AI and labor markets: the future of work explained

Hello everyone, welcome and thank you for joining us today. We're really excited to kick off our fall virtual event series, Rewriting the Future of Work, a collaboration between MicroOne and Microsoft AI Economy Institute. This series is all about exploring how artificial intelligence is reshaping the world of work, from labor markets and mobility to workforce enablement and the opportunities and challenges that come with building a more inclusive AI economy. Today's session, AI, Labor Markets, and Mobility, brings together two thought leaders who are at the forefront in this conversation. We're joined by Morgan Frank, research fellow at Microsoft AI/EEI and professor at the University of Pittsburgh, and Joshua Browder, member of the board of directors at MicroOne and CEO of DoNotPay. And leading the discussion will be Mark Esposito, professor at Harvard and chief economist at MicroOne. With that, let's dive into it. Mark, over to you.

Thank you so much, Sophia. Nice to see you again in this series that you've been helping me with since last week. Morgan and Joshua, thanks so much for joining. I really appreciate it. I'll start from you, Joshua, if that's okay. I remember the news that you were joining MicroOne not too long ago and the fact that you brought a lot of excitement to MicroOne. So look, you are operating in the industry. You are co-founder of company operating in the technology spectrum. You joined MicroOne now. Just tell us a bit more about your, your background. What brought you to where you are right now and what has been the gravitas in your career, Joshua, when you're looking at AI? Why did it become an important part of your, your story? Because I think none of us started as AI native, but we eventually we converge in that direction. Joshua.

Yeah, so thanks so much, Mark, for having me. So I'm an accidental entrepreneur and an accidental builder in the AI world. I started my company, DoNotPay, because I got a lot of parking tickets and I didn't know how to get out of parking tickets. And also they were very expensive. This was back in college at Stanford. And so I became an expert at getting out of parking tickets and I built the first version of DoNotPay really just to help a few of my friends and family get out of their tickets. And I launched the product and it just went internationally viral. I went from like 10 users on the first day to 50,000. And I thought maybe there's something here in terms of helping people fight for their consumer rights using AI. And so fast forward 10 years and DoNotPay now operates in over 100 different areas that helps people fight big companies and the government, like getting refunds, negotiating bills. We have AI robots log into someone's Comcast account and start chatting with the big company to negotiate the bill. And what's interesting is the big companies are using AI and we're using AI. And so sometimes it's actually an AI versus AI negotiation. And I started investing 5 years ago. And I really just did that to relive my journey. And in that process, I am fortunate to be the first investor in MicroOne. And so I speak to the amazing MicroOne team every day, and big fans of what everyone is building there. And recently I decided to join the board as well. So it's been an amazing journey of both building and also investing in MicroOne.

Wow, Joshua, that's— I didn't know the story about do not pay.com. Fascinating. And I really appreciate the fact that you decide to join this company, not only as an investor, but also as a member of the board, right? Because I think it's really about, as you know, as a company going through a significant phase of its own life, right? As the valuation that happened not too long ago kind of positioned the company in a place where now things can no longer be accidental, but they have to be intentional towards the next step. So really, really nice to meet you. Morgan, I'm going to go over to you. Look, a little bit of the series that we started with Microsoft, they start with you. You're the first fellow from that cohort that we are— we're kind of dedicating a session about. So first of all, it's an honor that we can start this. And look, you are a scholar, you're a professor, but you also decided to join Microsoft with this cohort of fellows. So tell us a little bit more about your journey somehow as, as a researcher that decided to, to go for this. And what was the research question that really brought you to think that it should not have been just a journal, but it could have been a company like Microsoft, right, to really bring your research to the next level? And welcome, Morgan.

Yeah, thank you for the warm welcome. Um, so I started out my academic journey as a researcher, actually as an applied mathematician interested in things like chaos theory and dynamical systems, and eventually shifted to studying social systems as these complicated things where there's a lot of interacting parts and individuals bumping into each other. And when I started my PhD at MIT in the Media Lab, it was clear to me that this framework is a useful way to think about labor markets too. AI is changing people's careers by changing what kinds of workplace activities are in demand. And so there's this really microscopic change that's happening, producing macroscopic things that are more common to talk about. So since then, we've been trying to develop this idea and finding new data and new insights. That help us build it out even further. So this opportunity with Microsoft as a fellow in the AI Economy Institute is really tremendous step forward because Microsoft is, I think, uniquely positioned to connect very valuable data on this. For example, Microsoft in its umbrella of assets includes things like LinkedIn and Copilot and GitHub. So there's a potential to really study how people are actually using AI functionally and ground truth and what is happening to them later on. So I'm really intrigued by this idea. I think we can cut through a lot of the confusion about AI that's in the media right now by actually following individuals to see what happens. And I think the AI Economy Institute is very motivated to move the field forward.

