AI on campus

0:00 I think AI and especially how students use AI,

0:02 it's very telling of those motivations.

0:04 You know, there are some students who are using it to complete work for them,

0:08 you know, to do it on their behalf.

0:09 And there are some students who, you know,

0:10 are staying away from AI, or using it proactively.

0:13 They're using it in ways that reinforce their learning.

0:15 It's our responsibility now as students to, you know,

0:18 use this tool to, you know, achieve your own individual outcomes.

0:23 Everyone is talking about how AI is changing education, but we figured,

0:26 what better way to learn about these changes than by asking actual students?

0:31 My name is Greg; I'm from Anthropic,

0:32 and today I'm joined by four university students

0:34 who are here to give us the inside scoop.

0:36 So, why don't you all introduce yourselves?

0:38 Hey, my name is Zain.

0:39 I'm a final year student at the London School of Economics,

0:42 and I study accounting and finance.

0:44 Hi, my name is Chloe.

0:46 I'm a junior at Princeton, studying psychology and computer science.

0:50 Hi, I'm Marcus.

0:51 I'm a senior at UC Berkeley, studying econ and data science.

0:54 I'm Tino.

0:55 I'm a second-year grad student at the Thunderbird

0:57 School of Global Management at Arizona State University,

1:00 and I'm studying a master's in digital transformation.

1:03 Amazing.

1:04 Thank you for being here.

1:05 Thank you.

1:06 So, let's start by setting the scene.

1:08 What are the vibes like on campus these days with AI?

1:12 How are people thinking about it?

1:14 Yeah, so I did a survey not too long ago on how students are using AI,

1:18 and I saw that, you know,

1:20 90% of students are using AI in their day-to-day workflows,

1:24 using it to summarize lectures, using it to answer problem sets,

1:27 to help give feedback on assignments that they had written.

1:31 And so, really a diverse sort of, like, use case for AI within students.

1:37 It's having an impact.

1:39 Universities are having to manage that.

1:40 We're seeing changes in rules and regulations.

1:42 We're seeing, you know, some courses ban it, other courses encourage it.

1:46 And so, students are in a bit of a gray-zone

1:48 right now where they may not know how to use AI.

1:52 Yeah, I also concur.

1:53 There is a lot of chaos in understanding

1:56 AI and what role it'll play in universities, but at the same time,

2:01 there's a lot of energy surrounding it,

2:03 especially being at Berkeley and being so close to AI hype in the Bay Area.

2:07 I also agree that, like, over 90%,

2:11 if not basically everyone uses AI in some way or form,

2:16 mostly in the form of chatbots,

2:19 yeah, like, summarizing lectures, doing assignments,

2:22 answering questions where or when teachers,

2:27 TAs, like, can't answer them for you.

2:29 I will say there is also a lot of confusion on, like, administrations or, like,

2:35 a professor's end on how AI can play a role in the classroom,

2:40 and we're slowly seeing some changes around that.

2:44 As business students, I see even myself and my colleagues,

2:48 like, we use AI for a lot of different things.

2:50 We use AI to understand and analyze business cases, do market research,

2:56 just come up with financial research, as well,

3:00 so people use that for that as well.

3:02 People also use AI, as well, to complete quizzes,

3:06 you know, like, when you don't have time, 'cause when you're a grad student,

3:09 sometimes you've got multiple jobs that you're working,

3:11 and you don't always have time.

3:13 So, sometimes, you can see someone who just,

3:16 you know, quickly submit answers and everything.

3:18 So, that's the bad side of it that, when you're in grad school,

3:22 you know it's supposed to be, like,

3:24 a time for you to expand your critical thinking,

3:26 a time for you to be someone who is, like,

3:30 more decisive, someone who has, like, substance in how you make decisions.

3:34 And so, that's, I think, the bad side of it.

3:37 Yeah, I would say definitely the vibes are, like, really chaotic right now.

3:41 Both I guess in a good and bad way.

3:43 I think the good, obviously, like, Zain said,

3:44 there's a lot of exploration and cool projects and stuff popping up.

3:48 The bad is because everything is such a gray area.

3:51 It can be very difficult to stay resilient and hold yourself accountable.

3:55 It's very easy to just be like,

3:56 "I'm just gonna give up and feed this all to AI and not do any of the thinking.

4:01 I've noticed that there's a lot of tension, as well, between,

4:03 I guess, like, the line of how much is over-relying on AI,

4:06 or how much is it good to actually have,

4:09 like, an actual cooperation between those two.

4:11 And I have also noticed that some people are really into it,

4:15 so they use it a lot in terms of, like,

4:17 all of their different workflows, while others,

4:19 like my humanities and maybe some social science friends are

4:21 a bit more hesitant and have a bit more concern.

4:24 So, there seems to be a growing, like, identity,

4:27 polarization effect that I think will be really interesting to see how it goes.

4:31 I'm curious, you say that they're hesitant.

4:34 Are they hesitant but still using it pretty regularly,

4:38 or is it a mix of some are using it a lot, some are not using it at all?

