WHOOP Podcast - The Hidden Cost of AI: Creativity and the Future of Human Work with AI Philosopher Aleksandra Przegalińska

Episode Date: October 7, 2026

On this week’s episode of the WHOOP Podcast, WHOOP SVP of Research, Algorithms and Data, Emily Capodilupo sits down with Philosopher and AI Researcher Aleksandra Przegalińska to explore what happens when humans and AI collaborate. Drawing on her research on human-AI interaction, Przegalińska examines how AI can reshape productivity, creativity, education, and the future of work. This episode addresses real concerns over AI establishing diminished creative ownership, growing sameness of content, and skill atrophy in the way we work and communicate. Przegalińska illustrates the human-AI connection as something that has the potential to grow everyday capabilities, but warns that it doesn't necessarily mean less work or stress. The conversation makes a case for the use of AI by keeping humans at the center through thoughtful delegation of tasks, allowing organizations to embrace AI while keeping that important human touch. (00:39) Aleksandra Przegalińska and the Philosophy of AI(01:54) Defining the Human AI Interaction(05:42) Research on The Human Experience of AI(07:37) Trust, Rebellion, and Adopting AI(12:11) The Productivity Paradox of AI(16:08) How To Keep Human Joy In Your Workflows (17:16) The Founder Study: Ideation vs Execution with AI(19:39) Researching AI Development and Pace of Adoption(22:44) The Skills That Need To Stay Human(25:23) AI Rebellion: Why Teens Are Pushing Back Against AI(30:03) Writing with AI: Concerns About Authenticity(32:41) AI Use Case for Non-Native English Speakers(38:52) Importance of Human Judgement and Review of AI (41:05) Advice For Leaders: Is AI Beneficial For Growth? Support the showFollow WHOOP:Sign up for WHOOP Advanced LabsTrial WHOOP for Freewww.whoop.comInstagramTikTokYouTubeXFacebookLinkedInFollow Will Ahmed:InstagramXLinkedInFollow Kristen Holmes:InstagramLinkedInBuy Aligned by Dr. Kristen Holmes: Follow Emily Capodilupo:LinkedIn 

Why is it important to study whether or not humans like AI? So we're trying to understand better how this future of work with AI will unfold. Are people stress when they talk with AI? Are they happy? Do they feel overwhelmed in any way because of the interaction? We're not so much focused on AI itself. We're most interested in how people perceive this technology and how they discover different ways of working with it. Because ultimately, you know, judgment, selection, decision, these are very human things. And I believe that that part actually judgment and decisions should stay with humans.

What is it that worries you? Like, what's the danger? That's a great question. Hi, everybody. I'm Emily Capitaluppo, Whoop Senior Vice President of Research Algorithms and Data. And today I am joined by the absolutely incredible Dr. Alexandra Pshagalinska. Thank you so much for being here with us today. I'm very happy to be here, and I love how you just pronounce my surname so correctly.

beautiful Polish accent too. I do my best to pretend to be. So Dr. Shagalinska and I met, I guess about almost a year ago now, at a conference. And I was just so inspired by everything that you had to say. And then we remet again at another conference a couple weeks ago. And I was just like, okay, we absolutely need to get you on the podcast because you are, among other things, one of the leading voices in Europe on AI. and the future work and wearables.

And this whole idea that I think does not get enough airtime, which is like human AI interaction, you have this very techno-optimist, but mixed with like reality kind of perspective on how AI is going to totally change our lives and the world. And it just felt like a perspective that more people needed to hear. So thank you so much for being here. I'm very happy to be here.

