does it answer what needs to happen in the classroom?
[00:25:05.640]How can I make it better?
[00:25:07.680]What are the biases that are embedded in there?
[00:25:09.810]And I think it's actually easier if
[00:25:12.030]some of it is created.
[00:25:14.010]Also, because I think most, or many of our students
[00:25:18.030]will actually end up teaching lesson plans
[00:25:20.880]that they didn't necessarily write.
[00:25:22.770]That are coming from
[00:25:24.750]curriculum creators of one sort or another.
[00:25:27.570]And so, they're just reinterpreting existing lessons
[00:25:31.470]more than creating them from whole cloth.
[00:25:33.570]So, that's an interesting and somewhat different issue.
[00:25:36.510]And what's in the back of my mind is,
[00:25:37.890]how long before these big curriculum providers
[00:25:40.530]provide an API that helps
[00:25:44.427]teachers create a version of their lessons
[00:25:48.390]that are much more fitted to the needs of
[00:25:51.660]their specific classroom, and maybe even-
[00:25:53.110]Well, and we're seeing that now, right?
[00:25:55.110]Like, one of the things that I think
[00:25:58.470]is important
[00:26:01.976]is that level of
[00:26:05.580]engagement
[00:26:06.750]with these
[00:26:09.870]chat bots.
[00:26:10.703]With these portals.
[00:26:14.325]Like, one of the places I've been
[00:26:17.130]really playing with is edua.ai, not sponsored.
[00:26:20.970]But, like, it does exactly that.
[00:26:23.340]You can feed it a lesson plan,
[00:26:25.680]and then tell them characteristics of your class,
[00:26:28.110]and it will differentiate.
[00:26:30.870]And it will spit out the,
[00:26:34.275]"Okay, here's the version of the assignment
[00:26:36.150]for this group of students,
[00:26:37.350]because they have these characteristics."
[00:26:39.067]"Here's the stuff for this group of students."
[00:26:43.005]And so, one of the things that I'm finding
[00:26:45.660]really heartening about this is,
[00:26:49.500]about these technological developments, is it's
[00:26:52.200]allowing for
[00:26:54.450]differentiation at a scale that doesn't come at
[00:27:00.390]the cost of, "Oh, you need to spend
[00:27:02.820]your entire weekend doing this," right?
[00:27:05.220]Like, instead you can ask
[00:27:09.180]a portal to do that kind of work.
[00:27:11.970]It will design the resources.
[00:27:13.680]You have to check them over, of course.
[00:27:15.510]And this is where the, like,
[00:27:17.460]one of the things that I am-
[00:27:19.380]One of my personal arguments, I think in this age of AI, is
[00:27:24.300]the level of expertise, the role of expertise,
[00:27:27.870]is going to be more important than ever,
[00:27:30.810]because you are going to have to have
[00:27:34.620]a lens
[00:27:36.300]with which to bring to these products
[00:27:38.730]that are going to be generated.
[00:27:40.770]And you're going to have to have a sense of
[00:27:44.250]when is it going awry?
[00:27:45.390]When is it not?
[00:27:46.350]When is that, is that actually going to work, in terms of
[00:27:50.580]the differentiation for the kids
[00:27:52.260]that you're seeing in your classrooms, right?
[00:27:56.547]And so, that's where,
[00:27:58.140]like, that's why I start with the,
[00:27:59.700]you're designing this lesson, because I need you to have an
[00:28:02.014]internal understanding of how a lesson
[00:28:04.830]is structured and how this instruction
[00:28:10.357]is forecasted, right?
[00:28:11.400]What does it feel like to plan for this type of work?
[00:28:14.940]And then, how does that feel when you're interpreting
[00:28:18.420]that same work that someone else has done.
[00:28:20.550]And how can you reconfigure that?
[00:28:23.910]And so, thinking about that,
[00:28:26.940]that role of expertise, how do I give my students
[00:28:32.280]enough background
[00:28:34.920]to be dangerous, and enough background to be like critical
[00:28:41.925]without having to cover it all.
[00:28:43.110]'Cause you can't, right?
[00:28:44.190]Like, the world is too complex,
[00:28:46.350]but you have to have enough.
