How AI Will Reshape the Economy, w/ Anindya Ghose, the Director of AI at NYU Stern

Hello and welcome to AI Curious. My name is Jeff Wilson. I'm a journalist. I'm an AI strategist and I am curious about AI. Today's episode is framed around a big, slightly terrifying question. How will AI change the economy? The new job ladder and who gets squeezed? Because here's a thing. Most of the AI conversations is either too small potatoes, like frankly will chat to be tea, take my job. That's real, but kind of a small question in the grand scheme of things. Or it's too sci-fi, AGI, robot overlords, the singularity. What I wanted for this episode is a cleaner map of the middle. The real-world mechanics of an AI-driven economy and what that means for work, business power and policy over the next few years. That's why I invited today's guest, Professor Enendo Gosh of NYU Sterns. He runs NYU's Business Analytics and AI program and he spends a lot of his professional life in the world of big tech, thinking about competition, market power and what regulations should actually do in practice. He also corrode a new book called Thrive, Maximizing Well-Bean in the Age of AI. Now in the conversation, Enendo offers a simple but very sticky mental model. The AI economy has a stack. Not just a stack of apps and chatbots. He starts at the bottom, the foundation, energy, chips, cloud and data centers and then then large language models and finally the consumer facing applications. And once you see that stack, you kind of can't unsee it because it raises questions about consolidation, dependency and which companies get to dictate the terms. Then we go where everyone eventually goes. Jobs, not an vague AI will change. Work sense, he lays out a job ladder of view of who's likely to be okay, who will likely need to reinvent themselves and why we may be headed toward a world of portfolio careers where a lot of people don't have one stable role, but instead multiple roles stitched together across a week or a year. And yes, we do talk about who is likely to get squeezed. We also get practical about where AI is working inside companies right now and where it's failing. Finally, we turn to regulation, anti-trust, IP and copyright. Disclosure questions like, should consumers know when content is AI generated? And even the geopolitical layer of how the US, China and the Gulf states are thinking about open source models and strategic dependence. This episode is not about hype or doom. It's about a blueprints of AI and the AI economy and what that could look like. So with that, please enjoy the conversation with Professor Enendo Gosh. (upbeat music) And Indo, welcome to AI Curious. - Thank you, Jeff. - I'm so excited to speak to you because you are head on tackling one of the biggest questions that everyone has on AI. And most people kind of look at this in small ways in a narrow sense. Oh, how will AI impacts this piece of the economy or this kind of job or their own job or what have you? You are full-throated, answering and exploring a question of what does an AI-driven economy look like? What is that blueprints? So let's start there. You're giving speeches now and giving a talk on the AI-driven economy blueprint. So let's start at a high level. What at a 30,000 foot view does an AI-driven economy look like? - Yeah, again, thank you for having me, Jeff. So I think this is one of my favorite topics, like you mentioned. And I was trying to help you visualize what this economy looks like using a tech stack like structure. Okay? So think of a five or six layer tech stack where the bottom most layer is the energy centers. Okay, so the energy centers that power the semiconductor chips, which in turn powered all the data centers and the cloud computing infrastructure, which in turn powered the large language models, which in turn powered the final set of applications. Okay? So think of this as this five step process. So the consumer-fressing part of this is the AI application. So any application on your phone, on your desktop, et cetera. But behind that, now we see below that are the large language models. Then we have the data centers and the cloud computing, seven kind of chips, and then finally the energy centers. So that's the high level tech stack. - That makes crystal clear sense to me in the AI worlds of what that landscape looks like. I guess the question is to what extent is the rest of the economy, as we know it today, is that also reorganizing around that tech stack, or is this the economy fully transforming into this tech stack? But there's all the verticals of law and accounting and food production and shoe manufacturing. Where was that fit in this kind of like the tech stack framework? - Yeah, so think of like pick your favorite three companies, Citibank and banking, you know, Macy's as a retailer and maybe Delta as an airline company. They're all sitting at the very top layer at that application layer. So when I open the Delta app or the Macy's app or the Citibank app as a consumer, I'm essentially operating at the application