Wonderful. More, I'm gonna ask you in my follow-up question later on more about the cohort of colleagues that you're working with. I think it's for people that don't know that kind kind of magnitude of that. And there was a call for a second cohort of the fellows, right, that just happened. So there's something I want to ask so that you can help more of the conversation moving forward. There's one thing you said, Morgan, I'm going to bounce it now over to you, Joshua, right? So Morgan talked about data in our conversation prior to going live. You were mentioning to me, Joshua, that one of the things that has always been of interest for you is the need for high-quality data. And I would imagine, and correct me, my assumption is wrong, that what also brought you close to MicroOne, like in my case, was the fact that data was a serious business for MicroOne, right? It was not just like, like, like an accessory, but it was really defining the business model. What do you mean, Joshua, when you talk about the need for high-quality data? Where was the fundamental problem you encountered? Do you find that we are now better than when you maybe started to realize that we had an issue with that?

Yeah, there's so many areas of the world where there's not reliable data to train these AI models and products. And one area is what we work in, which is consumer rights. And so, for example, we have a product that helps consumers get money whenever they get a spam call. There's an amazing law in the US called the Telephone Consumer Protection Act that allows consumers to get $1,500 every time they get a spam call. And the problem is these robocallers use fake names and fake numbers. And so it's very difficult to actually track them down to get this compensation. And so what we did at DoNotPay is we built a trap. And the way the trap works is whenever the robocallers try and sell our customers something, they can use a special DoNotPay virtual credit card. And they say, I'm interested in what you have to buy. And through the payment network, it gets their business name, details, address and everything. And it traps the robocallers to get a demand letter to get this $1,500. And when we were building that product, this was a completely new area. Like no one had ever done things to kind of track robocallers and things like that. And we really lacked a lot of data and it took us like months and months and months of like Freedom of Information Act requests. Kind of pulling legal data to kind of build this product. And DoNotPay, not on the data side, but on the talent side, is actually a customer of MicroOne. And now that there's more of a focus on data, we should have used MicroOne to kind of parse all of this data. And so that's just one example of where AI and products are only as good as the data inputs. And we had to scrape all of this data to kind of make it happen. And it really slowed us down. And it, but had we had the data refined from day one, we would've been able to launch the product much more quickly.

Yeah. Joshua, just as a, as a quick follow-up on this, do you find the issues that the data is simply not good quality or is it more about scattered, not concentrated, or eventually not available? So it's more a scarcity issue or more of a quality issue in your case, or both maybe?

It's not necessarily about quality. It's about turning unstructured into structured data. You can think of like some data, like on the parking ticket front, it could be like a scribble on a parking ticket and that's very unstructured. Or it could be the kind of, everyone has seen in San Francisco, the 4 parking signs that say opposite things. So that's an example of very unstructured data and it's how can you structure it and get it into a model that systems can understand and you can kind of operate on top of it.

Fascinating. Morgan, maybe to you before, maybe on the question on the, the, of the cohort, just to, to continue on the same chain of thoughts. Data is also important in the work you do. And I think you equally have mentioned to me before that the moment that you had access to data, you could really study the evolution of data. Not necessarily in a short period of time, but probably more longitudinally or eventually across a person's journey moving forward. So from your point of view, I think Joshua was extremely, extremely persuasive understanding the challenge with data from his point of view as an entrepreneur. From your point of view as a researcher, where was the issue with data and how do we use data today to simply improve our understanding of people's journey? And they could be career-wise, but I would imagine applicable to any part of of structure and data for a specific purpose.

Yeah, absolutely. So I got it. I got started with this interest in this question, I think the same way a lot of people get interested, which is they see a headline that says AI is automating jobs and you say, oh, that sounds awful. And then you start reading the article and you realize there's actually no data there. There's nothing behind it. And actually, if you talk to labor economists, the prevailing wisdom almost unanimously, at least before generative AI, was that technology shifts, they do destroy some jobs, but it's rare and it creates more jobs than it destroys. And many jobs change in the process. And so this made me think, that's really interesting. It's a really subtle, nuanced understanding of what's going on. Let me go learn more about the research. And what I found is that the research is very departed from that story. The scale is actually very different. It's really hard, for example, to understand who's losing jobs and who's gaining jobs when you just look at the employment for software developers in San Fran— in San Francisco or in California. It could be that this year and next year the employment is about the same, but everyone's been fired and everyone's been hired. There's been a complete turnover. You would actually miss all that if you were forced to just look at data from the Bureau of Labor Statistics. Another version of this that I'm actively working on right now is none of these studies I'm aware of actually look at unemployment directly in a way that's convincing, at least to me. And that's because the government, the Department of Labor, only reports total unemployment. They don't break it down by what kinds of workers, what kinds of sectors, what kinds of education, even things like ethnicity and gender are absent from the unemployment statistics you can typically get. And that seemed silly to me. So there was a while when we were in the middle of COVID and I was freshly stuck in my apartment. I was able to work with some, I won't say who, but some well-known economists at MIT and Harvard at that time. And I was emailing them, is there any data on unemployment by occupation? And one of the replies I got from one of these ultra-famous MIT labor economists was, That's a silly question. By definition, unemployed people don't have occupations. So strictly speaking, that's true. But I think that that's sort of a silly answer to the question. So the journey has been to break through these aggregate cross-sectional statistics, find ways to actually find data on individual outcomes, break apart unemployment. And in my case, we're actually working with each US state's unemployment insurance program. Where they do actually collect information on what your most recent job was when you claim unemployment. And trying to find ways to actually track people's journey through these different bumps in their careers. So looking at, for example, worker histories from LinkedIn is a really powerful complement to the more comprehensive data you get from the Bureau of Labor Statistics and the Department of Labor. And The type of information you get when you follow individuals is just incredibly rich. We know things like where you were educated, what did you major in, when did you graduate, all of the jobs you had since then, who were your employers. We even have ways to estimate what your salary was. This creates a tremendous opportunity to focus on questions that we care about right now. For example, what is happening to recent graduates? Are they having a harder time finding employment? Or are there other frictions in their labor market experience that we can actually quantify? Things like lower salaries or more time job seeking. What kinds of people are actually getting by and not having this issue? And what kinds of people are instead experiencing these different frictions? We can answer this, these questions, and I'm not aware of really any other resource that gives us these kinds of insights.