[00:28:47.520]Like, I'm thinking about the coding part.
[00:28:51.120]ChatGPT has been really helpful as I'm coding
[00:28:55.016]within my Canvas sites.
[00:28:56.160]But, again, I know just enough
[00:28:59.040]to know what I'm asking for.
[00:29:01.020]And just enough to
[00:29:04.170]tweak as needed.
[00:29:07.080]But, if I were to do like anything
[00:29:09.360]fancier than
[00:29:11.460]tabs with embedded da, da, da, da, da, da,
[00:29:15.150]I'd, you know, have some problems.
[00:29:18.240]But it gives me enough to
[00:29:20.880]kind of go on.
[00:29:23.355]It kind of feels a bit like
[00:29:24.300]when the internet first came out.
[00:29:26.010]Or when the internet-
[00:29:27.157]"When the internet first came out."
[00:29:28.620]Listen to me sounding old.
[00:29:30.150]But, like, when
[00:29:32.250]we realized the utility of the internet,
[00:29:35.130]the widespread utility of the internet.
[00:29:37.560]Where like it allows me to learn stuff I didn't,
[00:29:40.110]and do stuff that I didn't, wasn't able to do before.
[00:29:43.710]Like, how do I make these things happen?
[00:29:46.230]You're making notes. You're making notes.
[00:29:47.702]So, I wanna hear.
[00:29:49.875]It was a good conversation.
[00:29:51.090]Well, I usually use a medical,
[00:29:53.610]or, rather, a musical analogy for this,
[00:29:56.640]but maybe I'll also use a cooking one.
[00:29:59.040]So, initially, I would say,
[00:30:01.560]teaching is a game of chicken, right?
[00:30:05.070]At some point, there's the moment when the teaching happens.
[00:30:09.270]And that's the moment where, you know,
[00:30:12.330]you're, the game of chicken comes in the point where,
[00:30:14.640]when can I start planning?
[00:30:16.980]Because I know I'm, like,
[00:30:18.270]I'm never gonna feel like it's perfect.
[00:30:20.010]I'm never gonna feel like it's all done.
[00:30:22.230]And so, you might have to give yourself
[00:30:24.060]a certain amount of time.
[00:30:25.260]But what are you doing in that time?
[00:30:26.690]And how are you spending it?
[00:30:28.680]So, the musical analogy might be, you know,
[00:30:31.770]you're playing the oboe reasonably well, okay?
[00:30:35.790]But, like, you have this,
[00:30:37.830]you need this whole symphony to happen.
[00:30:40.080]Or, you know, you're supposed to make this meal,
[00:30:42.990]but you're spending most of your time
[00:30:44.490]peeling carrots and potatoes.
[00:30:46.740]And that's a frustrating thing.
[00:30:48.060]And it's a necessary part of teaching,
[00:30:50.130]where, at the beginning, you're grabbing from
[00:30:52.380]other people's lesson plans, other people's curriculum,
[00:30:54.630]because they're, like, because you don't have the content.
[00:30:58.260]And it takes a long time to get, you know?
[00:31:00.240]So, you can sort of chunk up, like, level up
[00:31:02.610]from peeling potatoes and cutting onions
[00:31:05.760]to sort of really thinking about
[00:31:07.080]how all the flavors are blended together.
[00:31:08.970]That's similar, you know, in terms of a musical analogy,
[00:31:11.880]of going from playing the oboe, to leading the ensemble,
[00:31:16.410]and then conducting the orchestra.
[00:31:18.570]And that's something that generative AI
[00:31:20.773]gives us a chance to sort of try our hand at,
[00:31:24.330]to do a little sooner than we would've otherwise.
[00:31:27.000]Especially when we've got people that are sort of
[00:31:29.070]averse to human,
[00:31:31.710]human given formative feedback.
[00:31:35.070]In the research that that we did with graduate students,
[00:31:37.800]we had a student who was
[00:31:41.010]more capable than I am at coding,
[00:31:42.930]which isn't saying that much, but it's saying something.
[00:31:45.780]And, in the interviews with this participant,
[00:31:49.950]they said the, like, for them, it felt like a failure
[00:31:53.010]if they had to go to office hours
[00:31:55.050]to get help on a project.
