Building Resilience in the AI Era: What Faculty Need to Know (Live from ICISER)

Welcome to the very first live stream of AI Goes to

College, the podcast that helps you figure out just what in the world is going

on with generative AI and higher ed.

So we want to thank the organizers of the International Conference on

Information Systems Education Research Workshop, the

conference committee. Tanya over there, who. You can't. Well, it's

a 360 camera. You might see her back there in the back. And Rosetta

Romano, who is online. She was the one who came up with this idea.

All we did was say, okay. And this is sponsored

by the association for Information Systems Special Interest Group on Education,

better known as SIG ED because I'm not saying that again. So this is

sponsored by SIG ed and we want to thank the SIG ED board for helping

us. And we also want to thank Hazel Crossler for

handling the sound. That's a really clever way

of saying, if there's any problem with a sound, blame it on

Hazel Crossler, who is, yes,

the child of Rob Crossler and. A

master's student in music at Middle Tennessee State University, which makes it all. The

more embarrassing if the sound really stinks.

We want to make this very loose. As you can tell, we've got some

questions, but we sincerely welcome questions from the audience,

either online or those of you who, for some reason just

have to sit down instead of all these tables that we set up our entire

arrangement around. Well, let's get started. So, Rob,

what are some surprising generative AI

Uses that you've run into? Surprising. None of the normal stuff.

Surprising. The surprising one that made faculty

squirmish was students learn how

to create transcripts for their presentations

and then feed those transcripts for their presentations into an AI

tool that automatically created voices for

their presentations, which stayed under the five minute limits and then were the most

boring presentation you ever saw, perfectly dictated

by a machine with a transcript created by a

machine. Squirmish is a word. Now,

is that okay? Okay. We've coined a word

that is surprising. What's surprising is that had to be

at least as much work as just recording the stupid video. It

was, but the students were comfortable because they didn't have to stand up in front

of people and talk themselves. They were able to push off

what I would call that friction point of professionally teaching and push that

to the machine. Yeah, that. That's a problem.

That's a problem. We're going to get into that sort of problem a little bit

more later. I've been doing vibe coding.

We'll talk. Thank you. Yes. By the way, all things AI goes to

college are available@aigostocollege.com so who is

Vibe coded? It is the strangest

thing. You tell the

AI interface what you want and it creates it.

So do you all know what a landing page is? So you just have this

landing page. It's a website, it's not a big deal to create. But if you

want one that really looks good and that drives the

kind of action you want it to drive, it takes a while to

do. I created one in about 15

minutes. Deployable out there on the web,

ready to go in 15 minutes. And I did several other things. I did these

micro apps, which I think is a made up word,

like squirmish. The little personal

world of new technologies. Craig, we're allowed to make up new words. That's right. That's

what we do. And so these, these are little productivity apps that I don't

want to pay a subscription for. A time blocking app, one that

helps me keep track of my projects with like a Kanban

board. These are things that I was paying 20 or 30 bucks a month to

do over Thanksgiving weekend. I

created these apps. And there's a payoff to this.

I'm not bragging because anybody could have done this. I think

this fundamentally changes the game for knowledge work.

That's what's the most surprising thing to me. I'm convinced

that information systems, our field, is

what I call the canary in the coal mine. That's the

metaphor that I'm using. It's the early signal for

what's going to happen with knowledge work. This thing that used to take

somebody with deep expertise days to do,

I did in a matter of a few hours. And I quit

coding 30 years ago. So I see this

coming for knowledge work of all sorts. So let me ask a follow up question

to that because I've seen that argument made. But the counter argument

is it's easy for the simple things, but when you take that to

the level of the enterprise and do

complicated enterprise apps, it falls down. And people spend more time

debugging and getting it to work than if they would have just done it themselves.