Let's break down some of those terms. It's like, what does it mean to be working on these sort of human AI interaction? Like, what does that actually look like? That's a great question. And I often wonder myself how to put it in words properly. But my team, also back home, so the team here at Harvard, but also back home in Poland, we've always researched, you know, different ways of communicating with AI. It started with virtual assistance, you know, when they were certainly not as good as they are right now.

we actually started 14 years ago, which I think makes us some sort of dinosaurs in the AI space really way before the hype. And we were very interested in how those interactions really work for humans and what are the outcomes for humans? Are people stress when they talk with AI? Are they happy? Do they feel overwhelmed in any way because of the interaction? Does AI lack anything? Is there a lack of expertise, are there any limitations that prevent this conversation from having some sort of natural flow? So these were the questions that we were really interested in. And then that's sort of, you know, multiplied into different areas because currently we're very interested in human AI collaborations. So how does it feel to work with AI? What are the best modalities of working with

AI? What is it good at, you know, as your teammate, collaborator, et cetera? So you could say, I guess it's a very interdisciplinary field of inquiry. Some of the questions have a psychological angle. So they're about your state, your emotions, et cetera. Some of the questions are about your productivity, right? So they're more focused on management. Some of them are, I would say, closer to labor economics. So we're trying to understand better how this future of work with AI will unfold and you know, which professions stay and which would go away. So there are many different questions that we're trying to ask, but indeed we're not so much focused on AI itself and the technological or technical side of AI. We're most so interested in how people perceive this technology

and how they discover different ways of working with it. Yeah, because you're a philosopher, right? And so... That's my background, you know, more of a data scientist, I guess, right now. But yes, I'm trying to keep that philosophical angle, you know, in our inquiry. How are you merging data science, AI philosophy, what does the composition of your research group look like? What are the backgrounds? Because you didn't go to school for future of AI. No, I guess maybe such schools will open now, you know, maybe that's a new field, a new discipline that will develop itself with some sort of rigor. But certainly all of us researching human AI interaction, the future of work, we come from other fields. And in my case, I have studied philosophy. And I did my PhD actually, and I would say something

that we could call philosophy of AI, because my main research question was about the Turing test, which is now obsolete because AI speaks so well that it could well pretend to be, you know, a human. So many people, I think, talking to AI in some sort of blind sample would not be able to even say that they're interacting with AI because it can have or hold such a good, conversation that's very hard to recognize. So yeah, that's been sort of, you know, my main research question at some point. Then I came here. Actually, I came to Boston, to Cambridge for the first time, and I did my postdoc at MIT at the Center for Collective Intelligence. And I think it was very helpful because it added more of, I would say, methodological rigor to my studies. I

dwelled into machine learning, natural language processing. So then I could ask new types of questions, thanks to these methods. And indeed, we've been conducting various types of experiments. Actually, also experiments focused on the use of wearable technologies. I'm happy to speak about it later. But in those experiments, what we were trying usually to detect is, you know, okay, let's have a conversation between yourself and AI. How about you and AI do a task together?

And then we're going to look at the division of labor, who did what, for instance. And then we're going to hook you also to some electrodes. I know it sounds scary, but just to passively record your affect. So your emotions, your stress levels, your frustration. We can even hook the electrodes to the face. And then we can detect frowning and smiling, right, which are very often involuntary reactions to something happening. So we've been trying to integrate multiple different methods, surveys,

declarative stuff with objective output, like, okay, how did the productivity look like? What did you achieve with and without artificial intelligence? Which group did better, the one that worked with AI or the one that worked by themselves? And then on top of that, we would add the stuff that I know WOOP is very much exploring, which is wearable technologies, electrodes, different types of sensors that can help us in figuring out how people felt when it was going on. Because very often, and I think this is very important to mention,

the declarative data, so what people say, you know, how they felt and, you know, how did it feel for them to kind of go through this experiment is very different from what the body signals tell us. So, you know, it's a different story that the body tells and a different story what people essentially report very often in a post survey or something like that. So that, I think, makes for an interesting combination. And this is something we've been, you know, doing for a couple of years now. And for anyone listening where this might not be obvious to them yet, like, why does it matter if people like working with AI, right? In a capitalist society, if AI is more productive than humans, right, AI is going to be used,

right? There's all this economic pressure to improve margins or drive down costs or however you want to think about that. Why is it important to study whether or not humans like AI? Oh, I think it's a great question. especially because AI, to some extent, has been also failing. For the past couple of years, I think we saw, you know, AI rising in organizations being implemented. And in some processes, it went really well, but in some others, there were some very serious bottlenecks.