Fascinating, Morgan. While you were talking, my mind was going in so many different directions. If you allow me for a follow-up question, just on a very quite trivial level, imagine you are scrolling on the feed on social media and you find— I think I happened to read this yesterday where there was this announcement that Accenture by restructuring was letting thousands of jobs go. Right. What do you think when you read those? Because you would imagine many people will read this and it will validate the fear they have that technology is replacing jobs, right? Because I think the magnitude of the numbers where this news are reported, if factual, they feel definitely a paradigm shift. But from a research perspective, you will probably filter that information through a different lens, right? Probably less dramatic than what it seems like. What would be your interpretation when you see these large numbers? I can't remember the exact but it was 1,000 that were basically let go by Accenture just by restructuring. And always in the headline, there is a conversation about how AI is kind of prompting the conversation to a restructuring, which is fair. But what— how do you make sense of those news, Morgan?

Right.

So the first thing I've learned when you see these announcements about companies reorganizing and letting go of some workers, you need to also look at their hiring patterns and their hiring intentions in the year to come. So we had this conversation with Microsoft when the fellows all visited their main campus, where we had exactly this conversation and said, well, what about Microsoft? And Microsoft, of course, had representatives in the room and had recently had an article run about their announcement to let go of a certain number of workers. And their perspective was the concern is, in their case, is overblown because they intend to hire more people than they intend to let go this year. Now, this might not be the case across every company, but just be cautious about that. My second thing is that I am worried that AI is kind of a scapegoat for a potentially for a correction for overhiring during the pandemic that a lot of tech companies underwent. This is a concern a lot of people in our field will list, and I think actually we can control for it. So we're going to see some studies in the near future that make some claims about AI's impact on workers and are able to control for this decrease in hiring we've seen in tech. The more concerning thing is to disentangle some of the other economic factors that were happening at exactly the same time. So pandemic recovery policy was changing a lot in about the same time that AI was launched. Again, I think this is something in my own experience that I think we can control for. One of the real big elephants in the room is that at about exactly the time that ChatGPT was launched, federal interest rates went way up. And unfortunately, tech companies are particularly susceptible to dynamics that result from that. So it could be a correction from overhiring during the pandemic. It could be that it's now harder to raise capital than it was 2 years ago. And all these things correspond exactly with the launch of AI. Another reason to be a little unsure about AI is you're starting to see more and more discussion about things like work slop and how AI is not fully automating a lot of work and actually creating new work in other areas. For example, managing and correcting and, and sort of checking the quality of work. And so this type of dynamic actually falls well within the theory that labor economists apply to technology broadly, which is that technology changes things, but it creates new needs, new work, and new opportunities.

Yeah, great. I love this. We can keep on building on this. It's really the fact that maybe we are almost fixated on a measurement of value that is no longer kind of commensurate to the time we live. It's almost like the zeitgeist that we're experiencing. Pushes us to rethink whether the value of— the measurement of value is what we really are able to understand, or whether we're using a legacy system, most likely inspired by industrial policy of the last 56 years, to basically try to understand a phenomenon that is largely different from the other industrial revolutions, because it's the first one that is impacting mostly the middle class working in an office job because of the cognitive nature of this technology. And I find this to be Again, fascinating to continue to talk to you about this. I'll jump over to you, Joshua, now to try to build a bit deeper the conversation that Morgan has helped us to craft. And in this case, Joshua, so you work in the field, you're working as an entrepreneur, you're basically looking at the people that are working for your companies. Did you notice a change in the kind of direction of the employment of the people working for your companies? Did you start noticing any differences in the way, for example, talent was being found, whether companies like yours were having higher or less chances to find the right people? And I'm asking this question to, to determine whether the issue was not employability, but mostly fit on the kind of job that we're dealing with for technology companies. So what's been your experience on that, Joshua?