4:41 Yeah, great question.

4:42 I think there's a spectrum.

4:44 A lot of, especially, like,

4:45 the pure humanity students have just completely opted out,

4:48 I think, because, often, in their classes and research,

4:51 there's a lot more of just close reading, while other, I think, like,

4:54 for social science, I have noticed a slow trend where they're

4:57 trying it out more and just seeing AI being applied beyond just,

5:01 like, pure computational or, like, machine learning,

5:04 like, context, which has been cool.

5:06 And a lot of computer science and also, like,

5:09 other engineering classes, it's still kind of a taboo to use AI.

5:13 I mean, in application these days, we're using a lot of, like,

5:18 AI coding assistance to build actual projects outside the classroom.

5:22 But in the classroom, we're still using, like,

5:24 VS code and blocking out these AI features because professors,

5:28 at least at the moment, are still kind of discouraging it.

5:31 But we might see a shift in the next few years.

5:34 I mean, I know Stanford is beginning to have

5:36 a course about learning to use AI tools in, like,

5:41 software development and engineering.

5:44 I think that's the number one, I guess, breakthrough with these AI tools is

5:50 that the accessibility and barrier into building something,

5:53 like a project or software in general, has gone down a lot.

5:57 And especially with a lot of courses, like,

5:59 with, like, Claude and, like, the developer docs, for example,

6:03 it's been really helpful in teaching folks who don't come from, like,

6:07 computer science background, like in political science or in, like, psychology,

6:12 or even something like math,

6:14 be able to build their own projects on the side from, like, ideation to, like,

6:20 a working prototype that's on a website or some kind

6:24 of app deployment within the span of, like, a few days.

6:28 Yeah, I've seen that a lot at my university where students who,

6:32 you know, don't typically have the confidence to go and build with, you know,

6:38 raw code have now, you know, started using the terminal,

6:41 for example, which is incredible to see.

6:44 And, you know, Claude Code, for example,

6:45 makes that so much more accessible, so much friendlier,

6:48 which I think has been one of, like, the most crazy changes so far.

6:54 Like, myself even, I don't have a computer science background,

6:57 but I'm comfortable in the terminal now,

6:59 which is crazy, and I've seen it within societies as well.

7:02 So, we have a number of societies at LSE, and they each have,

7:06 like, an Instagram page; pretty basic, easy to put together.

7:10 But now, we're seeing societies have websites,

7:13 and these websites have a load more information,

7:15 and they're building them with Claude Code,

7:16 because it's just so much easier now.

7:18 So, it seems like the AI transformation for students has already happened,

7:22 and we have mixed feelings about it.

7:24 One thing that you all share is that you all are Claude campus ambassadors,

7:28 and you are each leading a student organization

7:31 called the Claude Builder Club on your campus.

7:33 So, first of all, maybe can one of you, like,

7:36 give a quick summary of what it means to be a Claude campus ambassador,

7:39 and then the club that you're leading?

7:42 Yeah, I mean, as Claude campus ambassadors,

7:44 our number one job/role is to be the point of contact of engagement

7:52 between what Anthropic and Claude is offering and between that and students,

7:58 and basically being a facilitator for that on campuses.

8:01 Cool.

8:02 And since it's a club about builders, what are people building?

8:05 What are you seeing happen at your clubs?

8:07 A lot of cool things have been built.

8:09 I'll reference an example from a recent Vibe-a-thon I did.

8:12 I think a lot of the most fun ideas is not the most technically-savvy ones,

8:16 but the ones that really start with human emotion.

8:18 So, one that was really cool was there's called the Princeton Prospect.

8:23 There's, like, kind of a bucket list of things people would like

8:26 to do before they graduate and kind of gamifying that through a leaderboard.

8:30 And the best part of it actually was, the winning team,

8:33 they were just a bunch of freshmen and they were all roommates,

8:36 so they just came into this for fun, and with that human insight where it's able

8:40 to build out something that resonated with everyone,

8:42 and that was something really cool that I enjoyed seeing them build.

8:46 I think one cool tool that my friend and I built

8:48 was this place where you could basically put in your lecture slides,

8:52 and it gives you sort of, like,

8:54 professor annotations down the side of each slide.

8:56 So, it's so cool.

8:57 I've been using it so much for, you know,

9:00 just revising through content in preparation for end-of-term exams,

9:04 and it's so good because it kind of preempts my questions,

9:07 and so I've prompted it such that it knows that I

9:10 want to know the definitions of certain things on the slides.

9:13 The slides can sometimes be a bit abstract and missing context.

9:16 So, adding in the context on the side.

9:18 Did you get a good grade in the class?

9:20 We'll see.

9:25 I think one of my favorite things that someone has built with AI is,

9:29 it's an app called Courseer, and we have this challenge where,

9:33 like, the most, like, amazing fun classes,

9:36 like, when it's time to register for classes, they just run out so quickly,

9:39 and you can, like, wait weeks and weeks to get, like, a seat in that class.