How do you see those two views coming together? So vibe coding is not for

that. That's not what it's for. What vibe coding is for

is for doing little stuff quickly. There are other

systems like Claude Code and codecs from

OpenAI and Google just released theirs, which has a very

weird name, Anti Gravity. Those

are environments where professional coders can make themselves

more efficient. But that's not what vibe

coding is. But One of the things we do, those of you who may not

be in information systems, one of the things we've been trying to do

for a long time is to not try to spend a lot of

time figuring out what the requirements for a system are by talking to users,

because they never know. You build these prototypes and let the

users use the prototype and say, wait, it doesn't do this, or I don't like

the way it does that. That's what you can do with vibe coding, because that

never gets rolled out up to scale with all the security controls

and that kind of thing. But it takes hours and hours and hours to do,

and now it doesn't. So what I hear you saying is that the way we

do systems analysis and design, that process is

completely changing and is able to do things

perhaps differently than we've ever done before. I think that's the case.

But I think that we're going to see this with other types of knowledge work.

You know, right now it's us because, you know, we're gearheads and we were the

early adopters for is and I mean for AI and

it just kind of makes sense for us. But you're going to see this go

into accounting, into finance, into law,

and it's not going to be. This does everything it's going to be.

It does this one thing really well,

small scale, or it makes other things easier. One of the

biggest mistakes people make with AI is they want it to do 100% of something.

That's the wrong way to think about it. When I give talks, I have a

slide that says 50% is greater than 100%. Try to get

AI to save you half of the effort. We have

faculty in the audience. What would you pay for something that cuts

your grading time down in half? If I

had kids, I don't know if it'd be my firstborn, but my second born would

be in trouble. I have to say the same thing right now. My

firstborn standing right next to. So you're, you're cool, but, you know,

I don't know, but, but that, that's what vibe coding does. It

may not do the whole scalability thing,

and it shouldn't, but it can do a lot. And I

really think that's what's coming for virtually

everybody that gets a degree to get a job. Yeah, I think

we're going to see some real changes. The prototyping, I

think, is going to be the first easy one. And, and where I see that

being great for information systems students in some ways and is it

doesn't require them to necessarily work with the engineers to

create the prototype solutions. No

offense to you engineers out there, it's not that we don't love you. But you

think about then where does the engineer's time get spent? If we

aren't quite yet at that point where we are able to

completely redo the building of those enterprise systems, the focus can

be on there. Another thing that I think is really important to point out as

we think about these things as large language models

which are part of the generative AI systems is there's a lot of pieces

to the generative AI systems that need the technical know how to make

sure that they work. And a lot of engineering thinking as

businesses start to customize and make their systems work just for

them, are going to be around that engine and that system that helps to

automate this in ways that are specific

to each company and to each industry. All right,

Anything else on that one? No, that was good. All right. Well,

we have questions from listeners. First one came from

a student who I'm going to refer to as AM because

that's their initials. So that's how I'm going to refer to them. So AM

sent us an email a couple of weeks ago and he

talked about desirable difficulty, which is a term that

was introduced by Robert A. Bork. And it's something

that we've been calling useful friction. So you don't learn

if you don't do any work. So

one of the problems with AI used inappropriately is it removes friction.

That's useful friction. And so he was commenting on that

episode where we talked about that. But here's

the question. In your experience as educators, what is necessary

for desirable difficulty and how might generative AI

be leveraged by the teachers and students to create

that kind of difficulty? So we've been

talking about this at Washington State University where I work a lot.

And one of the solutions that we've been leaning towards

and trying to push more towards is live projects in

more classes where we take something that is ill defined from

organizations that are willing to work with our students and give them these

real world problems that can't just be fed into

the machine to find an answer. And so the machine might be part of getting

there, but that process of large scale problem solving

is where we're trying to do more of.

Okay, is that trying to maintain the friction? Tell me more. I don't

get how that maintains the friction. Well, it helps. The students aren't

going to be able to come up with a simple Solution. They're going to have

to say, here's the process we need to go through to solve this problem. Here's

the inputs we have. They're going to start down a road and then they're going

to be given more information that is going to change their thinking. They're going

to have to pivot, they're going to have to adjust. And it's not going to

be the simple creation of a thing, but it is going to be

solving a problem that really doesn't have a known solution. So it really

is their process of problem solving that becomes more

what you're looking at as opposed to what is created. So it

shifts the friction. Correct. And also raises the bar

a little bit. This actually came up during the workshop today where we

talked about. And again, I hate this term, soft skills

and problem solving skills and that sort of thing. Isn't that soft skills?