layer. But what is happening, the tech stack I described is it's pavering everything that's happening in that application layer, but also pavering everything that's happening inside the organization. Okay, so when Citibank is saying, okay, let me try to figure out who's the right audience to send this credit card off or two. Or Macy says, who do I send this promotion to? Or Delta says, you know, what is the best, the most optimal route between US and, you know, these five cities in Europe, right? All of those questions are being answered using a variety of data and applications. And sitting behind all of that is that infrastructure. So you have the cloud computing, here are the semiconductor, through the data centers. Some are obviously using minimalistic resources, some are using a lot of it, but basically that's where the, that's where the, that's how companies are stacking up, right? So we only get to see the very first layer as consumers, right? We only get to see the first layer. But if you are somebody inside Citibank or Macy's or Delta, you know that sitting behind your entire organizational application layer are all those things behind that. There's some large language models powering your internal processes, there's some cloud computing that is powering the large-long in models, and there's some chips that's powering that, okay? A lot of them are not necessarily using things that they have built. So they're outsourcing everything to a handful of companies, Microsoft, OpenAI, Amazon, you know, Google, Meta, okay? So you see that the significant amount of consolidation happening, essentially a handful of companies have a presence in each of those layers. So they have a presence in chips, cloud computing, LLMs, okay? So we don't get to see those companies behind Macy's, Delta, and, you know, Citibank, but they're there and they're essentially powering this economy. Interesting. So if I hear you right, that is a pretty fundamental sweeping change. Every company will have to reckon with and to reorganize themselves in a way, if this framework holds, that their entire operations are lying with the idea of, okay, you have chips, you have. I mean, obviously Delta probably doesn't have to worry about the data center, someone else doing that, right? But they certainly, at the very least, LLMs and the app layer, they need to think about what their data sources are, where the data is living, is it on premise, in the cloud, and then the apps. So it might get it right that pretty much every, whether you are an airline company or a shoe manufacturer or a law firm needs you fundamentally reorganize yourself along those lines. Yeah, assuming, yeah, absolutely right. You're assuming that you really want to get on this AI first AI transformation journey. You really want to use data to power your internal decision making. Then you have to figure out beyond your internal applications and infrastructure, what else, who else can you figure back upon to have the remaining infrastructure? So like you said, Delta or the shoe manufacturer or the restaurant, a large restaurant, they're not, when we're building foundation models, they're not building chips, they're not building cloud computing. But they have to figure out, do I go with Amazon? Do I go with Microsoft? Do I go with Google? Do I go with meta? Because our, you know, OpenAI and video, right? So they have to figure out like, who are my downstream partners? Okay. And while there are, you know, choices, it's also interesting that those four or five names seem to appear at each level. Okay. So in theory, like this is entirely hypothetical. Citibank might go to Amazon and says, hey, look, I have for my internal applications, you know, credit card, retail banking, new derivative products, right? I don't want to build my own large average model. So I'm going to leverage your models. And Amazon might say, sure, you know, we can give you a license for that. but you will also have to use my cloud computing, AWS, you have to use my chips, trainy m chips, you have to use my energy centers back in small, small, down in Texas. So, in theory, there's a lot of power that some of these companies have courtesy of the fact that they can vertically integrate the entire textile. Okay. Yeah. And so that's where we are at the moment. So this definitely perspective of from the corporation, which is super important. Let's kind of widen the lens and look almost macro economics. So what the economy looks like from perspective of jobs, of what human beings will be doing. I think a question on everyone's mind is, okay, how many jobs will be impacted and to be even more like fundamental about this? Will we have jobs? Well, I mean, there is one scenario, right? They're people are, okay, if people are trying to build a one human being led billion dollar company, the logical follow up to that is that there will be less jobs, right? So as AI does more and more of our work, in your minds, what happens to jobs for most people? Yeah, that exactly is the 800 pound gorilla in the room, which