9:44 So, what they did is they built this AI,

9:47 and you can, like, just input the course that you want,

9:50 and then it's gonna alert you the moment a seat is open in that class,

9:54 so you can register for it.

9:55 Oh, I like that at our school.

9:57 Yeah, yeah, instead of you, like, going back and checking every day,

9:59 class search, "Is this class available?"- You

10:00 get a notification that you jump on.

10:02 Yeah, you just jump on and you get a seat.

10:03 Yeah, I love that.

10:04 I need that.

10:05 Your next project idea.

10:06 No, exactly.

10:07 It's actually funny, we have, like,

10:09 a shortage of seats at university, at my university.

10:13 I'm talking about, like, actual seats, like in the library for example.

10:16 And so, again, my friend built this amazing tool that basically scans all

10:20 of all of the data that you can get the data on, you know,

10:23 which classrooms are free.

10:25 And so, it basically points all

10:27 the free classrooms and tells students, you know,

10:28 if there are no seats in the library, then go to these ones.

10:31 And again, like, non-technical student building this, which is insane,

10:33 unheard of, but, you know, these are some of the possibilities.

10:37 Yeah.

10:37 Yeah.

10:38 I've seen in the past few hackathons or entrepreneurship classes

10:44 where a lot of students have been looking into, like, healthcare use cases,

10:48 mixing computer vision with a Claude API to interpret a person's,

10:55 like, emotions for, like, a mental health use-case, like, signs of stroke via,

11:02 like, a camera on, like,

11:04 someone's phone or, like, a separate, like, medical device.

11:07 Or even signs of, like, dementia for example.

11:10 And all of them has been really interesting.

11:14 It's so cool that people are spending their time doing that in school,

11:16 'cause that is kind of the magic of being a student is you do have time to just

11:19 work on ideas and try new things and come

11:22 up with projects that are just for fun.

11:23 They're just the side projects.

11:25 Yeah, absolutely.

11:26 Yeah.

11:26 Cool.

11:26 So, let's talk about learning with AI.

11:29 I think one of the more tricky parts of this is that, you know,

11:34 AI can be a tool to help you learn about anything you wanna learn,

11:37 but it can also be used as a crutch to maybe prevent learning if you lean on it.

11:41 So, I'm curious how you each personally balance

11:45 that and how you see students balancing it,

11:47 and if you see students at your university balancing it well.

11:50 I think initially what we noticed was that even, like, amongst our classmates,

11:56 at first it was just like, whatever the AI gives you, that's what you put.

12:00 And then, it's, over time, attitudes have started to change.

12:03 We're like, "Let's just put a little more effort," and not even just,

12:06 "let's just put effort in what we're putting together."

12:09 Because let's say you have a group project and they're,

12:11 like, four or five people on that group project; everyone gets a different part.

12:15 And if everyone just does the first thing that AI gives them,

12:18 that's not gonna produce a very good project at all.

12:20 I think one thing about AI and education

12:22 is that it's very telling of students' motivations,

12:26 like, why you're at university.

12:27 I think students, you know,

12:29 you can typically group three objectives for university.

12:33 The first I would say is to learn,

12:35 to, you know, to deepen your understanding in your chosen topic.

12:38 I would say a second objective is to, you know,

12:41 position yourself for a career, you know, get a good job.

12:44 And I think the third is the social

12:46 element of university where students are coming to network,

12:49 to have fun, enjoy themselves.

12:50 I think, like, those are the three broad objectives for students,

12:53 and every student weights those differently.

12:56 Like, some students prefer, you know, they're coming to learn,

12:59 and they don't really care about the social aspect of uni,

13:01 and there are other students who,

13:03 you know, they're coming because they want to get a good job,

13:05 and they want to enjoy university,

13:06 and they don't really care about the learning really.

13:09 And I think AI and especially how students

13:11 use AI is very telling of those motivations.

13:14 You know, there are some students who are using it to complete work for them;

13:17 you know, to do it on their behalf.

13:19 And those are typically the students who want

13:20 to save time and want to, you know,

13:22 put their efforts and motivations towards other things, which is fine.

13:26 And there are some students who, you know,

13:27 are staying away from AI, or using it proactively.

13:30 They're using it in ways that reinforce their learning,

13:32 that make them better and make them stronger.

13:34 And those are typically students that, they want to learn themselves,

13:37 they want to, you know, have some depth to their knowledge.

13:40 And so, I think that's what AI is revealing,

13:41 like, why you're really at university,

13:43 because we have the tools now, to be honest,

13:45 to get through university without actually learning much.

13:48 It's our responsibility now as students to, you know,

13:51 use this tool to, you know, achieve your own individual outcomes.

13:55 If you want to learn, you can.

13:57 And if you want to bypass, you know,

13:58 a lot of the exams and assignments, you can pretty much do that.

14:02 And I don't think there's gonna be any sort of, like,

14:04 rules or regulations that come in place that can change how students use AI,

14:09 because, like, fundamentally, I don't see how that would be possible.

14:12 And so, I think the responsibility is in the student's hands;

14:15 it's like, you're in control.

14:17 Yeah, definitely.