And those bottlenecks are very often associated with, you know, that very, I would say, interesting, but also problematic combination, which is humans working with artificial intelligence or supervising artificial. official intelligence or delegating stuff to AI. So humans are usually the ones that are held accountable. You know, this is also the system that we live in. Obviously, you mentioned capitalism, but the other thing is obviously the legal framework where you are the one delivering the task

and you can use technologies for it, but you are held responsible for it. And that is something that I think makes this collaboration between humans and AI difficult and interesting at the same time because we still hire humans, thankfully, and very often we expect those humans to work efficiently with artificial intelligence, and I think for different tasks,

this efficiency will look differently. What I think is quite perplexing, at least to me, I had that thought recently, is that we very often use, or on a daily basis, we use LLMs, right? Language models and different agents that are built on top of LLMs. And now they're so good at many different things,

reporting, you know, going through some databases, coming back with insights. But they're actually not so good with writing. When you think about it, that was their first purpose. But when I see an AI generated text, I kind of immediately detect it. Or very often I would be able to detect it if it's not edited. And I don't consider it to be a compelling thing to read. So it's kind of interesting how we are trying to spot the best uses for artificial intelligence and also those areas where we would like to keep more human intervention, right?

And I think that is something that makes this human AI collaboration and figuring it out very important. And a big part of it is how people feel about it. Because if people don't like AI, if they do not trust AI, if they reject this technology, which often happens, even if they're very skillful, they may still not trust it and then feel sort of unhappy about it or fearful. that is going to impact the trajectory of their individual productivity with AI and can translate also into big outcomes for the organization. So I actually had a PhD student, Conrad, who studied rebellion against AI and organizations. There are organizations where many people feel like they would like to pioneer this space and they feel very enthusiastic about AI, but there's also this other group, which is not necessarily ignorant.

It can be a group that knows a lot about AI, but because of their perception of this technology, fears related to the job market or privacy and many other things, they just don't want to use it. So they can hijack essentially the whole process. So I think kind of understanding that effective emotional layer of interacting with AI, is it something that soothes you, makes you feel happy because you've achieved something faster,

Or is it something that actually makes you feel very overwhelmed and generate some sort of brain fry? This is something we want to find out more about. So you touched on so many different interesting things. I've definitely read lots of sort of anecdotes and articles about people intentionally sabotaging corporate AI initiatives because of fears of job loss and that, you know, some companies are more overt than others when their intention is, you know, to train AI to take your job. jobs, people aren't super excited to help that go even faster. And then there's sort of all sorts of versions of that story, like people going out in the streets and like throwing rocks at Waymo's because that's the own version of that. Another thing you said that I think has been a really interesting story and one I have

definitely felt personally, which is, you know, there's all this, there was a lot of rhetoric like six months ago when agents started getting, you know, good, that, you know, this is the beginning of the four-day work week or, you know, all this time we're going to get back. And I think anybody who's deeply using AI has experienced so much the opposite where you feel so capable of doing so much so quickly that, like, an hour of inactivity feels so much more, like, guilt-inducing than it ever used to. And I just feel like, oh, if I just kick off this agent before bed, if I just do this, if I just do this, I can have all these agents doing this stuff and working for me all night. And it, like, I feel like I've never been more of an I-C, you know,

at least not for several years, never worked as hard as I've been working. And it's a mixture of personal projects. I've automated a ton of stuff at home. And of course, work projects as well. But it's somehow, like, the more powerful I've become, the more I can do, the more I feel like I should be doing. Whereas, like, in sort of our pre-AI at work kind of moments, things needed to be more collaborative. And so at some point I'd just need another human to weigh in. And it would be, you know, eight o'clock at night. And so I would wait for another human to weigh in tomorrow.

And that would be the end of my, you know, working on something. Whereas now with AI I can, like, do all those things by myself. So I'm never waiting on anybody else. But then it's like, okay, so I can just keep going. And then I find that I have to like, you know, really stop myself and like, go to bed. but I hear this from a lot of people that like when what you're capable of is greater, it feels like there's more to do. And I think one of the things that's where those two points are maybe related and what

people get wrong about AI job loss is that it assumes that there's some kind of like finite bucket of work and like the more we delegate to AI, like the less there is for humans. but I actually think there is an infinite backlog of work in our philosophy here at Woop, which is why we're hiring so aggressively right now instead of firing like so many other companies are. Congratulations. Thank you. We're hiring 600 people this year, so if you want to be one of them, check out our career page,