as you said, everybody, my students asked me, the sea suit folks asked me, everybody asked me. Obviously, it's hard to predict this certainty, but I can tell you maybe what is happening now, what is likely to happen in the near future, right? It's hard to pray beyond, I don't know, two years, but here's where we are. At the moment though, and it's not necessarily good news, at the moment, there are more jobs being displaced or disrupted than being created. I think at the moment, unfortunately, the net is negative. Now, this is not just because of AI, just a week here, and some of that is just macro economic forces, you know, rate cuts, you know, inflation, etc. But AI does have a AI does give public company, you know, sea suit people the ability to say, AI, look, AI is automating a lot of jobs, we need to, instead of hiring 10 people this week, you're going to hire four. That is happening. That's definitely happening. Now, you know, we'll talk about skills maybe after this, but at the high level, this also brings up the question of what kind of maybe careers or, you know, knowledge, workers, or skills should the next generation of people gravitate to so that they can actually save their jobs, right? Or so that they can actually leverage their newly acquired skillset to go and find those new jobs. There are going to be new jobs being created. They're not as many as the ones being displaced. So at the moment, it's net negative. I think one of the really important skills that I tell everybody that I know, well, is you need to know what Gen AI, which is this new shiny object, you need to know what Gen AI cannot do, not what it can do, but you need to know what Gen AI cannot do, why it fails, because it does fail, and we can talk about those, it does fail, the white fails, because the more you know when it fails and white fails, the more valuable you become to the organization. Otherwise, you're going to be replaced by some Gen AI product. Yeah. I think it's a great advice, and I want to get that in a little more depth in a minute, but I do want to, I fully appreciate and agreeing. It's hard to predict beyond two years, but you're a very smart thinker in the space. I would actually love your thought anyway. I promise you in five years, we're not going to come back to you and play this tape and say, oh, you said this. So if you had to, I'll make a video show. Yeah, give a sense of like, are we, are we all working half time in the rest of our times on a beach? Are we, are we, is 30% of the current workforce completely unemployed and we have some kind of UBI to fill on the gaps? Like, what does the world look like to you? Yeah. As AI gets more and more prominent? You kind of pretty much read my mind. Here's what I say, that top 10% based on any reasonable metric of skill sets, education, experience, etc. They're going to be okay. They will have, they will continue have that single career. Okay. So you started off as a banker in Citibank, you will end up as a banker as a bank, right? That one career. And they're going to be work, they're going to be working five days away. The next 30 to 40%, pretty good smart people. They will have to learn to juggle multiple careers at the same time. Meaning, I see a wall where Monday to Wednesday, you may be a banker or you are a yield management specialist in Delta. But then Thursday, Friday, you have to pivot to a completely new or relatively new hat and a new industry, maybe in a new company because Thursday, Friday, your job is going to be given to somebody else in the society who also needs a job. Okay. So I see, and this is obviously required, it's going to require a lot of government intervention, etc. But the next 30, 40% will be okay, but they need to juggle two or three white college jobs. The next, I would say, 30 or percent will need to juggle not necessarily white college job, but more like freelancing, creator economy jobs. Okay. So they're, they may be skilled enough, but they're not going to be enough white college jobs to fill that that next year. And unfortunately, I do see a world where a good 10 to 20%, maybe even slightly more will struggle to find stable jobs. And you know, this is a slightly slight deviation that I always tell people that one of the most important skills in life is financial literacy, systematic investing from a young age, so that when that day comes, that is going to come to all of us, right? We have some sort of stable revenue just from investing systematically in the market. So that I think naturally gets us to the concept of reskilling in one of a three, the three planks of your your blueprint economy talk are reskilling regulation and transformation. So we're kind of covered transformation of what this wild world can look like. How in your mind, do we embrace reskilling to get us to a world where people are able to make that transition to like, to okay, Monday, for the top 10%, there, there's power movers within city bank and flourishing. The next 30 is percent they have their and also if I folks are releasing this freak