14:18 I actually agree, and I think a lot

14:19 of, for how I use and approach AI is, like, intention.

14:22 I think, even before I actually start prompting or asking it to do stuff,

14:26 I like to think about am I asking it to, for example,

14:28 directly complete a task for me?

14:30 Or is it more of, like, something that I'm brainstorming and I'd like

14:33 to think about it from different perspectives?

14:34 And I think that piece is, like, I'm starting to see a lot more happen,

14:38 'cause I think AI is very good as, like,

14:41 a catalyst for especially implementing and building things.

14:44 But the intention, I think, really comes from the students themselves.

14:47 I really resonate with that.

14:48 I think, when these AI chatbots start coming out a few years ago,

14:53 either because of the technical limitations back then,

14:57 or just of how little we understood about AI at the time,

15:02 the typical workflow was just you ask the chatbot a question, you get an answer,

15:06 and you do that maybe, like,

15:08 50 to a hundred times across different conversations.

15:12 Now I think people are becoming smarter and, like you said,

15:15 are becoming more intentional with how they're using it.

15:17 We're starting to have, like,

15:19 more extended conversations across talking about one specific topic.

15:24 I've started, like, when I'm studying, I'll have projects on Claude where I

15:30 would have one for each class; upload, like,

15:33 the syllabus and a bunch of different

15:35 course content I'd take in for each project,

15:37 and have a bunch of conversations acting as, like,

15:41 individual files in, like, a folder for example.

15:44 And with these chatbots being able to, in recent years,

15:48 manage context better, manage memory better,

15:51 be a much more helpful assistant and, I guess,

15:55 conversationalist when, like, working with me on a specific task.

15:59 You wonder how long it'll take before the societal aspect of things

16:02 are gonna catch up to how fast the technology is evolving.

16:07 Right now, with one example, is that in, like,

16:11 CS classes, I know a few professors who do say,

16:15 like, "Hey if you do use AI, like,

16:19 you can put a disclaimer in, like, your assignment and also describe,

16:22 like, how you use it in, like,

16:24 each homework or a lab assignment." But there isn't really,

16:28 like, a integrated, like, framework thinking about, like,

16:32 using AI in the class as part of the curriculum.

16:35 And I think we're still kind of waiting on integrations like that into, like,

16:40 education that we may see in the next, like, five years.

16:44 So, you feel like, in general,

16:46 your professors and the administration might be a little bit

16:49 behind the students in terms of AI literacy and adoption?

16:52 Not mine.

16:52 Yeah, I think they're still adapting to it, and I think, naturally,

16:56 students are more like the fastest

16:58 adopters because we're just reacting to, like,

17:01 what's out there, and we access information a lot quicker because we're,

17:06 like, native to the internet.

17:09 Yeah.

17:10 Yeah.

17:10 I have to say, I've seen some, like,

17:11 pretty cool advancements in some of the courses at my university.

17:16 So, we have a course called LSE 100,

17:18 and every first-year student has to take it.

17:20 And when I did it two years ago now, there was no, I mean,

17:24 we had AI, but there was no guidance on how it should be used for this course.

17:28 My brother now actually is in first year,

17:30 and he's doing the course at LSE, and he's told me it's completely changed.

17:34 So, they basically give you guidance on how to use Claude.

17:37 So, they say you should have a conversation with Claude, give it a persona.

17:42 So, they're giving guidance to students on how to actually use

17:46 these and use Claude for ways that aren't just direct outputs,

17:49 you know, like getting the answers for your problems,

17:51 but actually a conversation with it.

17:53 And then, they ask for the conversation log because they wanna see,

17:57 you know, how are you interacting with it?

17:58 Are you asking, you know, good questions back, and is it a good conversation?

18:03 And then, they film a video instead of putting an essay together.

18:06 So, now it's a video of yourself.

18:07 And so, you are encouraged to use AI, but now in terms of, like,

18:11 the marking, you know, you can't use it irresponsibly.

18:15 I have also noted that for some of my classes.

18:17 Like, the machine learning class I was taking this semester,

18:20 they have their own chatbot actually

18:21 they built to specifically answer student questions,

18:24 and if they wanna refer to lecture notes specifically,

18:26 it's pretty helpful for it.

18:28 I do think, however, that this is more of a bandaid approach

18:31 because it doesn't really prevent students from just going

18:34 to other types of AI tools that is not

18:36 the school one to just ask for answers and advice.

18:40 Yeah.

18:40 Yeah.

18:41 University is a one-size-fits-all route at the moment where,

18:43 you know, you have one lecturer for potentially 200, 300 students in a class,

18:48 and those students all learn, you know, in different ways.

18:52 And so, AI is acting more as a personalized tutor

18:56 if you prompt it in the right way and if you, you know, encourage it to do so.

19:00 And I've seen the learning mode from Claude where,

19:04 you know, it's asking questions back to you.

19:06 It's more of a, like, a progressive development of understanding,

19:09 which is good, and there are students that are using it.