So when I started my company, talent was much more rigid. This was in 2016, 2015. Everyone would just go in the office and work on one thing. And then given the pandemic and now kind of the various opportunities that everyone has, it's much more fluid. And so we have employees who we are just contracting with them for their unique special expertise in a, in a small area. and they might do something else. Um, they might like, for example, our designer, um, contracts with, with several companies. Um, and he's a very talented designer and we wanted to get his talent. And the only way we could do that was, um, allowing him to, um, contract. And so I think especially kind of the younger generation, they're seeing their skills as much more transferable between roles. Um, We had someone else who, we had a very specific problem with a robot trying to appeal someone's bank fees. And it's a scraping problem where the robot wanted to kind of click on the right buttons and send this appeal off to banks to appeal people's overdraft fees. And this was about 2 years ago. I got introduced to this amazing engineer who just wanted a contract to solve this specific problem for us. And so he solved the problem for us. It didn't work for us as an employee, just as a contractor. And then he actually went on to found his own billion-dollar company called Hightouch, which is now $1.5 billion. And so you even have like multi-billion-dollar founders offering their skills in different ways, which is interesting. And so I think that the traditional job being fixed, kind of like in the unemployment statistics that Morgan was mentioning is the wrong way to look at it, and people are much more fluid now.

Just if I may follow up on this, um, because it's an area where I think, um, people working at universities like Morgan and myself, we get exposed a lot. Do you find that universities are still feeding the kind of talent that we need to have? And I know it sounds like a cliché question because I'm hearing sometimes two sides of stories, like, well, if you measure this in terms of direct employment, maybe it takes longer before we could think that a specific degree leads into a specific occupation because we change, right, the way we work. But do you find the university is still kind of giving you that critical mass of people that you can really use? Or do you find that we have expanded the boundary of where we are recruiting people, where no longer is just directly coming from, from graduate jobs? And, and this is important because many universities are suffering of a real crisis, right, of fit between the graduates and their ability to enter the job market. So what's your take on that? Did you find that somehow that narrative is fair? We do have at the moment a crisis, maybe from a model that was designed in the past, or do you find that actually, no, we still have a relatively good conversion rate between graduates and their entry into the job market?

So, um, it's such an interesting question because, um, I actually, uh, in 2019, I took a program called the Thiel Fellowship. Um, for the audience that's not familiar, the Thiel Fellowship is an award from the, um, investor Peter Thiel where he gives, um, students $100,000 to take a 2-year break from college. And so I did that in, uh, almost at the end of my Stanford journey. and, um, to work on my company. And the, the reason I did that is because the world is moving so quickly that, and I will talk about some advantages of college. So Morgan and you, Mark, don't kill me as faculty. But, um, I think the reason I did that is the world is moving so quickly that people see their life as kind of paint by numbers. And, um, by the time that they've learned everything they need to learn, the world changes under them. And so for me, it was particularly helpful to get practical experience, um, with my business. Um, on the other hand, the one big benefit of college is that there's no other place in the world where you have thousands of other talented young people that you can work with on projects, can be referred to different opportunities. And certainly, uh, for my company, I actually recruited a lot of my really smart friends. From Stanford. And I wouldn't have been able to do that if I didn't go to college. And so it's very useful to kind of think about ideas and meet people, but maybe in terms of the fixed curriculum, the world is changing much more quickly than that. And then the last thing I'll say on kind of hiring and credentials is that as an employer myself, I really value what people have done. And what they've proven versus any credential that they have. So I think it's important for everyone to start building up a portfolio on X to kind of show proof points of work. And I think that's 100 times more valuable than just having a certificate.

Yeah, fascinating. Oh yeah, please, please, please. Of course, of course.