daily, oh my god, I need two jobs. It sounds like if I hear you right, it's not that people will necessarily be working any more. It's that they're pie chart, which was once like 40 hours a week for one company, they might it might look like 60% of the pie is is Delta 40% they're now doing something in some creative agency, whatever it is, but they're still working 40 hours a week, but split up different ways because but once took them five days a week at their original job, they now do in three days, thanks AI efficiency. So yeah, the human we're all working not necessarily harder, but accomplishing a hell of a lot more. Am I getting that right for in base case? Yeah, very much so. I think and the reason like you said correctly is you know, the productivity gains the efficiency gains from AI, which will continue to explode, will force organizations to hire you know fewer people and out of ones who would they hire they will be working fewer hours in that company, but they can make up for the disposable time by working somewhere else. So there's going to be but again, it's like no, those the next 30 to 40% people, these are still super smart, super competent people. That's why they will they will be in demand. That's why like two organizations will be willing to hire and work with them part time, right? So so that's what I'm getting at. Yeah, got it got it. So so what reskilling is is a what does that look like? I'm guessing to some extent that there's some onus on actual workers that on their own. Okay, you should be exploring AI on your own, watching YouTube videos, listening to AI curious podcasts, educating yourselves, but also and also I'm guessing there's some onus on the companies to help up skill workers, but I'm guessing I just got my gut feel here knowing nothing is that that's not going to be enough. There needs to be especially for a big chunk of the economy and population that is not really thinking about this companies that are not investing in this company. So I have no freaking idea where to start there to truly if the transformation is as widespread and profound as you suggest and believe me, I think it sounds quite plausible, there needs to be a much bigger effort at a societal likely government's investment level for reskilling. Am I warm there and what is the risk that we look like? Yeah, no, so I'll talk about Singapore's Marshall plan, right? So I mean, we were the first in the US to move the Marshall plan for the Second World War. Singapore, after what they saw happening with AI, has an equivalent Marshall plan where they actually giving two thousand dollars a month to every citizen Singapore to upskill and re-skill and re-pool themselves in AI, right? Fantastic idea. Now they obviously have a lot of resources, they have the pockets so they can do this. I think I do foresee a world where this will become necessary. It's sort of like UBI, but in the context of UBI, be I typically the government would say, look, here's a couple of thousand dollars, you can do whatever you want by your groceries. And in Singapore said, we demand it, make sure that this is only spend in AI upskilling and retooling. I think some version of that, I do see what would be necessary for governments, different parts of this, you know, and then the world to institutionalize that. Now that being said, you back to your question about rescilling, right? What should organizations do? There are obviously different choices. So there are formal educational programs in universities, but those are costly. Then there are semi-farmal programs, certificate programs that are not degree programs. There are far less costly, but they also don't give you the whole pie, they give you a big piece of it. Then there is self-learning. So I feel like a lot of this will be initiative driven by the individual, right? So obviously if you have the resources, you want to get as much exposure as possible on a formal, semi-farmal basis, otherwise you do self-learning. And I think that is where individuals who are deeply serious about upskilling or tooling themselves, they will have, they still have choices. There's free online resources. There's also semi-farmal certificate programs, degree programs. You have to figure it out like what books for you under the circumstances. One other thing I'll say and then we can pause and build on this is, I am not just emphasizing the hard skills. I actually think that as AI becomes even more ubiquitous and we gravitate between automation and augmentation, it is the softer skills that most of us have, right? Those become more important because we'll have to be using our judgment, our creativity, our inference and all of that. Jayne, it tax back to the point you made earlier, which is thinking about what at Gen AI, at least now, cannot do. And influencially, that is the human stuff, that's judgment, that's compassion, that's empathy, that's listening, that's all of those, it's connecting seemingly disparate ideas and schools of thought, making a surprising connection between some indie folk song you