19:13 But I think, you know, it's about finding the students that, you know,

19:17 want to learn and want to progress

19:19 because there are many students that, you know,

19:22 if one AI tool goes away from, like,

19:25 giving direct output or giving direct answers,

19:27 we're gonna see just a shift of students to the other.

19:31 Tino, were you gonna say something about this, by the way?

19:33 Yeah, I think I was gonna piggyback on what Zain said,

19:36 like, 'cause at my school, Arizona State University, we are super pro-AI.

19:41 Our career management center, they built, like,

19:44 a prompt bank for us for prompts that we can use to, you know,

19:48 work through different scenarios and roles.

19:51 They also built, like, for our sustainability class,

19:54 as well, the professor built her own bot, as well,

19:57 and we actually, there's a new class that they introduced

20:00 called Artificial Intelligence Chip Strategy and the Future of Work.

20:04 And it was taught, like, for one semester, but people were like, "Yo,

20:08 we need this class," and now it's taught, like, the whole fall and spring.

20:12 This is all very positive, which is great,

20:15 but I know that it's not all positive, it's not all roses.

20:18 So, I'm curious, what are things that you

20:21 are seeing that are not on the right track,

20:23 or things that you're afraid of, or things that scare you?

20:27 I mean, cheating is, like, the top three use case;

20:29 if not, like, top one in universities without a doubt.

20:31 It just comes from, like, what we discussed.

20:35 You put in a prompt or some input, and chat gives out an output.

20:42 And a lot of students, what they started off doing,

20:45 and a lot of them are still doing,

20:46 it's just taking that output and, you know, submitting it in a cycle.

20:50 I think, I mean, if you look at the interface, it's waiting for a question.

20:54 We are given the questions from the university.

20:57 It's never been easier to take that question and put it into the chatbot,

21:00 and get the mark scheme pretty much.

21:03 And so, it's just so easy to get the answer,

21:06 and you really have to be strong as a student to go

21:09 and work on that problem by yourself and do it yourself.

21:11 Yeah, I think a bit more of a nuanced take.

21:14 I have also noticed that, for even students who are

21:17 using AI to build their own projects and, for example,

21:20 to try out different types of, I guess, technical implementations,

21:24 there's been a really strong sense of ownership shame

21:26 that I've noticed whenever AI even gets mentioned that, "Oh,

21:29 when I was building this project,

21:31 I used AI a little bit," just because, like I said,

21:33 I think the line between of how much the human is using the AI versus

21:37 how much is the AI actually just controlling

21:39 the whole project is very blurry right now.

21:42 So, especially at the vibe-a-thon when I was for, example,

21:44 asking the winners, like,

21:45 "How did you use Claude in your projects?" I had seen a lot of them build out,

21:51 brainstorm, think through, and, like, really iterate with Claude.

21:53 But when I asked them that question,

21:55 a lot of them just defaulted to, "Oh, Claude just, like,

21:58 was very helpful and they did everything," which I think right now,

22:01 like, there's a lack of vocabulary and frameworks to, like,

22:04 regard these types of AI usages, which I also think is what's causing a lot

22:08 of this polarization effect where schools are just either completely banning it,

22:12 but students are still using it regardless,

22:14 hence a lot of the cheating and just, like,

22:16 not really being intentional or using

22:18 their brains when they're interacting with AI,

22:20 which I am a bit skeptical about the direction of this, just

22:24 because I think students are now required to be the resilient ones

22:28 in the age of AI where they really need to be skeptical

22:31 of every single time they use it without guidance from schools and institutions.

22:35 So, I feel like, if institutions or schools

22:38 can't really adapt to this quick enough, there is a danger in it just kind

22:41 of skewing and going into a more polarized direction.

22:45 I will say, though, the sentiment and, like,

22:48 how we interact with AI among students is changing.

22:51 I think, like, as university students, we naturally do want to use our brains

22:55 and use it for something that's interesting to us.

23:00 In the past couple years, yes,

23:02 people have just been pasting in questions as, like,

23:05 prompts and taking the outputs to submit as, like, deliverables or assignments.

23:11 But people are beginning to be more interested in, like,

23:15 doing something more of that; like,

23:17 taking more ownership of maybe their assignments,

23:20 but even more importantly, like,

23:21 I guess projects on the side or things they want to make or explore.

23:27 And I think a lot of students just kind of need

23:28 that little push to see what's available and what's out there.

23:32 And back to the point about cheating,

23:33 I think a lot of students are also realizing that AI is pretty bad at cheating

23:37 in context because there's all these patterns

23:40 that start to come up like, "Oh, there's, like,

23:41 a lot of em dashes," or AI has a specific voice or tone,

23:45 or it doesn't actually understand to the level of what you know about the class,

23:52 which could be a whole conversation about how

23:54 students actually know more than they think they do.

23:57 Yep.

23:59 Okay.

23:59 Yeah.

23:59 Yeah, I agree.

24:00 And I think students are evolving, you know, with AI.

24:03 I think, when it first came out, everyone's very excited.