Yeah, yeah. So I'm also very interested in like the million-dollar question for universities is what is our value proposition moving forward? Does it need to change? Do we need to change how we teach? Is there a pedagogical shift that we need to make and what is it? And I think it comes down to should we, should universities lean in to teaching skills that are very close to what AI can do? With the hypothesis being that we're training students to be the most effective users of AI. They understand the nuance, and so they understand how to get the most out of it. Or on the other hand, should we lean out? These are skills and abilities that AI can do, or we think they'll do in the near future. The demand for them is just gone. So it's redundant. You don't need to know that to be important on the job market. I think it's actually an open question. I think both hypotheses are plausible and have people who believe in it. It's just a matter of actually figuring out what happens to graduates who are exposed to different types of curricula. One of the things that I'm doing with the AI Economy Institute at Microsoft is the worker profile data we have, like I mentioned, it's very special because you can actually bridge the gap from education to what happens to graduates. And we're combining this with another dataset from the Open Syllabus Project, where they've given us 3 million university course syllabi representing the learning objectives and learning activities for courses at universities across the US. And we can actually say, well, how different were these? What kinds of things were being taught or emphasized? And then what happened to graduates from, say, computer science at MIT compared to computer science majors at Stanford? Or maybe even graduates from the same university at different years. What happens to them when they enter the workforce? So we can actually get some insight into this question moving forward. And some of our preliminary analysis suggests that teaching students AI-related skills, things that might be automated by AI, might not. There still appears to be a value to that, at least in the years since ChatGPT has launched. The other part of this question is that the— it's a two-sided job market for college graduates. There's employers and then there's graduates, and universities are of course trying to support the graduate side and don't necessarily have a ton of influence on the employer side. And we need to better understand why entry-level jobs are sort of disappearing across the spectrum. So for me, one of the eye-opening things, I was part of a cybersecurity labor conference here in Pittsburgh, and they had all the big cybersecurity firms in the Pittsburgh area together. And the, we went company by company and said, are you hiring? Do you have any entry-level positions? And one by one, they said, they said, yeah, yeah, or no. And one company said no. The one company that said no said that their strategy is to poach people from the other companies. And the companies that had entry-level job postings for cybersecurity all required at least 5 or 6 years of experience. So that means that effectively there are no entry-level jobs. And I wonder how widespread that is and how such a disconnect could have been allowed to grow and what it takes to get over that. But again, just to highlight, there's two sides to this market and there's a bunch of confusing dynamics on both sides.

That's right. And I would imagine, Morgan, only more data will help us to understand, at least to a deeper level, right, the relationship between companies, uh, can employees or, or jobs and And of course, the journeys that are actually ensuing out of this, which brings me to kind of follow up on what you just said now and what you were saying before. So if I kind of understand correctly, the reference we're making about the study of individual careers, especially with the support of tools like LinkedIn, that really tell us a story different from what we might be capturing at the Labor Bureau, right? Is that careers are no longer as linear as they used to be. And this is, I think, coming from a behavioral change, generational change, a perception of identity that has changed greatly in the last few years. So when you're kind of looking at that evolution in a less linear fashion, do you find that this is coming at the expenses of employability? In other words, we're kind of trading off choices and arbitrary and discretion in exchange of what might have been a much more continuous career? Or do you find that we're simply improving the labor mobility because we're having people that keep on cross-pollinating skills from one job to the other, but they don't do it from the same industry in a way that we can eventually determine a trajectory that tends to be nice and neat, right? So do you find this to be a reassuring trend? Do you find it to be a trend that concerns you when you're looking at this this non-conventional evolution of individual careers and looking at the data?

Yeah, so I would say there's been— it's varied sector by sector, but I would say that there's been a general trend towards labor markets being more liquid. I think that's a nice way to think about it.

Yeah.

Actually, where I saw this the most was I spent a summer in Silicon Valley at LinkedIn years ago and was— I thought it was really crazy to learn that it's very common for people to bounce between all the big tech firms there. And even return to a former employer at some point later in their career. There's really a lot of flow of knowledge and talent that way. And that's not something that happens elsewhere where things like non-competes can completely limit. For example, if you're a doctor in one city and you would really like to make a switch, well, you better move, you better go to a different city. So it's not true for every sector. And I think on one hand, that's really good. It means that it's very, it's easier now for whatever reason to allocate talent to problems and match those things up much more easy than it was before. But it comes at the cost of stability and the ability to plan for the future. There's, in generations past, there was sort of a stronger agreement between employer and employee that this was going to be a more stable relationship and more common for people to spend their whole life with that company. The, the drawbacks of that are maybe you're not realizing your full potential and maybe the economy is not being as efficient as it could. But the benefits of that outside of your economic life are a bit harder to quantify. Things like you have, you're married to one community, presumably your kids are going to grow up in one household and you just have a better stable foundation besides your work life. I think now we're part of the issue is that these promises that were made to have a certain skill set and you'll have a lot of control over your career, those promises don't seem to apply anymore. So they're the classic example in my mind is when I was growing up, everything was, was about learning to program. Like you must learn to program, it'll be a tremendous asset to you once you try to get a job. You can get a job at one of these great tech companies. They're amazing, high salaries. They have, they have cafeterias and gyms. You'll be so happy. And of course, that story has a bubble associated with it. It had to burst at some point. And so I think we're just living through that. What does this mean for what students should do? How should they allocate their time and attention? That's a much harder question, but I think this current uncertainty, potentially because of AI, is something that's happening more generally and has been happening across a variety of sectors.