Doesn't that make it sound like they don't matter? I got a new term that

I like to use, and I've heard this in a couple places, social

skills. It has more to do with their social interaction

as opposed to soft skills, which kind of does have a different

nomenclature to a different perspective. Yeah. So I like social skills.

Humanistic skills. That's way too geeky, but I like that one too.

People skills. Yeah, it could be. Could be. But

the. Well, I don't know. I'm gonna have to think about

that. So social is going to extend to AI here soon. So I don't know.

We're gonna have to think about that. The, the thing that what you were talking

about, Rob, triggered in my mind is, is this

idea of complex thinking and problem solving. You know,

you're not talking about just doing some little thing. You're having to think through

and kind of push the envelope and come up with a

solution to some sort of a problem. Where I see

this going with generative AI is what I've been

calling co cognition or co

created cognition with generative AI. So if you

use the AI to help you think

in ways that you might not be able to do on your own, but

also in ways that AI couldn't do on its own, putting thinks

in quotes, then you've created something with the tool

that couldn't be created without the tool, and the tool could not create on its

own. And I think that's where we need to be pushing. And that's what you're

talking about, I think. Yeah. And I think this is really coming up

with creative ideas, and that's what I want to empower people to do, encourage

people to do is to find ways to look at

a expecting more, because I think we can expect more in this world of

AI. But to begin looking at walking

alongside the students in their processing to be able to

evaluate their critical thinking skills, their problem solving skills, all

of those things that we've had a hard time doing before, I think the door

is open to that. And I think

if we were to think about an experimental world where every faculty member gets

to run their own experiments and how they're doing things in their classroom, we're

going to come up with a lot of things that don't work, but in the

process, we're going to come up with a lot of. Things that you're an administrator

now. That's a horrifying thought. Every faculty just going off and doing what they want

to do. Let's take a tangent on this.

So we've been having some email conversations with our colleague Franz

Belanger, who's here in the audience and is also

our co author on our book Information Systems for Business Colon An

Experiential Approach, which is in

Edition 5.1 and is published by Prospect Press of

Vermont. And the publishers are here with us,

Beth and Andy Golub. And we appreciate that how. That's our

first commercial. That's our first commercial on AI goes to college. And

I was talking about something. Oh, that plug, Craig. I was talking.

Yeah, they do pay us for that plug. Yeah, they do. They do quite well.

So we were communicating about how

you can not

AI proof, but how you can either enhance

or. We didn't use the word resistant. Resilient. Resilient. AI

Resilient. I like that term. And Franz, would you like to tell us about what

you've been doing? She would not. So you have to come up to the microphone.

I'm going to pay for this later. This is something that we're about to talk

about that every professor should be thinking about. And that is the AI

resilience of every class they teach. Because if you can cheat your way through higher

education, what we offer is worthless. So this is one of the most important

things that we're dealing with right now. Well, they are

able to cheat their way through and they're really smart. Students

will still be able to be what I've come up with this

semester. So I teach an online only

class, asynchronous. I'm not allowed to have them

be there at a specific time, so I can't do any of the live

exams. So I have a

semester Project. It's real life in organizations, as you were saying. So this

works really well. But I also want to know that they're actually reading the

materials for each module and that

they actually understand it. How do you do that? Well,

can't do tests, can't do essays. I've tried every single

essay into AI and it gets perfect grades.

You do quizzes? Well, they download the

PowerPoints, they give the AI to write a

script, they run the scripts to the questions and they

get full grades. So what do I do? So what

I decided to do this semester is I have four cases

that they discussed during the semester and

I created with the help of the AI, I

created knowledge questions. And so the

questions make them use the content of

the modules and apply it to the case,

but really not based on case questions that you would normally have.

And so they have to be able to use the content, each of the

modules, before I use three or four modules, put a question

in. They have to demonstrate the ability to think

and then to. Now, can they cheat this? Of course they can.

But. Hopefully most of

them will demonstrate some not. I have a question for

you. This just occurred to me. So

we say they can cheat with this and they can cheat. You know,

cheating was not invented by generative AI. Ask me how

I know. But maybe if they

cheat with an assignment like that, they're actually learning something.