whoop.com slash careers. But we are, we believe that there's never been like a higher ROI on a human. I love that. Thank you. But this is your podcast, not mine. So I would love to kind of hear about from, you know, the research perspective instead of just my lived perspective, what are people not getting about human AI collaboration and what are you seeing? Well, I think a lot of the things that you just

mentioned are very important. So the fact that we have this obligation to constantly prototype new things, right? So with AI, it's possible to start working on this application or whatever functionality that I've always had somewhere in the back of my mind, but I never had time for it. And now I'm thinking, oh, it's actually doable. I'm going to have those agents do it. And then I'll ask them to kind of do that. And then I'll ask them to get something else. And then we're going to try and improve it.

So I think a lot of that does contribute to this brain fry. And I think people do not appreciate how difficult it can be for them. So kind of really figuring out how much you want to push yourself with this technology and and striking the right balance, I think, is very hard now. So I have those friends who are completely obsessed with AI agents, and it feels a bit like an addiction, frankly speaking. There are good sides to it in terms of effects, but I am not sure about themselves, how did they feel about it?

Some people do tell me that I feel like micromanagers of artificial intelligence, constantly talking to AI. And then, you know, I think what people, if they're so immersed in it, and maybe to forget is like the joy of creation. So I have to tell you that I obviously use AI on a daily basis, and I think it's an incredible tool for research as well. But I did discover joy in writing by myself, especially, you know, I don't know, for LinkedIn and other social media.

I just feel I want to re-find, try and find again my own voice somehow, and that it's very important that it's my own flow. And actually, that is something we have researched. because we've looked into how people kind of how invested they are in something if they write by themselves. And it turned out that the fact that you've written something reignites that need to write even more. So it opens that state of flow that is hard to achieve when you just delegated something to AI and you read it. If you read it, you're much less invested in it and it doesn't feel like it's your own. So there are certain areas, I think, where we would like to keep that part, where you want to keep that flow,

you want to be re-engaged in something that you've created by yourself. Another example is a research we've done recently, and it's also about that proximity to AI, how to regulate, you know, how closely you want to work with that technology, how much you want to involve yourself in it, et cetera. Actually, it was a research with startup founders in Germany and in the Netherlands. It was with our partners from Groningen, University in the Netherlands. and we had 200 startup founders come up with different startup ideas, select the best one, and then come up with a plan for implementation.

So there were three stages. So the first one was the kind of draft different ideas, and they had actually two conditions. One group was working very closely with AI, and the other one would only use AI every now and then in more assistive capacity. So you have two different variations, and what happened? So for the group that worked very heavily with artificial intelligence, they came up with pretty decent ideas. But then the idea they've selected, because they were relying on AI very heavily, was not the best one. According to the judges, you know, evaluation.

So we hired independent judges and they said, oh, this is not the out-of-the-box idea. For the other group that worked more so by themselves, they've selected far better, more interesting and surprising ideas. Now, when it comes to implementation, the group that worked with AI did so much better, right? And I think it shows you quite a lot about like how much there is still to discover and, you know, how to regulate that space between artificial intelligence and yourself not to feel so overwhelmed, but then really, on the other hand, to max out on this technology to maximize its potential. And I think it's just, you know, that's one of the things that I think many people maybe don't get wrong, but they just don't spend time thinking about that.

You know, where is the space for me? And where is that space for that technology? And what's my desired outcome? Which parts do I really want to leave for myself? They just tend to kind of push everything to this technology because maybe they think it's an oracle of sorts, you know, that would do everything better either way. But it's not true.

According to the research, there are still areas where we do as humans so much better. So I think this is a misconception to me that some people feel like, oh, AI is this end game. and it knows better about everything. So why don't I just submit myself to it? I would like to avoid that. So AI is moving incredibly quickly. Yeah.

And so how do you reconcile the fact that, you know, every two weeks there's a new best frontier model with sort of doing research that almost by definition requires that you freeze a point in time and evaluate the state of the art? Oh, that's a great question. That's a great question. And I don't think it's asked enough. It's a question for all of us, essentially, you know, scientists, but not only people who are trying to grasp what's going on with AI.