Fascinating. My only regret about our conversation is that time will never be enough to cover everything that comes to my mind. So allow me, Joshua and Morgan, maybe to start bringing our conversation to the questions It happens that there are two questions and they kind of fit, one for you, Joshua, one for you, Morgan. And, you know, we have at the moment 90 people currently connected, right? We have a pretty high peak throughout the entire session. So for those of you watching, if you have any question for Joshua and Morgan, please use the Q&A function. So we'll try to moderate as much as we can within the time we have left. Joshua, there's a question from Alexander, which is fascinating. So He says regarding the labor market, right, apart the people that have an AI background, which you could say they're, they're kind of easier to hire simply because they feel to be the right kind of fit for the right kind of conversation we're having. And besides the job where we're labeling and training, so things that I have started when I did my first induction into MicroOne where I was training, I was a human data expert, right? So what are kind of the jobs that you think are kind of redefining the labor market. And I think the question that Alexander is asking to you, Joshua, is mostly related to you as an entrepreneur eventually defining those needs as they are coming and as you can see the companies in the ground. How would you answer this question?

I think depth is much more important than breadth in this kind of hiring market. So what I mean by that is Um, if you are the top 0.1% expert in something very niche, that's very attractive to us as an employer. So for example, um, we, we get, um, a large number of our customers from, um, SEO and increasingly, uh, GEO, which means AI is actually referring our website to potential customers. And if you're a top 1% expert in SEO or GEO, we're very interested in talking to you. Because you, you have so much specialist knowledge, or maybe someone who's really, really good with Android development who can know exactly, like spent their entire career focusing on that very niche thing. And so I think in today's market, you wanna be an expert at something and, and be really top 1% of something very specific. And the kind of all-round type of people, it's still good to be an all-round, all-round expert and all-round type of employee, but it's much, much more exciting if you're really passionate about one specific skill.

Yeah, that's great. That's great. Thanks for this, Joshua. Margaret, the next question, as you can tell, is coming directly from Edwin, who was hearing you speak before. I think the more concerning part of the conversation we're having is that it seems like that maybe it's not an issue with entry job being unfit for the employment. It's just entry jobs are no longer happening as they used to, right? And he asked actually a fair question, what can be done to change this culture? And I think I was reading not too long ago the fact that we shouldn't evaluate the job only as a job. It's a whole social inclusion journey that really allows people to expand the boundary of what they do when they start working for organization. I think there's both implicit and tacit and intangible that are coming with a job, not just necessarily the material part. So how would you— and I know these are not questions that easily can be answered just through research, but maybe your own take on this as you've been working with this data is what do you think we could do to change that culture? Should we change the culture? Are we basically trying to fix what has already been broken?

So I think this question is actually pretty related to what Joshua was saying. You know, I think it's fine for companies to want to target the top 1%, but The downside to that, or maybe not the downside, but the other side of that is that there's, there's a very small number of people that are within the 1%. So that's not the problem of the employer, but it does raise the question, what should these other people be doing? And I think what we need to emphasize right now and in general too is not that you're on a fixed path, but that you need to find something to be passionate about and really dig in. And when the time comes, switch what that thing is you're passionate about. Now, often there's a lot of expertise from one experience you can take to the next thing. I do a lot of research with data. I use very similar statistics to answer a variety of very different questions. And so the, the tools I use to learn for one project, I carry with me to the next thing and I become a deeper expert on those things as I apply them in different ways. So I think the way to sort of combine my thinking, what Joshua was saying, was that you need to be very good, you need to be very passionate, but you need to also be adaptable and creative about what it is you can do and how to apply that. The way to think about this if you're entering the labor market right now or entering college right now is to try to learn a lot of skills that are applicable in a variety of domains and a variety of problem areas. Take that with you. Don't be too married to one thing. So for example, if you're majoring in computer science right now, don't be married to the idea that you're gonna work for Facebook, Google, or Microsoft. Be more open to adventure and think about how these tech skills are now in need across the whole economy basically because of things like AI.

That's great. Morgan, I really appreciate this. I think it's a responsible call for the level of engagement that is not expecting the system to fix the problem for you, right? But really build your own journey and in a way that it requires the degree of interoperability as opportunities are really emerging. So I guess it's an acceptance that the fluidity of the times we live in are not necessarily negative per se. They are only prompting us to consider some of the assumptions differently from before. And this is not coming as an imposition from outside. I find this a direct reflection of who we have become over the years, right? So sometimes we think technology operates outside of our own thinking, but it is maybe an extension of our brain, just probably amplified, and which I think is where it becomes sometimes fearful for many. Look, I think there is a— folks, we're not too far from eventually wrapping this up. There was, I think, a question from Jorge about the revolutionized area of AI in teaching and learning languages and translation. Jorge, this is a question I wouldn't have an answer for. So I don't know whether it's anything that any of you wants to take on this. Morgan, maybe you have a take, a first stab on this?