heard five years ago and some chemical experiments and doing something cool with that mashup, right? So, yes, totally agree with that. Before we get to the third kind of plank of your talk of regulation, I would like to explore a little more of what industries and jobs you think are most likely to be transformed sooner and what's most vulnerable. I know that the easy answer is, well, all industries, all jobs. I use that answer myself. My running joke is, if you were a good-hunter, good news, your job was not impacted by AI. If you were not a go-hunter, it almost certainly will be, right? I'll do it. But I'm guessing there are some industries that are especially susceptible to massive change very soon. What in your mind you see is kind of the industries and jobs that are most at risk of being radically changed and jobs being radically cut. So at a high level, I think I would say B2C, you know, is typically two to three years ahead of B2B, right? So any consumer-facing industry or business is far more likely to be ahead in terms of AI transformation and AI first strategy, which also means disruption, displacement, and somewhere between automation and augmentation, right? Like automation, we typically think is sort of, you know, not good. Augmentation is great. AI can do both, right? We are definitely seeing that in the B2C world, which means any consumer-facing, you know, those examples between airlines, retail, you know, banking, all of that. It's already happening, right? We are also seeing the advent of something called AI agents, agent-to-AI, where basically it's literally extremely smart software programs or that function like ages. They have their own autonomy. They can make decisions. So agents are ones where they're not just retrieving information, but they actually execute on an action. You know, if I wanted to book like tickets on booking.com through OpenAI's charity, right? We already seeing a world where that's happening in small steps. It's happening. I can complete my entire airline hotel booking because Expedia or booking.com or anyone of these, you know, online gatekeepers have essentially been using OpenAI's APIs to integrate their business. I don't have to get out of it. So those things are already happening. So, you know, I think B2B will catch up in two to three years, large insurance companies, pension funds, manufacturing companies. They've all been exposed to the importance of AI. So those conversations are trickling down, but with a two to three-year lag compared to these B2C companies. That's really interesting. What one thought I just had is we tend to look at this as which industries are most affected. I'm wondering if we could slice it differently by size and size of company. What I'm going with is it just from my layperson's perspective, it seems like some of the massive companies that have ambitions to invest heavily in AI and automate a good chunk of their workflows because they are so big, because they have so much often bloats. They struggle to really fundamentally do it. We're at small companies that are more nimble can take more chances. In a very simple example, I think most listeners will recognize its chatbots and customer service chatbots. It's amazing to me that when I use voice mode on chat GPT, it is awesome. It sounds like a real person. It sounds like the movie "Her" is brought to life for better or worse, but it sounds very real. When I call, we keep picking on Delta Airlines. When I call Delta Airlines or almost any big company, I still feel like it's the year 2013 saying, "If you would like to discuss your reservation, press two." If you would like, "What the hell?" Why can't they implement the good stuff? One way to be asking our smaller companies being quicker to effectively implement AI and the core of the area is there's a job vulnerability in smaller companies as well. It's an interesting point. I would say, "Here's a trade-off. The trade-off is between data and workflows." What do I mean by that? Smaller companies have smaller datasets, so less juice to power any AI transformation, but they also have fewer workflows that have to be automated, which means any agentic AI unleashed on the infrastructure has a higher chance of succeeding. Larger companies, they have a lot of workflows, which means they have to automate and code those workflows, which is a Herculean task, but they also have a lot of data. The algorithms in larger companies have a lot of juice and horsepower because they have a lot of data, but they also have a higher bar because they have many more workflows internally. I see this quite, really frequently. This summer I traveled quite a bit. I mentioned this earlier from on my book, Ten Countries, Three Content and Sixteen Cities, talked about two dozen companies really figuring out when is AI succeeding and when it's failing. To my surprise, I shouldn't have been surprised, between 40 to 45% of agentic AI projects are failing. You know why they're failing? Because these organizations, they grab this new, shiny object, but they forgot