[00:31:57.144]But that's where generative AI came in for them.
[00:32:01.110]So, they could hear the, you know,
[00:32:02.730]they could hear the critical feedback
[00:32:04.830]and it didn't feel real.
[00:32:06.720]Or we were talking, before we came on camera,
[00:32:10.620]about feelings.
[00:32:12.180]And this idea that, you know,
[00:32:14.160]if you can remove the feelings from it,
[00:32:16.050]and maybe, that's partially what generative AI does.
[00:32:19.290]Is the affective place that sort of makes us
[00:32:22.200]sort of wince inside when someone says this isn't good.
[00:32:25.770]Or this could be much better.
[00:32:27.120]Or you forgot four things.
[00:32:29.280]I think that might be, at least, less,
[00:32:32.160]if not, gone for a lot of people,
[00:32:34.410]when they're using generative AI,
[00:32:36.360]that there's no human judgment attached to it.
[00:32:38.850]It's just them, in their dorm at 2:00 AM
[00:32:42.840]getting some help on a lesson plan that they've developed.
[00:32:46.230]And they're iterating at a level
[00:32:47.610]that's above peeling carrots and potatoes.
[00:32:50.640]And I think that's really exciting
[00:32:52.200]for teachers as well as other fields.
[00:32:57.780]Yeah, I like that.
[00:32:58.890]Because we both kind of talked about it in some way,
[00:33:02.070]and it reminds me of work I've done many years ago on
[00:33:05.820]kids practicing fluency, which is a totally different thing,
[00:33:08.640]but it was the same thing.
[00:33:10.290]And that is, when they had automated response
[00:33:13.860]that gauged, they were willing to practice
[00:33:16.857]20, 25 times reading.
[00:33:18.990]It was actually too many times reading the same text,
[00:33:21.450]but they were willing to go back to it.
[00:33:23.310]Whereas, when there was a person there, they wouldn't do it.
[00:33:25.980]They would do it two times, and then they would shut down.
[00:33:28.410]With the machine, listening to the same text,
[00:33:31.260]offered the same way, they would go again,
[00:33:33.600]and again, and again, and they would be
[00:33:36.480]willing to vocalize and practice everything.
[00:33:40.860]So, affect has,
[00:33:42.210]that idea of affective filter, as a way,
[00:33:45.677]it's a really interesting way to think about,
[00:33:47.940]this is what the copilot can do for me in teaching.
[00:33:51.510]Here's a voice, here's a coaching voice.
[00:33:53.790]And the other piece is, of course,
[00:33:55.170]that it's available when I need it.
[00:33:57.870]That immediacy is really important.
[00:33:59.850]I mean, we talk about this.
[00:34:00.900]This can be late at night, or early in the morning,
[00:34:03.840]or on the weekend.
[00:34:04.673]And, suddenly, you don't have to bother anybody,
[00:34:06.930]who will, if they're reasonable, not respond in real time.
[00:34:11.310]You can get a response in real time and move on.
[00:34:13.860]I can do, and I did what you talked about,
[00:34:16.110]and that is, as I'm planning my own syllabus,
[00:34:18.390]I asked ChatGPT to create a syllabus for a similar class.
[00:34:21.570]And I just looked to see, do I have everything?
[00:34:24.480]And I was like, "Yeah, I'm okay."
[00:34:26.580]But, again, expertise comes in.
[00:34:28.680]But, you know, I've asked for other syllabi.
[00:34:31.740]And I've looked and compared.
[00:34:33.060]But it was really nice to have that
[00:34:35.550]immediate response to what I'm doing.
[00:34:38.340]Yeah, I am curious about this, and this is a
[00:34:41.460]topic for another time.
[00:34:42.990]But, like, the lack of affectiveness
[00:34:46.470]is like-
[00:34:48.870]I'm curious about what that means for feedback, right?
[00:34:51.990]Like, I'm curious-
[00:34:54.803]When
[00:34:57.270]something is-
[00:34:59.490]'Cause when, I'm thinking about when I'm
[00:35:01.200]personally using ChatGPT
[00:35:04.440]as a thinking partner.