Yeah, I was thinking about that too. So if they go to the

trouble that every step that they need to do

to beat my new thing, they're going to learn something.

I use AI all the time and I know

that I even get to the point of challenging the AI. Why

are you telling me this? And it's quite interesting.

If they get to that point, that's what they're going to do in their lives

later. My problem is in an online

class, I don't want the majority of them to just

download PowerPoint's lectures and send that

to me. As I've given the course.

I think what you said is something important. If we're all using

AI to help make our jobs more efficient and

better at them, why shouldn't we expect our students to? So in many

ways, guiding them in proper uses of these tools and

how to do that is the other part of what we need to be thinking

about. Why wouldn't we want them to use it? I want them to use

it. I taught doctoral seminars last year, all

year, and I encourage the students to use AI and show them how

to use it and told them how I was using it. It's a tool that's

going to be out there. It's something they're going

to use and if they don't, they're going to fall behind. So I

think, I don't think anybody in this audience is just outright banning

AI, but if you know people that are, I'd

push back on them a little bit or just stay out of it because it's

not your business, but that's your choice. One thing I would say though is as

you're encouraging students to use AI or whether it's colleagues,

transparency I think is crucially important because if we

do it in the darkness, we don't know what's going on, we don't know how

it's being used and we can't talk about what is

the viability. How good is that information that was received

if we don't know how it was created? So talking about that

openly is important. And I think the first way we get to that level of

transparency is to say, yeah, we allow it, but we're going to talk about

it. Yeah, I mean that's, that's absolutely

the case. If we try to outright ban it, they're going to hide it and

we're not going to be able to help them adjust the way they're using it.

So I did require disclosures. Tell me how you're using it. And

this. They're also very motivated students because

if they didn't get what I was putting out, they're going to

have a lot of trouble when it comes up to their comprehensive exams. And so

I didn't have, I don't have to give them a bunch of assignments and stay

on top of them. I don't have quizzes. They're motivated students. And this is something

we've talked about in the past. That's our big job is

to change higher ed from this point based

transactional view to something that taps into

their motivation to learn. We don't have to give them points

to learn. Cheats on Grand Theft Auto. Is that, is that still a cool game?

I don't know. I don't do games well. We got, we got business professors

making jokes about Grand Theft Auto before gtx. There you go. Okay. That's right. Oh

yeah, that's right. I forgot about that. Delayed. Delayed again, by the way.

Delayed again, by the way. But you know, we don't, we don't have to motivate.

You don't motivate little kids to learn. They learn because they want to know stuff

and we have to. It's A heavy lift. It's going to take a long time

to do, but that's what we need to be doing. I'm. I think that's our

number one job. Well, and it, you know, one of the things that is

essential for students after they enter the business world is

to continue learning. Because three or four years ago nobody is

using generative AI in their jobs and now it's

expected and you've had to figure out how to do it. You've had to learn

how to do it. And that's just one example. New technologies come around

all the time that change how we do things. And if you can't continue to

learn, which really is what I hope you're learning in college more than

anything else, you're going to be stuck. One

thing I want to ask you about, Craig, is we were talking last

night a little bit. Did you see that? He backed away from me. Yeah, I

know. When he said it's going to be good. Yeah. Because you kind of asked

me and you didn't love my answer last night, so. But

one of the things we've seen a lot of is the importance of oral exams

and how in an oral exam, a student sitting in front of you, you can

get a good sense. Did they learn the things, did they know the things? And

we're discussing that and the idea of scale came up

and how do you we a lot of

higher education to make money, to be able to pay the bills. We have classes

that are larger than 10, 10 students, this is easy.

100 students, not so easy. How does this scale to be

able to do oral exams or to do these creative things that maybe take

more one on one? Getting to know your student, send students. Isn'T

even the first row of the auditorium. I teach undergrads and this is the

other part of the conversation we've been having online about scalability.

I don't know the answer to that. I'm really concerned about trying to

find a way to scale, especially if you've got a highly

constrained set of circumstances like Franz does. I mean,

what I'm going to do is just base the big chunk of the grade on

two exams they have to come in and take, and then everything

else drives them to prepare for the exams. You

do this stuff, you're going to do better on the examination. You

don't want to do it. You're adults with agency. Don't do it. I don't care.