Yes. Yeah. But I'll caveat it by saying that it's basically pure speculation at this point. It's just my best guess. I think that education, especially higher education, is going to change dramatically because of AI and not necessarily in sort of a substitutive, deleterious way. But I think that these AI tools have the potential to make education more accessible and more efficient for people. We just have to adapt the university system pedagogically to use this tool more effectively. Some evidence is just, you know, think about the value of a tutor and how hard it is to get enough TAs for really large popular classes. And now you can talk to ChatGPT. It's basically a 24/7 on-demand tutor that can meet you where you are with very fluid flexibility and complexity in how it thinks about a thing. So that's amazing. And I imagine that that will grease the wheels and people who are on the margin of going to college are now more likely to because they will see they have a tool they didn't before. It should be easier. And if this is correct, it might mean that universities need to grow at the same time, maybe change the way that they offer courses. But I think the demand for higher ed is going to be perfectly fine. Uh, the other reason I'm optimistic about this is that there's now a handful of good case studies where, for example, AI has been deployed in a random control trial in a customer service setting or in a medical setting, or Ethan Mollick does some experiments at the Wharton School with his entrepreneurial students and all these different applications across different domains. And my, in my reading, they all sort of point to the same general result, which is that generative AI is not drastically improving the performance of high performers. It's really improving the performance of low performers who start to look more like high performers. So imagine if this result is true widely and holds in the classroom. That means that students who are struggling for whatever reason, maybe they're first-generation students or they're a minority on their campus, they now have a better opportunity to succeed. And that, that's a good thing for society and for the labor market.

100%, Morgan. I have to say, even at Harvard, I'm sure you know the work from Karim Lakhani from Harvard Business School, of course, not my school. But I think in the BCG experiment they ran, they were able to find out that what the GenAI was doing was really lifting the floor, not necessarily improving the ceiling. Right. And I think it's an important, it's an important part to to actually, I think, normalize the use not as dystopian or draconian, but really it's just an infrastructural improvement of some sort. And so this is important because I think it's coming from another source of research that keeps on supporting, I think, the right level of engagement in this conversation. Many people, they come in with big questions, and although we might not have the answer for that, we might want to direct them in the way in which they can restructure their own thinking, which is appreciated a lot. Joshua, there's a question that kind of I found Quite interesting. It came from somebody, I don't know the name, is anonymous. But, and I'm going to talk to you more about somebody who really were able to be successful starting from your time at Stanford, becoming an entrepreneur, eventually really getting to where you're right now. And I do sympathize a lot with this, right? You know, a lot of fatigue. I think exhaustion is probably also equally the right way to feel. A lot of applications, lots of interviews, a lot of courses. It looks like No matter how much in an incremental way we're trying to, there's still a sense of not good enough, ghosted, discarded. I'm using the language from this question. So what do you think we can do? Is it a problem, the fact that we— is the attitude that we're having that keep on reinforcing kind of the fatigue? So somehow it's really about breaking a vicious circle that we are feeding. Or do you find there is a better answer for that kind of situation? You would probably find lots of people that come across with this degree of frustration as well.

So I see, I see this is very similar to when startups are pitching. So if one person is turning you down, that, that's fine. If 10 people are turning you down, that's fine. If 50 people are turning you down, that's fine. But if like 300 people are turning you down, there comes a point where one has to evaluate, maybe they need to change something about their pitch. And this is true when pitching companies, but also true in job interviews. So maybe it could be adjusting the type of role that they're going for. It could also be kind of upleveling their own skills to make them more attractive to the employment market. But I really think both with kind of pitching one's company and also pitching to join a company, it really shouldn't be about a volume game. It should be about a quality and depth game, and it should be much more, uh, of a focus to kind of pick the right opportunity and, uh, make sure one is a fit for that than just scattering, um, applications. Um, and then I would also say that, um, networking doesn't just have to be in person. Um, doing kind of this portfolio on LinkedIn and X and connecting with people, um, that's also a form of networking. So I would say, unfortunately, the majority of jobs are still getting filled with people, someone who they've known their work and they've had a real connection with them versus cold applications, at least that I've seen in my hiring. And so I would say further building an online network is important too. And maybe the network should come first versus just scattering applications.

Yeah, great. And folks, I'm kind of bringing that over to a closure. Alessandro made a generous comment, and thank you so much, Alessandro, for sharing the support of the work we're doing in this conversation. Russ, the same for you. Thank you for making the comment. And there are two quick comments that Kevin and another Alessandro have made. I'll take the liberty to just to give a brief just the direction on that. I think, Kevin, on the question about cheating, definitely we understand this is also true, that maybe we have to increase our understanding of governance oversight, but also the differentiation where humans and machines are not going to be measured by the same proxies, right? But understanding where humans are capable of doing better versus machines and kind of differentiating the use in the same way you would phrase your calculus on on a piece of paper versus a calculator. So I think this is a space for us to keep on engaging by redefining the boundary of the jagged frontier, like, like it was defined in that work from, from my colleagues from the business school. And to your point, Alessandro, I think the purpose of a company like MicroOne is really to expand the train of data to multiple domains, not just the more traditional ones, to the point that, yeah, people working in philosophy, humanities, literature, But even doctors, architects, right, lawyers, they could keep on getting engaged in training the data so that we are operating across a workflow rather than necessarily across silos of specific domains. And that's how organizations fundamentally work. And so I think the space of human data training will continue to create opportunities for people that were non-conventionally contemplated by the industry itself. And I think this is where maybe there's a space for your question. So I appreciate that a lot. So guys, I know, Rahim, I don't have time to maybe address your question. I really appreciate that. I think maybe my final question to Joshua and Morgan, I'll try to integrate a bit of a question about the future and maybe what we could do. So Joshua and Morgan, my very last, maybe a 30-second answer from each one of you. Where do you see the nature of your work going from here? And yeah, just share a little bit about the prospect where you see your contribution to this conversation that we had, this discourse, to move on in the future. And really, what kind of inspiration can we really share that you have learned by doing this work to the people that have been watching us today? Joshua, I'm going to start from you.