[00:35:05.670]Like, it's usually in preparation of
[00:35:08.850]going live with
[00:35:11.910]something else, whether it's,
[00:35:14.040]you know, an interview protocol.
[00:35:16.743]Or whether I'm thinking through my syllabus.
[00:35:18.840]Whether I'm thinking through my teaching, what have you.
[00:35:22.680]And, like,
[00:35:26.820]ChatGPT's not my stakeholder.
[00:35:29.070]And so, that feedback's slightly different
[00:35:32.460]in some way, shape, or form.
[00:35:34.410]And it doesn't feel like that.
[00:35:36.750]Like, it's nice to get that feedback,
[00:35:38.100]but it's not necessarily feedback
[00:35:39.540]that matters yet to me in my head,
[00:35:42.840]as we're having this conversation.
[00:35:44.340]And so, I don't know where I'm sitting on that,
[00:35:47.340]but that was interesting to me.
[00:35:49.050]Like, the-
[00:35:51.150]If it's not-
[00:35:56.034]If I am not engaged in feedback emotionally,
[00:36:01.110]am I going to actually take that?
[00:36:02.940]Am I gonna take that up?
[00:36:04.290]So, I was thinking- As I'm thinking that-
[00:36:06.510]I'm trying to think that through.
[00:36:07.890]And I was thinking about it exactly opposite.
[00:36:11.275]I mean, why are we not surprised?
[00:36:12.450]Yeah, we are not surprised.
[00:36:14.670]So, I was thinking about it as like,
[00:36:17.100]I like it in some ways, because
[00:36:19.525]I'm willing to take that critique.
[00:36:22.230]My challenge is,
[00:36:24.624]actually, the opposite.
[00:36:26.880]And that is, some of our natural resistance may be good.
[00:36:30.510]So, some of our natural resistance to feedback is bad,
[00:36:33.840]because we're not listening.
[00:36:35.160]But some of it is good, because, sometimes,
[00:36:36.840]we need to stand our ground, and we're actually correct.
[00:36:39.570]And the feedback is actually wrong.
[00:36:41.670]And, maybe, with the lowering of the affect,
[00:36:44.610]if you're like, "Okay, I'll take that
[00:36:46.530]critique and I'll move on, because I need-"
[00:36:47.963]Well, I don't know that it's actually critique.
[00:36:50.504]And that might be what I'm trying to figure out.
[00:36:53.580]I haven't thought about AI in this way before.
[00:36:56.944]My point is, let's say it is critique,
[00:36:58.770]or I'm accepting it as critique.
[00:37:00.720]And, maybe, because there's no affective,
[00:37:03.540]the affective filter is kind of lower,
[00:37:06.000]I'm actually taking it in when I shouldn't,
[00:37:08.010]when I should resist.
[00:37:09.930]And so, I'm wondering about that.
[00:37:11.982]Now, if it's something like coding,
[00:37:15.300]there's a very objective test of if it worked or not.
[00:37:19.470]But, if it's something like making a plan for a syllabus,
[00:37:22.710]then it's squishier, it's harder to judge.
[00:37:24.960]And there's definitely not an immediate response of,
[00:37:27.167]did it work or did it not work?
[00:37:29.430]So, that's where I'm a little bit,
[00:37:33.180]I'm wondering about the boundaries.
[00:37:36.480]Well, and I think asking questions.
[00:37:38.550]So, like, out in the world,
[00:37:42.720]around other future teachers or around colleagues,
[00:37:46.890]there's a certain level, everyone has a certain level of
[00:37:49.680]pressure to perform competence.
[00:37:51.990]Right?
[00:37:52.823]And so, in front of the bot, as long as you feel
[00:37:55.860]like your transcripts are your transcripts, you know,
[00:37:59.760]that feeling of performing competence diminishes.
[00:38:03.960]So, that's something.
[00:38:05.393]And so, like, that sense of vulnerability
[00:38:07.980]when you ask certain questions, isn't there.
[00:38:11.040]Or, at least, that's what our participants have implied.
[00:38:14.250]That they don't feel as vulnerable when they're asking
[00:38:17.613]certain questions than they would otherwise.
[00:38:22.304]And, you know, you take away that vulnerability,
[00:38:24.540]it's like when I was doing my dissertation study.