But you don't get to whine at me afterwards if you don't like your grade.

And I think we have. God, I'm going to

sound so old right now. We've

gotten these darn kids to where

they expect hand holding and they expect us to make the

decisions for them. You want to do the homework, do the homework. You don't want

to do the homework, I don't care. It's one fewer that I

have to grade. Now I'm exaggerating a little bit, but only a little bit. I

think if we treat them like adults, they respond more like adults. And that's not

my thought. I'm not the first one to say that. So I'm going to follow

up with that because I'm going to put my. Thank God, because I had no

idea where I was going. I'm an assistant professor and student

evaluations are super important to me. I'm not getting tenure without good student

evaluations. And it's easy for the full professor to say, who cares

if I fail a bunch of students because they won't rise to the challenge

when if that's me, I may not have my job at three or four years.

You know, I had the same attitude as an assistant

professor. I think they respond, well, as long as you're matter of fact, you don't

have to be a jerk about it. It's like, look, if you do what you,

what I'm asking you to do, you're going to do well in this class. If

you don't, that's kind of on you. I mean,

I can't do anything about that. I don't want to. I'm tired of being the

homework police, you know, that's not why I got a doctorate.

So, I mean, I've been doing this for a long time and I've never

really had students push back. But. But you cannot be a jerk.

You've got to be matter of fact, you can't be lording it over them,

bossing them around and just kind of, look, do it, don't

do it. I think you're crazy if you don't do it.

And I think a key part of what you say in that is the

importance of students understanding the why.

If we ask students to do something and we say this is what you need

to know to do well in this class, that's one thing. But if we can

connect that why a little bit further and help them to understand that knowing this

is going to help you with this other thing or

it's going to help you to succeed in this aspect of a career path

that takes you this way. When students understand the why? I found

that they are more willing to actually

do the hard thing because they set the payoff at the end. Not just busy

work. My professor asked me. Right. And that's another thing I think we need to

work on. We need to lean out a lot of our classes. Things get added

and added and added and added, and it's not

surprising that students don't see how it all fits together. And

so maybe we need to cut back on the number of topics or.

Okay, you need to really understand this stuff. Well,

these four things, when I used to teach database, it was like, you need

to understand conceptual modeling, logical modeling, and SQL.

We're going to cover a bunch of other stuff. You need to know what it

is so you can go look it up or ask somebody if you hear the

term. And that kind of thing works well because it's like, look,

when you go out, this is what you're going to do. And so you need

to know this stuff. But if we can't say that about every one of

the, I don't know, 50 or 100 topics we have in a class, because it's

kind of not true. And this is really hard in the Intro to Information

Systems class, because in is we could go

all the way back to teaching about how a mouse works. At what point do

we no longer need to talk about how does a mouse work? What are the

inputs and the outputs? And I think that's true even beyond those topics. But in

that intro class, it can get really, really big if we're not purposeful about saying,

what do students absolutely need? What do they really need to

know? An exercise I learned about that one university

was pushing their faculty to do. I'm not sure they did it across the board

or just in a subset, but they had their professors

throw away their syllabus and create brand new

ones, so that way they weren't anchoring on.

This is how we have always done it.

How do I tweak it to make it work? But it was truly building a

new set of scaffolding, if you will, or topical areas

based on what they really wanted to do in that course.

That's a great idea. Although I think we still need to teach the mouse

because I just spent like 45 seconds wondering why the touchscreen

on my Mac wasn't working well. But then you use the touchpad, Craig, and not

a mouse. Yes, I know, but it's close enough. Close enough. So

how do you pull this off? You're an administrator now.

What do you do for faculty that encourages them to

experiment. That encourages them to rethink

their entire course. That's a lot of work. Well,

a couple of things that I've done.

Have I convinced people to redevelop their class? I don't think I've succeeded with that.