24:06 Students, you know, were using the outputs directly, but now,

24:09 like Marcus said, you know, people, students are being more, you know,

24:13 intentional with their prompts, so potentially,

24:15 you know, writing a little bit longer prompts;

24:17 you know, directing Claude a little bit better than before.

24:20 And I think that's just because we're getting more used to it.

24:23 Like, myself, as a student, I must have spent,

24:25 like, a thousand plus hours, like, talking to Claude now.

24:28 Like, I know, you know, how it responds, and I'm learning more about the tool,

24:33 and as a result, my interactions with it are getting better.

24:35 And like you said, we're students.

24:37 You know, we want to use our brains.

24:39 The majority of us, you know, want to be intellectually simulated.

24:42 And so, I think we're moving to a time where students

24:47 do genuinely use AI tools to benefit themselves and to actually,

24:52 you know, go further, rather than kind of limit themselves,

24:56 I guess, by just relying on its output.

24:59 Yeah.

25:01 I think, when it comes to, like, cheating, for example, you know,

25:06 you've got that first level of you ask a question, you get your output.

25:10 But in my instance, the final boss is can you present to us what you think?

25:15 You gotta put together a presentation, 10 minutes,

25:18 15 minutes defend your position, and the AI is not gonna be there,

25:22 you know, at that time to speak for you or to give your ideas.

25:26 So, in that way, I feel that there's that, like,

25:29 first level of, like, using it like you mentioned, like, maybe.

25:33 But then, you get to a level where you need to, in our case,

25:38 explain what do you mean and everything.

25:41 So, it's not so much a case of like, yes there's that level of, like,

25:45 people are cheating, like, doing just, like, small quizzes.

25:48 But then, in our instance as well,

25:50 you actually have to always defend your position,

25:53 so you have to know what you're talking about.

25:55 Let's talk about, after college, entering the job market.

25:59 First of all, maybe we can do, like, a thumbs up, down, middle.

26:03 How does everyone feel about getting a job after graduation?

26:09 Like, constantly just like this.

26:11 Okay.

26:12 Okay.

26:12 Okay.

26:13 Tell me more.

26:15 Okay, well, I guess, like, the good ones I think is, like, just having AI to be,

26:20 like, a better, like, companion for, like,

26:22 practicing for interviews, brainstorming, tailoring the resumes, et cetera.

26:27 Unfortunately, the downside is that also companies

26:30 are obviously using AI a lot more, which involves a lot of Hirevues.

26:34 I've basically been talking to a, like,

26:36 a screen this entire recruiting cycle, which is great,

26:40 but also can feel a little less human because I don't feel like there's,

26:44 like, no chemistry, like, talking to a screen.

26:48 Are you doing interviews with like you're talking to a robot?

26:52 Not where it's explicitly, but it's just, like,

26:54 kind of a question on a screen for me,

26:55 and then I'm just, like, talking to myself.

26:58 And I have also heard just a lot of anxiety

27:01 about companies also using AI just to screen candidates.

27:05 And I think this also has just not been great for, I guess,

27:08 like, both my self worth and also just, like,

27:10 trying to figure out what the best, like interviewing strategy or even, like,

27:14 what jobs to apply to, 'cause now it just feels so much more random than before.

27:20 I'm curious, what do you guys think?

27:21 I agree with you, especially, like, the screening job candidates.

27:24 It's so painful because you can realize, like, from the entire, "Hi,

27:29 I would like to invite you to apply

27:31 for this job," right up until you submit your CV.

27:34 You've put time together, tailored your application,

27:36 everything, and then 15 minutes later, "Sorry,

27:38 we regret to inform you." when did you have time, you know, to review.

27:44 Yeah, exactly, the AI-generated email.

27:46 The AI-generated email.

27:47 So, that's, like, I think the really, like, big downside of that.

27:51 The upsides really are that AI fluency has become a major like,

27:56 for example, consulting firms now,

27:58 I know the top four consulting firms, they used to hire generalist MBAs,

28:03 but now they're looking for MBAs who've got AI fluency.

28:06 So, if you understand, like, how do you apply AI to different industries,

28:11 then you're, like, their number one candidate.

28:13 Actually, back to, like, Chloe's point,

28:14 I have had an AI, like, interview me before.

28:17 Really?

28:17 Wow.

28:18 And it was so nice.

28:19 It would give me responses like,

28:21 "Your response was super invigorating and informative and exciting." And then,

28:27 "Let's move on to the next question."- Did you get the job?

28:30 No, but it was because I didn't qualify.

28:35 I think they were looking for, like, rising juniors, and I was a rising senior.

28:39 So, I still got auto screened.

28:42 But it wasn't as bad as I thought, I guess.

28:46 Traditionally, like Chloe said, like, there's Hirevues right now where,

28:50 like, they take a recording rather than, like, an interactive conversation.

28:53 I actually kind of enjoyed having a nice interviewer as an AI.

28:59 I agree.

28:59 I agree.

29:00 Okay, speaking of, you know, interesting uses of AI,

29:05 Merriam Webster named "slop" the word of the year,

29:10 so I'm curious what AI slop means to you all,

29:12 and how do you see it impacting the people around you on campus?