So I view Do Not Pay as my life's work, but I'm very excited on the investing front. And my investing model is very unique. Which is I find people who are extremely talented, but maybe aren't getting exciting funding offers and maybe need repositioning. And I become their first believer and I go all in to help them. I host them in my spare bedroom. I put their visas on my personal credit card to move them to Silicon Valley and all of this stuff. And I really love making people's dreams come true. And I think on the talent front, um, people, more people should consider starting companies. There's still unlimited opportunities, and especially in the age of AI, starting a business is easier than ever with all of these tools. And so my focus is really on cultivating talent to start amazing businesses, and I'm really excited to do that over the next 10 years.

Fantastic. And I think just really also kind of help the question about what should we really think about for our children, and the idea is really to help them to become entrepreneurial, right? And not necessarily just about setting company, but it's the attitude of trying to think in a not necessarily structured way from a traditional perspective. And I really appreciate your comments. Morgan, I'm going to finish with you on two things. One, same question, what is really ahead of you in terms of your kind of research or your kind of direction of travel? And maybe if you can finish also by saying, tell us more about this cohort of Microsoft fellows, right? What was really so unique about this? And that would be a good way for us to also close our first series that was dedicated to, to the work you guys doing at Microsoft.

Yeah, sure. So in, in sort of the short term in my work, I'm really excited to find solutions to the problems we're uncertain about right now. Unfortunately, questions like, is AI taking jobs? Is, is still even just yes or no is not clear. So I think that there are data sources and opportunities to answer these questions that didn't exist a few years ago, and they're arriving just as we really need them. So I'm very optimistic about what we're going to learn from data about how to best adapt to the times we're in today. And I think that my research is going to help with that. There's a lot of other people looking at exactly the same question with a very similar perspective on how we can advance. So I'm optimistic. In terms of Microsoft's AI Economy Institute, this is an opportunity that falls well within what I just described. This is something that's starting their second year. I was part of the first year of fellows at the institute, and it's basically Microsoft's response to— there's a lot of important questions AI is creating. It's hard for us as a company to focus on all those things ourselves, and maybe having some outsiders, some academics, some outside experts who can look at these things without the business angle. So perhaps some more honesty and some more trust can be gained from those perspectives. They really want to answer those questions and they want to help experts produce answers that are believable to the public. The first cohort was very focused around just broadly what is AI doing to education? How should AI adapt? How should students adapt? How should employers adapt? And that was just the theme for the first cohort that I was a part of. The second cohort, again, very broad strokes focused on education, but also things like how is AI impacting people globally? A lot of what we discussed was pretty US-centric. Can we adopt a bigger lens and better understand who's being left behind, who has access, and even the multifaceted part of who benefits from building out AI infrastructure? Things like data centers, for example, are a potential opportunity for folks in rural areas and create economic opportunities that don't require a lot of education. And it's sort of separate from the discussion we've been having today. So Microsoft sees this wealth of questions. They want to bring people together who can work with Microsoft and their unique set of data and assets to really understand this stuff. And that's the goal moving forward.

Fascinating. Gentlemen, I am always the lucky one moderating this session to probably acquire the most learning from just incredible people like yourselves. First of all, thank you so much. I truly enjoy this. This is not just something that I do for the sake of animating a series is really that I learned so much from just exposure to people that are advancing the conversation in the right way, which is not necessarily by pretending that they know it all, but really by probably helping us structure questions differently. And I really find this to be the essence of what this new era is really about, is really helping the diffusion of information and knowledge that is so, so critical. For the way we're going to structure the question and also the conversation we're going to have in our heads since we're becoming more and more cognitive beings in many ways. Joshua, Morgan, it's been an absolute pleasure. Thank you very much for being part of this inaugural setup that we have started by honoring the fellows at the Microsoft AI Economy Institute. I hope it's not the last time we're going to have a chance to talk to each other. I think this is the very beginning of conversation we will have. And from Mark Thank you for all of you watching this. We had over 90 people constantly connecting to our conversation, and I found it great. You know, time was the only constraint we had, but it's a pleasure that I'm going to bring this over to closure after about an hour together. Thank you so much, gentlemen. It's been a pleasure, and see you soon.