But one of the things I think is important

is I challenged my faculty to do one thing that

was purposeful, stepping into AI, and to feel like they

could do it without the risk of poor teaching

evaluations, assuming they were purposeful in

learning from what it is that they changed. So creating a

safe space for people to implement change,

I think is important. Now that if you punish

failed experiments, you won't get any change. So we have a question

from the audience. I'm going to answer the other part of that question, and then

we'll get our question. Was there another part of that question? It's a real quick

one. The other thing I would say is we need to encourage

faculty not to be in silos and to get people talking

to each other. So we're not all reinventing the same wheel. Yeah,

absolutely. For both of you. First

of all, I want to congratulate you for jumping onto the

AI train really fast.

I was on camera and you weren't. I don't want to be on camera. Okay,

so congratulations for jumping on the AI train really

early. Now, what about other faculty?

So I know a lot of faculty who are not necessarily as keen

on keeping up with the AI

Everything. How much

of an effort is it? How much should they get involved

if they're not? And is. Is AI going to be one of the other tools

that's just going to become a tool? So are we ready for a

fishing metaphor from somebody who doesn't fish? So have you ever

watched them fish for tuna? They take. They slam

the hook in to a big, big old school of tuna, and they try

to yank a giant tuna fish out. It's

a very violent thing. It's forced. It's really

kind of makes me not want to eat tuna. And then there's

fly fishing. Does anybody fly fish? Fly

fishing? You put the fly out there, and if you have the right fly, it

looks attractive, and you kind of slowly bring it in and

twitch it around, and it's like, oh, that's a bug. I

like bugs. And eventually your trout, I think you fly fish

over trout. I don't know. You know, eventually your trout comes and hits the

lure. If you do it right, and then you reel it in. I think

that's the approach. This is fly fishing. This is not Tuna fishing.

So if you have colleagues like that, show them one or two things

they can do. Creating exam questions is great

because we all hate doing it. AI is pretty good at it.

Especially if you tell them you don't have to take every question AI comes up

with. It'll come up with 10 and two of them are good. But now you've

got two questions you didn't have before. But pick that low hanging pain in the

butt fruit that. Okay,

just do this one thing. That's it. Just do this one

thing. Don't try to drag them kicking and screaming because a. It won't

work and be. It's way too much work for any of us to do. So

just show them, show them. Do a little bit more

and then they'll find something else and they'll do a little bit more and they'll

discover these uses for AI because that's what we did,

you know. Have you heard this question before? I'm going to ask

Hazel. What do you think my very first chat with Chat

GPT was?

Get your mind out of the gutter.

Okay?

Yeah, yeah, sure. So it was to write a

poem about my cat Taz. My cat

Taz is this little gray fuzzball. Everything she does is funny. You've probably known

cats like that. Everything she lays down on something and

it's just funny. And it's not just me, it's not just her father. Everybody

thinks she's funny. So I wrote this. I said, write a poem about

this cat Taz. And it wrote a poem about the cat Taz. But it was

pretty generic, you know, it was kind of. Yeah, it was about a cat, but

it wasn't about Taz. So then I said, oh, she's a gray and white

cat. She's small, she squeaks instead of meows. She likes

to do this. And all of a sudden I got this poem about

Taz. Not about some random cat, about Taz.

That told me you have to have some context and some

specificity when you prompt. I started learning

and then it just went from there. And I think that's what you do. Get

them to do something either fun or

that that is that low hanging fruit that solves some problem that they

just something they just hate doing and let them come along and.

And if they don't, they don't. That's really. That's their

loss. Let them be like that.

So Craig, I think that works for 50,

maybe of people. 75. We can, by

not staying in our silos, we can get people to come along. I think at

some level, administrators need to

at least ensure that every class is AI resilient.

Because if you are cheating through someone's class today, let's say you let

someone cheat. You knew a faculty member just let people copy off everybody's exams and

you knew that we would have those hard conversations and say, your class

is not doing this thing right. We need to

work on fixing that. We need to be looking at every class and making sure

that it's being done and where they get the

AI tools. I'm not convinced it has to happen in every single

class. I think it probably needs to be happening in every single major because

there is some knowledge that learning, it is important to

understanding to be able to use those tools later on, perhaps

in another class. So I think part of it is where is

the tools being learned in the major? But the bigger part

that we need to, whether it's through part of the annual evaluation

process is some sort of a curriculum mapping that helps us understand

that every class

does not let you cheat your way through. And I think that's a different

thing than getting them to use AI. But I

agree with you 100%. They. They're really cheating the entire

institution if they don't at least make their assignments AI

resilient. If you have a question, please come ask us because

otherwise you have to deal with the things that Craig and I think we want

to talk about. Or what are you saying? So we do have another question.