It's so difficult. But I think there are certain things that are sort of universal, and particularly if you focus on the humans, right? I mean, feeling overwhelmed by working with AI or feeling joyful, right, or trust towards this technology. These are certain things that obviously need to be recalibrated and adapted to the new frontier stuff and what's going on in AI. but on the other hand, they're sort of universal questions. So I think even if it's a research that's been carried out, say, six months ago, it to a large extent is still valid. We also did another piece of research.

I think I mentioned that to you at some point about toxic artificial intelligence. We might get there. But I think, you know, so sort of, you know, AI not being helpful to you is a very important question. And it doesn't matter if it's just a persona, based on an LLM or an agent. It's still the case that if it's not going to help you, certain effects will, you know, trigger themselves in you, and you will respond in a particular way.

So we're trying to focus on the humans, because humans obviously will react to new stuff in maybe slightly different way, but then there will be also some areas that are always important. Like, for instance, trusting the technology, adversarial effects of the technology, feeling overwhelmed. So I'm just trying to think that, you know, since we're focusing on the humans, maybe what we do is more important and can tell us a bit more also about the technology

that's emerging, not only the technology that we've been experiencing before. One of the things that you've touched on a couple times now that I think is really worth digging into is like finding the stuff that like brings you joy and not like sort of delegating that to AI or finding the things that AI is not ready for. and not, you know, giving too much trust to the AI that you're just blindly delegating everything. But I'm curious about the stuff that maybe you enjoy giving to AI and AI is actually good at, but where you might create a risk of deskilling in humans or skill atrophy, however you want to phrase that. One of the things that I think about a lot as a parent, and I know you are a mom as well,

is like there are certain advantages that we have, you know, having learned to write with pencil and paper and no AI assistance. And so now like we can look at AI slop and sort of know how to write something better. But if let's say like AI did do certain things well. And so we just, you know, gave all those tasks to AI. Like what are we risking in not. exercising that muscle. Where do you get nervous about like your own children and like my children, like never learning to do stuff because just AI can do it really well today? And how should we think about the value of like almost like maintaining skills for like the art of it given that maybe

objectively we actually don't need to know how to do them anymore? Yeah. Well, I often think about languages and learning languages, which may not be a necessary skill at all, you know, taking into account the criterion of communication, because you may have devices that will translate everything well enough and you don't have to do it. But then for yourself, you know, it opens up so many, you know, areas and I think concepts and experiences that it's still worthwhile to learn new languages. I think there will be many similar things happening in the future in the sense that there will be skills that we will practice because we want to develop ourselves, not because it's an external necessity. And it's a good thing. Actually, my favorite book is the

Dune. And in the Dune, if you remember, you know, people, well, there was this battle with AI that people thankfully somehow won. It's not explained how. But then they've decided that it's important to develop themselves, like practice multiple skills, you know, and try to go deeper into their own experiences. And I think there's something extremely compelling about that vision that in a world that many things, or many things can be done without you, you would still try and learn and practice certain skills just for the sake of it. A good example is also, I would say, sports and fitness, right? So now it's an area where people, well, they treat it with care because they care about themselves. It's not so that they have to be fit because they're going to go and fight, right?

It's more so that they want to be fit, to be healthy, you know, and to be in a good shape and to experience beautiful things thanks to the effort they put into it, not because there is an external necessity to it. So that's how I think about it. And that's why I'm very often prone to think that maybe particularly early school should be without this technology so that you can just go out and practice multiple things before AI comes and shows you, hey, it's also doable with AI, and you don't necessarily have to put in the effort. My daughter is 13, and she is actually very rebellious against artificial intelligence. I can have that sensation, you know, we recently experienced AI booing at commencement speeches

and everything, that there is, how should I put it, there's like, there's this moment in culture now where teenagers, young adults, generation alpha, generation Z. they rebel against this technology, but the rebellion is far broader. It's a rebellion against the system. AI seems to be that thing that they rebel against, but it kind of, you know, encapsulates multiple different phenomena. So fear about the future geopolitical tensions, uncertainty, also on the job market, et cetera, and I think AI channels that.