[00:38:28.110]I was doing an ethnographic study at a elementary school.
[00:38:31.680]I was there, and I cared about those teachers,
[00:38:33.900]and they were great.
[00:38:34.770]And I hung out for over a year.
[00:38:36.870]And I didn't, I wasn't invested in the school
[00:38:39.840]like my job depended on it
[00:38:41.640]or those were my students in the same way.
[00:38:44.340]And so, it was sort of interesting to say,
[00:38:46.177]"Oh, if you have this ethnographic eye,
[00:38:48.000]you can sort of see things you can't see
[00:38:49.590]if you're too emotionally involved in what's going on."
[00:38:52.920]And so, at 2:00 AM,
[00:38:54.930]you're getting feedback that's, you know,
[00:38:58.200]magnifying you.
[00:38:59.490]Even if the feedback isn't as good as your own knowledge,
[00:39:02.790]it can still be useful.
[00:39:04.530]You know, I worked on a software development project
[00:39:07.410]with someone who just happened to be
[00:39:09.030]in the University of Texas library reading the newspaper.
[00:39:12.150]And he would talk to me every day.
[00:39:13.950]And, one day, I mentioned a project I was working on,
[00:39:15.600]and he said, "Hey, can I be a part of that?"
[00:39:16.837]"I don't know anything about technology,
[00:39:18.840]and I'd love to sit in on your meetings."
[00:39:21.000]And, you know, most of the time, it was sort of random.
[00:39:24.360]But, sometimes, he would ask questions,
[00:39:25.950]and it would be useful.
[00:39:27.090]So, even if it's not up to our level,
[00:39:30.690]in terms of what it knows,
[00:39:31.890]it can still be that irritant in the oyster
[00:39:34.800]that gets us thinking.
[00:39:36.210]And so, removing some of that affect.
[00:39:39.030]I mean, even when,
[00:39:40.830]Guy, when you mentioned about the syllabus,
[00:39:43.590]you qualified it afterwards by saying,
[00:39:45.937]"I also looked at other syllabi," you know?
[00:39:48.000]Yeah, yeah, yeah.
[00:39:48.833]Because there's that sense of vulnerability that,
[00:39:50.677]"Oh, you would have this question?"
[00:39:52.710]And our students are having questions all the time in class
[00:39:56.040]and they're swallowing them,
[00:39:57.540]because they're performing competence.
[00:39:59.460]And so, it's a nice sidebar,
[00:40:02.130]whether it's a back channel during class,
[00:40:04.650]or it's after class, where,
[00:40:06.480]if they don't know things, they can look it up.
[00:40:08.070]I mean, and you can do that already.
[00:40:09.420]You can be like, "Oh, I forgot what a word was."
[00:40:12.667]"I'm gonna look it up instead of raising my hand,"
[00:40:14.970]like you would've done in 1990.
[00:40:17.096]You know?
[00:40:18.810]And so, in that area, I feel like it's useful.
[00:40:22.740]And then, at least in the case of OpenAI,
[00:40:27.060]with RLHF, the way they trained from 3.5 to 4.0,
[00:40:32.880]it's a lot more, some would say, nice or PC.
[00:40:35.610]It usually starts with something positive.
[00:40:37.860]Sort of going along the-
[00:40:38.693]I find it really annoying, actually.
[00:40:40.107]I'm not gonna lie.
[00:40:41.700]I don't need it either.
[00:40:43.260]But, I think, for some people, it's really helpful.
[00:40:46.830]And that's, they think it's helpful at OpenAI.
[00:40:50.460]And so, those sorts of things continue,
[00:40:52.620]and it sort of softens if people are feeling,
[00:40:55.230]you know, that they could be judged.
[00:40:57.840]I think, eventually, you know,
[00:41:00.210]chat histories are gonna leak.
[00:41:02.430]And then, people might,
[00:41:03.824]you know, be a little more concerned.
[00:41:04.890]It's like, right now, a lot of people, like,
[00:41:07.590]never want to be in one of those YouTube videos,
[00:41:10.140]where it's, you know, police camera footage of
[00:41:12.960]them doing something embarrassing,
[00:41:14.850]because they don't wanna be that type of famous, you know?