This one is from Rosetta Romano, our president of Sig Ed.

She sent us an email a few days ago

that basically asks who's paying for all this AI?

How many of you are personally paying for one or more AI tools?

Right. Only 41 people raised their hand this time.

Yeah. So a lot of us are. And now some of us have funding

where we can get the tools, but a lot of faculty

don't. And she's concerned about

what happens in a cost constrained institution

around these AI tools because the numbers start to get big pretty quickly.

Rob, you had a. We talked about this last night. You had an interesting

story. So I think this is a really good question. And there's

two sides of this that I want to talk about. So one is

a university that I won't name because I'm going to get some of the details

wrong. I don't need my friends there coming up to me and saying something. But

they started a pilot program that required interested

faculty and it probably included staff to report

monthly about how they were using it so they could learn about use cases along

the way. And so they could see if they were

actually using what was being paid for. Because the fear was, we're going to pay

for these licenses, and then half of them aren't going to be

used anyway, so why are we paying for them? Right? And so then they could

deploy them to somebody else who would use them. And so over the course of

an academic year, they went from about 70, if my recollection of the

story is correct, people in this program to 700. And

with the, you know, reports being they were learning use cases, because that's one

of the big struggles we had. Microsoft came to Washington State University to

teach us how to use Copilot. When I was like, great, they wanted everyone to

go to this training. They invested a lot of money into training. And when I

went and asked them at that time, we were done about use cases,

their answer to me was, you have to figure those out on

your own. Which I'm like, if we're going to bring in products and

encourage people to use them, if we don't actually give a

handful of use cases that are going to increase productivity, then

what sort of a tool are we really giving people?

I think that's a really clever way to do it. Although the

numbers get big, you figured out it's probably $160,000 or

so per year, which a school of

the size of the school that you are not naming can afford,

but a lot of schools can't. That's a chunk of money. You

know, on the other hand, especially for business

faculty, you can probably afford

20 bucks a month if it makes your life easier. There are other

faculty that don't get paid so well, and maybe at smaller

institutions, community colleges, but I don't know. I'm willing to pay

a little bit, but I know that's a privileged position. You can also go

pretty far with the free tools. You can. But here's the caution and here's

why, you know, you call me Mr. Copilot when we talk, because I talk about

it all the time. Copilot has been reviewed by Washington State

University's attorneys and has been put on the acceptable use

matrix. They have verified that it is FERPA

compliant, HIPAA compliant. All of these compliances that we

need to care about. None of the other tools are. Yeah, you

absolutely don't want to share anything personally identifiable about your

students. I wouldn't do it about employees either, and

that's a bad idea. But that still leaves a lot of use cases.

I'm sorry, but every time I've tried to use

Copilot. It has failed miserably.

So I don't know. It's not my favorite tool. But I actually

quit my license to open AI this last month

because they're starting to talk about monetizing

by advertising to me. And it scares the heck out of me with

Gemini, because they're tied into the entire Google ecosystem

that already knows everything about me. I really, really, really think this

whole making money thing is going to be an

interesting conversation for the next year, because nobody

is making money with AI right now. And if we think

back to social media and all these sorts of things and how they make money,

it's based on the data that we give them. So, no, that's absolutely.

It is going to be scary because I was reading about the mental health of

teenagers the other day, and a lot of teenagers are picking up these

apps and they are using it as their counselor, and they are sharing so much

information with them that, I mean, it's actually led to

encouraging some to do some terrible, terrible things to themselves. But

more troubling to me is the machine

now knows even more about these children. We don't

have safeguards and safety trails. In place, and that's a problem. There's no doubt about

that. I do want to plant a seed for some of you. It

is not all that difficult to run some

large language models locally. You get something like LM

Studio, you need at least an okay computer, but you

don't need some $10,000 workstation to run

it. It's something that you can think about. They're not the Frontier

models, but like the one that that OpenAI released.