So my daughter is kind of in a healthy way quite rebellious. I still encourage her to kind of try and experience this technology in ways that she find suitable, but I think her generation is, you know, they have their own stance. And maybe it's a good thing, you know, it develops critical thinking for sure. So that that is something that I don't think is a bad thing. Some people are like, oh, you know, Generation Z rebels against AI, they don't understand their reality. I don't think so. I think they just have their own reading of it and they will have to carve out their own path with this technology, which is here to stay, but maybe they will just set up different conditions for it.

I'm very open to, you know, and receptive to listening what they have to say about it. And I think, you know, it's altogether a good process, right? To rebel against something and then try and approach it in a different way. But I did have today this, you know, before I came here, we had an examination, defense, actually. So the diploma, you know, paper defense of our students in Poland. and I was remotely participating. And we've noticed, actually, with my colleague who was also on the examination committee, that they tend to respond to the questions, you know, that we have for them.

And they had those questions prior on a list in very similar ways. And these were good responses. It's just that they haven't really looked for these responses. They asked AI for the responses because the phrasing was sort of similar. And it's not necessarily a bad. thing. They've used AI. I'm assuming they double-checked whether it's accurate, you know, their response, et cetera, but still they've used particular type of phrasing. And there is this, you know, there is this argument of an omni writer that, you know, we might end up speaking in the same way

because we overuse AI and we will shut down different styles of speaking and maybe also thinking about things because AI is that first answer provider. So we just spend a bit of time today, I think thinking about it. And at some point, we've asked one of the students, you know, after the examination, she was not stressed anymore. But we were like, okay, so how did you prepare for this exam? And she said, oh, I've used a couple of models and I compared the responses. And when I felt it maybe was off, I would try and check.

But she did not deny it. She said it was her source of information for good, well-encapsulated the responses. But then it's the 15th person that she's here saying things in a a very particular way. So that to me is something that worries me a bit, you know, that pervasive style of communication that AI sets up. And dig into that. What is it that worries you? Like, because my guess is the students also refined their thinking quickly and had a feedback buddy. And like maybe there was some good stuff that happened there too, right? And it doesn't sound like you would classify any of this as cheating or problematic.

No, they prepared, just in a different way. Yeah, different than it would have looked a couple years ago. And so what is the consequence we should be looking out for here? So we all start using similar rhetorical patterns. Yeah. Does that sort of streamline things in a nice way? What's the danger? Well, I think there has to be sort of the next step to it.

There has to be some sort of process that we haven't quite figured out where it's an okay to the fact that you've prepared things with AI and then there is an end that says, okay, and have you questioned the AI? Or maybe you tried, you know, answering yourself and then compare the responses. I really don't have the answer quite yet, but it just feels intuitively that there has to be some sort of continuation to that first step, because otherwise we will get to. stuck with it. And I remember very recently a Polish Nobel Prize laureate in literature. She admitted that she's used artificial intelligence for her next novel. It was a huge scandal, not only in Poland, but also beyond. And I was just trying to understand why people are so mad about her, you know, for using artificial intelligence. She just openly admitted that she's exploring this technology and she even feels it's an obligation for a writer to understand better, which resists.

with me. But then I started reading the comments, you know, to her post about it. And many people were like, well, first of all, they felt maybe cheated that it's not her. Like, you know, it's not her effort. And they felt there was something off about it. But the other thing was also, oh, so now everybody's going to write with AI and all these books will be similar, you know, which doesn't have to be the case. If you edit a lot and you put a lot of yourself into that writing, I still think AI can be a useful tool, but just like it feels like people were very suspicious about striking the right balance and maybe that those authors would fall into that rabbit hole of overusing AI and wouldn't give them their voice. That I think is a bit of a threat, you know, especially when

you, yeah, I guess when you read sometimes emails, you know, of co-workers, you know, and you know how they express themselves before AI. And now they're suddenly so polite, you're like, come on. I mean, be yourself again. We want you here, you know, with all your personality being, you know, I don't know, grumpy or whatever, we want to see that part of you. That's very human and just don't even have to be this polite, you know, and write those redundant messages. So I think that's something that LLMs have done to us. Maybe it's just a phase, you know, maybe as we move forward, we'll forget about that very specific style of communication that AI developed, but then, you know, it's all over the internet and it's everywhere.