29:17 I think, like, AI slop for me is when I receive an output from Claude

29:22 or any other AI tool that I know that if I had just used my own brain,

29:26 like, I could have come up with something better than that.

29:28 Like, that's kind of slop for me.

29:30 So, I think, just going back to job applications,

29:32 when I'm asking it to, you know, help me write a cover letter, for example,

29:34 which is a major use case for a lot of students,

29:37 and it gives me a cover letter which is so generic;

29:39 like, every other student is applying with this, and it's like,

29:42 "This is not gonna get me the job." Like, that's, you know, the AI slop.

29:45 I think it's really funny that, like,

29:47 AI responses can be so generic that it's its own voice at this point.

29:51 That's, like, a common, like, meme, I guess,

29:53 for AI to have a lot of M-dashes and certain sound bites.

29:57 Like, you're absolutely right, or like,

30:00 "Let me think about that." Or it has this, like,

30:03 two-sentence structure that it keeps giving me whenever I try to write, like,

30:07 letters or scripts, for example,

30:10 where it's like, "You're not reinventing the wheel.

30:12 Like, you're building the next Tesla."- Yeah.

30:16 Yeah.

30:17 Honestly, it's everything you guys have said.

30:19 Yeah, and then, like, you get the feedback,

30:22 you get the output, and then it's up to you.

30:25 You know, some people, if you work with them in a group,

30:28 sadly they'll just paste that, and you could see that at the end,

30:32 "Would you like Claude to keep" you know.

30:35 Oh.

30:35 Yeah, that's-- Claude can make mistakes.

30:37 Retry.

30:37 Claude can make mistakes.

30:38 Retry.

30:39 Yeah, that's my definition of AI slop.

30:41 So, you mentioned group projects, and I think this is a big thing, right?

30:44 When you have a group of four or five at university,

30:47 and you have maybe a 5,000 word report due, how do you guys go about it?

30:51 Because at my university, there is sometimes some students who, like,

30:56 don't want to use it, and I remember, like, one student was saying, like,

30:59 "I'm gonna do this project before you guys get

31:01 your grubby AI hands on it." And I was like,

31:04 "Okay." But, like, some students really-- Did he use that term?

31:08 Grubby AI hands.

31:09 Grubby AI hands.

31:10 Oh, man.

31:10 Oh my god.

31:11 Some students feel really strongly against it,

31:12 and when you're working in a group,

31:13 you know, you have to take into consideration other people's thoughts.

31:16 True.

31:17 Yeah, what are your guys' thoughts on that?

31:20 I can go first, 'cause we do a lot of those 5,000 words kind of projects,

31:24 like maybe create a business case out of this business dilemma.

31:28 And how we do it, like, how we've recently started working on it is

31:31 we'll take the paper or the question and we'll, like, create an outline.

31:35 We'll maybe ask AI, "Can you create an outline for this paper for me?" Like,

31:38 "What should be in this paper" and stuff.

31:41 And then, we divide it amongst ourselves.

31:43 One thing I like doing a lot is, yes, using that outline and for, like,

31:48 this example of, like, a 5,000 word report amongst,

31:50 like, four people split it into different sections,

31:52 and then for each person covering each section,

31:57 it's up to you and how you want to use AI, whether you use it or not at all.

32:03 And what I like to do personally is have a lot of, like, bullet points or just,

32:09 like, thought-dumping into Claude and working

32:11 with it to kind of structure my thoughts.

32:13 So, going from random, like,

32:15 bullet points or one-off phrases into more of an outline,

32:20 and then into paragraphs that I can kind of manually edit

32:23 the wording of so it's more like my voice and tone.

32:27 And then, one thing I really like asking Claude actually is

32:32 to give the context of who is usually reviewing my work.

32:37 For a job application, for example, it's, like,

32:40 this VP or, like, recruiter, and then, like,

32:44 in a class it's like a professor or a TA,

32:46 and I ask, like, "Hey, here are some criteria.

32:51 Rate my work score out of 10." And I would do that maybe,

32:56 like, two to three times, and it would always gimme reasons about, like,

32:59 why it graded, gave me a score, a certain score.

33:03 That's a good idea.

33:04 And what I could work and improve on.

33:06 A lot of times, I like the feedback.

33:08 Sometimes I think some of the feedback is a bit overzealous or ridiculous.

33:14 And in newer models, like Sonnet and Opus 4.5, they're starting to give, like,

33:22 a bit of urgency whenever I ask it to evaluate my work too much,

33:26 almost like they're calling me out for overthinking.

33:30 After maybe, like, the third try of, like, this, like,

33:32 evaluation, they'll be like, "It's ready to ship."- Yeah.

33:35 Yeah.

33:36 Nice.

33:36 Nice.

33:36 Are there AI slackers in group projects?

33:39 Like, people who you can tell are not turning on their brains for the project?

33:42 With their grubby AI hands.

33:44 I mean, definitely.