Their OSS model was kind of about 4.0

GPT. 4.0 in terms of its capabilities. LLAMA

has a model that'll handle a lot of context. They're not as good

as the Frontier models, but they're not bad. Facebook just announced they're going

to quit releasing LLAMA because they want AI that makes them money. I'm

shocked. Well, I've already got it, so. All right.

Any last questions from the audience? Yes, sir.

One question, because you mentioned it, so it came in

my mind about it. There's a huge discussion coming

about companionship, and that's

what they told him. Is it like a trainer, that he is

a light? I got, I don't know, a

personal mentor or something like that. And

you see it also in the academic world. There are people,

professors training

LLM with the rocket system, all the information about

administration, education and so on.

And they just auto reply students and they

just say, okay, I have no time to email every student,

where's the timeline? So it doesn't elevate. So it's

also a kind of companionship, another way around. So

what do you think about AI? So there are two pieces

to that, right? And let me comment first, if you don't mind.

The first piece is the companionship piece. The true

trying to find a substitute for human companionship. And I'm going to put that one

aside for a second. But the second one with

kind of loading up AI with all of the policies and that

kind of thing, I am 100% for that. And

cranky old man would say they're not reading the syllabus, why should I read their

emails? But really I think it's a very efficient way

for them to get a 24, 7 response because

I go to bed ridiculously early and I get cranky students because they

email me at 10 o' clock at night and they didn't hear back until 4

o' clock the next morning. And it's some question that was

actually answered in the learning management system. I think it's great

now, you know, you need to be careful because even with these

retrieval augmented generation systems, they can make mistakes and that

sort of thing. So you have to be a little bit careful around disclaimers and

all of that. But I don't see any problem with that. The

companionship piece, the human companionship.

Human companionship companionship substitution piece.

Dragon's not been drinking. That's. Maybe I should have been.

That's a tougher question. So Rob, what do you have to say about that? So

I think one, there's a really interesting research question in there about

the dark side of is and that is in this world we live in with,

especially since COVID increased amounts of loneliness, the pandemic of loneliness. You

hear people are looking this as a solution to that with

unintended consequences. So I think there's a lot of really interesting things to be

understood because we don't know. And I think that's the whole thing with a

lot of these AI things. I will say it's no different than any other new

technology. When technology comes out, we're very quick to jump

on the benefits of what we can receive from them without fully

understanding and knowing what the downside is or those unintended

consequences. And I do think companionship is one of those places

because we put them right into the midst of human life and meeting people where

they are. It could be good for A certain number of

people that might be a good thing. But how bad is bad for

the people who it ends up being really bad for? I don't know. I know

that's a really tough and interesting question. What I would like to see happen

is some group create

proper guard railed AI companions.

Because if you had to tell me, should

we have AI companions or should we not? I would come down on the side

of we should because I think there's so many lonely people out there

that they're probably being helped more than the number of people

that are being hurt. But I don't have data on that.

We have a question from the audience in.

Response to.

Something that affected civilization, civilization

that drove people apart. Why

should we accept AI as a solution to that problem?

Why not accept it? I mean,

you know that that's a really deep and interesting.

No, it does. And, and I think it would be much better if we address

the underlying cause, but we're not. So,

you know, is it better to at least have a band aid? I don't know.

I mean, I'm. I'm asking the question. I don't really know. And I, I think

we're getting the high sign. You want to. Yeah, we're, we're getting the.

It's time for us to wrap up.

Rob, any last thoughts? I will say again, we need to

understand these technologies, the impact they're having in the classroom,

the impact they're having on our students as they use them, the impact they're having

on society. And I think there's a lot of great opportunities to understand that

and doctoral students who are listening. I think there are some great dissertations

that could be written right now, and I'd be willing to bet that if you

did a good job that mis porterly isr top journals would listen

to what you're doing. Yep. All right, that's it. Thank you

for joining us on the first ever live stream of AI Goes to college. And

thanks again to the AIs Sig Ed folks for

inviting us to do this. Bye.