It's also in our communication challenge. Then it kind of is a bit of a feedback loop where because of the fact that it's already out there, AI feeds off of it and then multiplies it. So I am afraid of that Omni writer, of losing something important because of that. So this is so interesting. And I think like there's this sort of like element of it being a great equalizer. Oh yeah. And there are parts of that that are beautiful, but it sounds like also a very real.

fear that these incredible writers are going to end up sounding like, maybe take the easy way out and rely too heavily on AI and then will be cheated of their creative genius. The flip side of that is I actually have a colleague who is not a native English speaker and was very self-conscious of the fact that, you know, despite being a very talented engineer, felt, you know, maybe like judged or dismissed when he couldn't always get the grammar right or, you know, would make little mistakes. And so he used AI to create a plugin to sort of make, basically like auto edits what he writes. Okay.

To like correct any grammar or kind of misuse of language or whatever. And the sort of base prompt of this agent is, you know, I'm a native Spanish speaker, so I tend to make the kinds of grammar mistakes that, you know, somebody who never spoke English at home or whatever might use. And so just, you know, clean it up so I sound like a native English speaker. And, you know, it doesn't sound like him, but it is more elegant. And his perception is that, like, his requests are treated more seriously and all these things. So, like, you know, here you take somebody and, like, you're enabling them to maybe compete on more equal footing or anything like that, and that that could be a great thing. But then you take these incredible writers, right, like a ward link.

And obviously, AI's not trained on that, or at least not enough, because that's only a tiny fraction of all writing. And so if you take that incredible stuff and you make it sound average, you kind of get why people would get mad. Yeah. Yeah. And I think it's so true. For your colleague, I would love to see a version of that plugin that retains some of his phrasing and still, you know, to some extent, makes it all more professional. So I would love to see a plugin that would still signal that this is a person with a different background, you know, maybe uses different type of phrasing, has a different. type of imagination and references and how to retain that and then still achieve the effect that

you've just mentioned where he wants to be treated seriously. As a non-native speaker, I also can resonate. You know, I can relate to that. It's clear for me that you also want to be perceived seriously because you have important things to say. So I think that that is to meet the hard part. And I have to tell you that I had this, I did, I think, quite vicious exercise with my students. I pushed them really hard. I told them that we're going to do a take-home exam. And obviously, take-home exams now are a contested thing because, well, you can use AI easily.

But I told them this whole exam is going to be exactly about you thinking about how to use AI properly, because I allow you to use artificial intelligence. I allow you to use language models for this exercise. What I would request is that you also present some reflexion. on how you've used it and what was the limitation. But I'm very open also to you not using AI at all. And I'd be curious to see what you would write.

And maybe you want to explore both versions, you know, and compare and then think about it. And I told them, listen, everyone, like you said, Emily, is now able to use AI. It's a great equalizer. And it's true now. And it's going to be true when you apply for your work, for your first job. or whatever. And what will happen is everybody will send their beautiful Polish CVs. And how will your CV stand out in, you know, a situation where that great equalizer

kind of uplifts everything? What is your contribution? And I think, you know, for them it was so hard, you know, and many of them decided to just write by themselves, you know, and they're very skilled. AI is their field of study. They study AI and management, and they're very good, you know, in programming and they're very good in machine learning, and Gen. AI is easy for them. And they still decided, I think, you know, I'm hoping that's the truth, to spend some time writing. And it was visible somehow, you know, those imperfect papers

were somehow perfect. So, yeah, I think it's very interesting and important for me to try and ask these questions you asked about, like, how to bridge philosophy. And I think that's, That is a bit of that, right? Anthropology, philosophy. So how to try and ingrain the whole process of using AI with these questions that make you think? And, you know, just pause for a second and be like, oh, do I want it? Do I not want it? Is it a milestone for me so important that I would just rather write it myself, you know? And yeah, like I think this is so, so interesting. It's so interesting everything you're saying. And I really do feel for and deeply respect all of the. professors and teachers out there really at every level that are trying to figure out right now. It's so hard. Everything around. I've heard everything from like it being banished to like you can use it but you have to sort of send in your prompt history.