[00:41:17.520]Not that type of famous.
[00:41:19.350]Like, the goal is to avoid that.
[00:41:21.030]And so, I think we have a lot of students,
[00:41:23.760]even though we feel like youth are out taking risks,
[00:41:26.370]that are very risk averse in settings where they feel like
[00:41:28.830]they're being watched, observed,
[00:41:30.090]or their data's being collected, at least video data.
[00:41:32.940]So, until that happens,
[00:41:36.030]I feel like this is a less affectively
[00:41:42.090]dangerous, or they're gonna feel less vulnerable.
[00:41:44.790]And, according to the research we've started,
[00:41:48.420]that has some benefits.
[00:41:51.120]And we haven't had students saying,
[00:41:52.447]"Hey, I just," you know, "I thought whatever ChatGPT said
[00:41:55.740]was great, and I moved on."
[00:41:57.360]They've been concerned, one, they want to protect,
[00:42:00.090]they've said they wanted to protect their learning.
[00:42:02.520]They didn't want to get to a place where they were just
[00:42:04.680]feeding it everything, and it knew everything.
[00:42:06.900]And, two, for the students who tried
[00:42:08.700]to be really ambitious with their code
[00:42:10.290]before they were ready to understand how it was all working,
[00:42:12.870]like they couldn't be the,
[00:42:14.280]they couldn't be the conductor to watch the conductor
[00:42:16.830]to know when the conductor was messing up.
[00:42:18.900]You know,
[00:42:21.030]they outran, they out drove their headlights pretty quickly,
[00:42:23.940]and they're like, "Oh."
[00:42:25.170]And then, a piece of software becomes some sort of
[00:42:27.780]magical thing, where you don't want to touch anything.
[00:42:30.630]And you're like, "No, exactly how this is,
[00:42:32.880]I'm just gonna keep this here and never touch it,
[00:42:34.770]because it works and I don't understand how it works."
[00:42:37.216]Now, things like Copilot with GitHub and VS code,
[00:42:41.790]you can actually highlight that bit of code.
[00:42:44.010]And then, on the side say, how is this working?
[00:42:47.130]And I'm excited because I'm teaching a course
[00:42:50.160]in advanced web design and databases starting days from now.
[00:42:55.275]And, in that course, they're going to be using GitHub.
[00:42:58.560]They're gonna be doing computational literacies,
[00:43:00.540]learning in a analog or non-digital way.
[00:43:03.810]And then, we're going to, they're gonna be doing
[00:43:05.880]coding and using a Copilot in that way.
[00:43:09.870]So that, if they overdrive their headlights,
[00:43:11.850]maybe this helps them sort of come up to speed.
[00:43:14.280]So, they're not spend, you know,
[00:43:15.300]they're not peeling carrots and potatoes the whole time.
[00:43:17.820]They're not learning how to write a loop.
[00:43:19.770]Once they understand how loops work and where they fit,
[00:43:23.400]they can go from there.
[00:43:25.950]Yeah, the one thing that I'm wondering about.
[00:43:29.490]You're talking about performing a competence.
[00:43:32.160]I'm a little bit concerned.
[00:43:33.750]This is just a concern that I have in the back of my mind of
[00:43:36.570]people performing that
[00:43:41.550]sense of,
[00:43:47.070]the way you described it, as that sense of concern.
[00:43:50.460]So, are they really concerned
[00:43:52.050]or are they performing concerns,
[00:43:53.760]because they know they need to talk about it?
[00:43:56.400]And they know that they should not admit
[00:43:59.010]that they just kind of grabbed it and put it there.
[00:44:01.800]So, that's something that is in the back of my mind.
[00:44:05.430]Again, going back to that idea of expertise.
[00:44:08.400]If you have enough expertise, then you can evaluate
[00:44:11.550]and you feel more comfortable
[00:44:12.810]doing the comparisons and looking at things.
[00:44:16.912]But, if not, will that be the lesson?
[00:44:20.190]And that is, I need to say I'm very careful,
[00:44:22.680]and I'm checking, and I'm looking.
[00:44:24.750]Or am I just saying that, and then continuing with my life?