33:45 I think something that is most helpful for me is,

33:48 like, obviously besides alignment and just,

33:50 like, being very intentional when you're using AI,

33:53 a lot of face-to-face time actually.

33:55 So, what I like to do when I work

33:57 on a group project is just block out a time chunk,

33:59 sit down with my group, and we just talk about it as we work through it.

34:03 I think, very often, it's easy to feel like you're alone when you're

34:06 just working on a group project on your self,

34:08 which is what makes AI so tempting, 'cause you're just like, "Oh,

34:11 what if I just had someone write it for me?"

34:13 But if we were all forced to, let's say, sit down and talk together,

34:16 like if someone had a problem and work through it,

34:18 I think that definitely helps a lot

34:20 with the more human piece of working together.

34:22 Yeah,- I agree.

34:24 Okay, I'm gonna shift us to some rapid-fire questions.

34:26 So, each should be, like, one to two sentences maximum.

34:30 So, my first question is, what is a tip that you have for students right

34:35 now who are navigating this whole world of AI in education?

34:39 Learn it.

34:39 Learn it.

34:40 Learn how to use it.

34:41 It's only to your advantage if you understand how it can optimize your career,

34:46 or if you decide to be an entrepreneur, how it can optimize your business.

34:51 If you're trying to learn new concepts or revising for exam,

34:55 start a new project for every class you're taking in university.

35:01 Try and paste in all the relevant files, and perhaps you already have existing

35:06 conversations where you've worked with Claude

35:08 to go through certain assignments and set the writing style to concise mode.

35:15 That's been most helpful for me to get

35:18 a quick rundown in an efficient manner of, like,

35:22 every concept I need to cover for an exam.

35:25 Substack and open-source materials.

35:27 There are so many cool people out there who know

35:30 the best or newest ways to use different types of AI tools,

35:33 and what I've found most helpful is just soaking that up like a sponge,

35:36 and then applying it to my own projects.

35:39 Nate Jones on Substack.

35:40 He's pretty good.

35:42 My tip would be use the styles.

35:44 So, you mentioned it briefly, the concise mode.

35:47 The learning mode is fantastic.

35:48 If you want to augment your own brain and augment your own skills,

35:52 use the learning mode.

35:54 It will ask you questions back.

35:55 Be confident in your replies back, and you genuinely will get a better

35:59 output than just leaning on Claude by itself.

36:02 All right.

36:03 Next question.

36:04 How do you, in one sentence, personally draw the line?

36:08 How do you draw the line between using AI as a tool and using AI as a crutch?

36:13 Where do you find that balance?

36:16 If I was in a room like this, and I can't explain or defend what I've built,

36:22 even if someone asks, like, a super critical or specific question,

36:25 I think that's the line where you

36:28 kind of don't really understand what's going on.

36:31 I totally resonate with that.

36:31 It's a mix of, like, the ownership and intentionality.

36:35 If you can't really explain what you've done along with also

36:41 including what AI's role was in your work or what you're doing,

36:44 then that's a line for me.

36:46 Yeah.

36:46 That's another line for me as well.

36:48 Like, I should be able to explain it

36:50 like I'm explaining to someone in fifth grade, whatever the output is,

36:55 and I should be able to present it as well,

36:57 even at a graduate level, anything that I prepared.

37:00 So, that's my line.

37:01 Anything I create with AI,

37:03 I should be able to give that lower level and that upper level explanation.

37:09 Yeah, I agree with all of you.

37:10 I think, if you're not comfortable with the content that you've produced,

37:14 at the end of the day,

37:15 like, is that really yours or are you just stealing that content from Claude?

37:20 And so, just feeling comfortable, having some sort of, like,

37:24 feeling of ownership that I've produced this work, that's the line for me.

37:28 You know, there have been times where

37:31 I've submitted pieces which are, like, fully AI,

37:33 and it's just like, this is not gonna take me anywhere at the end of the day.

37:38 But you learn that, and I think that's the biggest thing

37:40 with students is that it takes time to learn those feelings,

37:43 and you kind of have to give it that time.

37:45 Like, a student might have to submit

37:47 something 100% AI to realize that, actually, this was not beneficial for me.

37:52 And I think universities need to be

37:53 conscious of the fact that students will learn,

37:55 and you've gotta trust the students, right?

37:57 At the end of the day, they live their own lives,

38:01 and, you know, you wanna set yourself up.

38:03 You have that equality between students, and we'll figure it out.

38:07 Like, we'll figure out what works, where it's good, where it's not.

38:10 Yeah, like holding space.

38:12 Exactly.

38:13 I feel like that's a fantastic place to end.

38:14 I just wanted to say, you know, that "We'll figure it out" mentality.

38:17 This whole time, I kind of expected this conversation to shift into doomerism,

38:22 and it never quite did.

38:23 I think, like, all of you are, you know,

38:26 thoughtfully positive about the future in a way that, I think,

38:30 is really exciting and really encouraging.

38:33 So, thank you all for being here,

38:35 for being honest, and yeah, I really appreciated this conversation.

38:39 Thank you.

38:40 Thank you for having us.