Oh yeah. Like with your, just everything. And I just, I can't imagine how difficult it is. It's trial and error, you know. I think for all of us, it's like we really don't have the recipe. We're trying to explore it with our students, but ultimately what stays, I think, for instance, with me are our discussions when we have a class on site or online, and we have those vivid discussions about how to use AI. I just feel like this is the part that's very rewarding, because students also have very interesting ideas about it, you know, and if they justify to me well why they decided to use AI 100% in this assignment, I'm happy to do it. it. Maybe we'll just do, you know, the oral version of this, of this assignment where they

talk to me about, you know, why they've selected this, because ultimately, you know, judgment, selection, decision, these are very human things. And I believe that that part actually, judgment and decision should stay with humans. And so I think this is the most interesting part, not really how they've used AI, but, you know, why they made that choice. Yeah. I do think that that's super interesting. As we get close to time, I'm curious, what are the misconceptions that you most want to clear up or mention here for our listeners? Well, you know, I don't think that there are like maybe misconceptions, but I would say that for me as a person who studies sort of this approach where human is in the loop or at the center, I am really very eager to understand the pathway forward

because I see that there is this great temptation now and it makes perfect sense to like you said before to use artificial intelligence fully to have an orchestrator and agents that are going to fight with each other and then they will produce the best outcome and then you know you'll be able to use it elsewhere and that is great but I'm just

very curious about where is human exactly, you know, in that loop. When we say human in the loop, how are we going to plot it together? How are we going to design that process? And I'm very curious to hear, you know, from organizations about how their journeys look like with artificial intelligence, which parts they feel is okay to delegate fully and where that division of labor is still very important for them. And we've seen how with agents, I guess some people maybe do not appreciate enough how much agents can do. And then you have a deleted mailbox, you know, because agents decided that that's the best thing for you now.

So, and that has happened is not just like a science fiction scenario. So you can really go far with this technology. But is it fully aligned with what you would want to see? That I think is a big question. So I just feel like whenever there's that messaging that we're all in for AI, AI or nothing, we all have to be AI native. for me the interesting part is to try and ask some questions about it. You know, what does that mean for you?

And, you know, how do you see the pathway forward with that technology? What is it going to multiply? Because it can multiply the bad stuff as well. Sure. Yeah. And if you were speaking to, say, you know, various company leaders, like, what's the one takeaway you want them to have? You know, as you're sort of asking some of these questions,

encouraging people to think about it, what would you want them to know in setting up their corporate AI. Well, I would say an open dialogue with everyone about our journey with artificial intelligence is very important. You mentioned at the beginning that some people just work to train artificial intelligence. To me, it's like so sad. I mean, I see how it makes sense. You want to develop physical AI.

You want a robot to learn certain things. Maybe it's going to help in the hospitals in the future or something. So I'm not like negative about it. But I just feel people. deserve an explanation of what's going on and what's the next step. Even if we don't have all the answers, I think just like an open dialogue about how we want to use this technology forward. What do we think about it? What are the limitations? What's the pathway forward for various professions, you know, and jobs that we have in our organization? This for me is very, very important.

I don't think we've had enough of that. Maybe in some organizations that that is already practiced, but it feels like very often it's just, oh, we have this new tool, and now we're all going to use it, you know, and yay, hooray, AI is here, we're going to serve the AI hype. And I feel like it's not enough, particularly for those people who are rebellious against the technology, it's not so that you won't persuade them. You just have to have a very serious dialogue, and you have to take their considerations and concerns very seriously. So that is something I would like to see more of best practices, but also limitations.

just open conversations about AI, more honest than the hype. I think that's a great and important message, right? Like the hype is hype for a reason, and there's a lot of great stuff, and we don't want to deny any of that. But I think there's a lot of pressure on companies, on individuals and all these things, and that pressure leads us to maybe ignoring the bad that comes with the good. And so if companies want employee buy-in and, sort of a smooth upskilling, these opportunities for dialogue are super important.

So thank you for sharing the beautiful message. Thank you so much. And thank you for being here today. Well, I'm very happy to be here.