[00:44:28.320]Because, especially in higher education,
[00:44:31.140]where they need to perform eventually,
[00:44:33.690]because that's what they do in classes,
[00:44:36.544]they send us something and they say the right things,
[00:44:39.030]but did they do them?
[00:44:40.590]Yes, and if they say by writing,
[00:44:43.632]there's a level of obfuscation, right?
[00:44:46.750]In my J-term course, I've had my students shoot
[00:44:50.340]one minute mini videos and put 'em on YouTube,
[00:44:53.850]and embed 'em in the slides, our sort of shared slide deck.
[00:44:57.210]And so, they sort of narrate their take.
[00:45:00.450]So, they make a multimodal slide with their ideas,
[00:45:03.450]based on the readings.
[00:45:04.770]Images, some text, and then their video.
[00:45:09.150]And to do that video, it's really clear
[00:45:13.230]from one, you know, conductor to another,
[00:45:15.900]or a, you know, wannabe conductor, a future conductor,
[00:45:19.110]if they're not, like, if they didn't understand
[00:45:20.820]what was going on, you know?
[00:45:21.990]As long as they can't read
[00:45:24.060]some sort of script they wrote for their video,
[00:45:27.240]it sort of gives them, 'cause I used to feel like,
[00:45:29.647]"Okay, well the multimodal slide is my insurance,
[00:45:33.030]that this isn't just a copy paste."
[00:45:35.010]Because, you know,
[00:45:38.264]some terrible, I know this is a clean podcast.
[00:45:40.560]So, I'll just say, you know-
[00:45:41.904]Is it?
[00:45:42.737]It's rated as clean.
[00:45:44.490]When I looked on iTunes.
[00:45:47.550]Anyways, or the Apple Podcast thing.
[00:45:50.995]I wanna make sure that they're,
[00:45:54.300]what they say is how they're thinking and feeling,
[00:45:57.210]and that they can sort of do it like jazz
[00:45:59.130]more than something that's totally scripted.
[00:46:01.530]And so, I've gone from the multimodal slide
[00:46:03.780]to a video that they shoot.
[00:46:07.222]And, you know, I have a problem
[00:46:09.270]not leaving YouTube after I watch that video.
[00:46:11.940]But, if they embedded in the slide,
[00:46:13.800]then I don't have to go to YouTube.
[00:46:14.820]I can watch their video, and I can get a good sense,
[00:46:17.970]and then move on.
[00:46:20.580]All right.
[00:46:21.420]I think
[00:46:23.640]we will need a follow up for this episode
[00:46:26.190]to talk about the way AI can
[00:46:29.760]be used in research.
[00:46:31.140]I think that's an interesting thing to explore,
[00:46:33.210]but I think that we need to give it enough space,
[00:46:36.300]considering our time and everything else.
[00:46:38.880]Yeah, no, like,
[00:46:40.500]we've talked about teaching a lot.
[00:46:42.000]Yeah.
[00:46:42.833]We haven't talked necessarily about the abilities of
[00:46:46.380]AI to help with data analysis, and/or writing,
[00:46:49.170]and/or those other kinds of pieces
[00:46:53.310]that are connected to the conversation that we've had today.
[00:46:56.640]But we'll pick that up at another time.
[00:46:59.670]Yeah.
[00:47:00.503]So Justin, thanks.
[00:47:02.670]Thanks for inviting me.
[00:47:03.821]This was-
[00:47:05.460]Now I'm thinking about emotion,
[00:47:07.440]and feedback, and the role, and technological tools.
[00:47:12.840]A lot to think about. All kinds of fun things.
[00:47:15.540]And so, this is an exciting time.
[00:47:19.232]And I love that we have something new to think about.
[00:47:23.220]And a new way to question, what does it mean to teach?
[00:47:26.820]And what does it mean to learn?
[00:47:28.650]In many ways.
[00:47:30.232](upbeat music)

The screen size you are trying to search captions on is too small!

You can always jump over to MediaHub and check it out there.

Comments

0 Comments

Not That Kind of Doctor - Artificial Intelligence

Description

Searchable Transcript

Comments icon comment

Related Channels

Comments