the basics of coding the workflows before they unleashed the AI agents internally. Now you have the situation where really smart people are running a really large public company, but they haven't done enough background work to code and automate the workflows before they unleashed the agent. The agent goes, "Okay, so here's an example. In case your audience is wondering what does this all mean." Let's say the procurement division has to place a large order for these products. Typically the workflows is the procurement manager needs approval from the head-off supply chain, who needs approval from the head of manufacturing, the head of finance, etc. The buck stops at the CA4. There's a workflow. Now the agentic AI, when it's unleashed, it needs to be told, "This is the order of decision-making, procurement manager, supply chain manager, ops manager, CA4." Without coding the workflows, the agent will just go crazy. And I saw this this summer across the planet, right? Large public companies just went and grabbed the new shiny object. So to sort of summarize, I would say, "That's where I see. I see larger companies. They have a lot of data to power the AI agentic AI, but they struggle with the workflows." And that's a problem. So while we're on this, that goes to aggression real quick, what given your kind of in the fields exposure and visibility to what so many companies are doing, where do you see AI most effectively used right now? So the percentage of agentic workflows that are working, what are the most promising use cases, what workflows are actually moving the needle, and where do you see companies struggle and, "Ooh, this is actually, they thought this would be super easy and turns out it's a nightmare." Yeah. So assuming the organization has done a good job or a great job in automating and coding the internal world, I think we are seeing a lot of success with Agent K.I. in doing the simpler task, like information retrieval and decision execution. Okay, so in the procurement example, right, this Agent K.I has to figure out what is the best lowest price among those 10 suppliers who are bidding for my budget, right, figuring that part out, like who's not just the lowest price, but also lowest price at the highest quality maybe, right. And then figuring out the decision making goes in that hierarchy that I talked about. If you have coded the workflows, we see a lot of success with what I call simpler tasks like this or you talk about customer service chatbots in contrast to some of the airlines who are not doing a great job. There are some really good examples. I see the in credit card and banking, for example, or in brokerage accounts, for example, you know, investment companies, they do a pretty good job with like customer service chatbots. And it's really a function of how much effort you are for in thinking through what these agents will do. Yeah, yeah, yeah. Got it. Let's segue to the third kind of leg of this stool of what the economy blueprint looks like and how we get there. And that's regulation. The shift we're discussing is about as profound as you can imagine in the history of our economy regulations. Obviously a hotly contested topic, pretty polarizing. What in your mind is the most sensible playbook for regulation or put differently. If you could wave a magic wands, what's the regulation you would want to see? So I think it's a great topic to think through with different context, right? So because the world of AI is so massive and pervasive, they're different, you know, context in the different regulations and relevant. So what I mean by that is let's first talk about that AI tech stack I talk about, which is, you know, what is AI economy looking like, right? So energy centers, chips, cloud computing, LLM's AI applications. Now one of the possibilities there, I just talked about that Amazon example, right? Like city bank goes to Amazon, but Amazon says you have to buy everything from me, what equal integration. Now in economics, what we have seen in prior anti-trust regulations is what equal integration may not necessarily be a bad thing because sometimes what equal integration increases product quality. So if service quality, a product quality improved, that's great, but if it comes at a much higher price, well, that's not great. So regulators have to figure out that hey, is there sufficient competition in each of these layers so that no one company becomes too powerful and by vertical integrating and forcing its will on its employees. So that's like one kind of regulation that's happening at the that's that is probably going to be happening in a couple of years from now, right? It's early days. The second kind of regulation is, for example, let's say things like in the context of large language models, there's been many high profile lawsuits about does the content that these large language model companies trained algorithms on is that fair use is a copyright violation of failures. And the jury is still out there on this one because I have seen instances where the courts are sometimes ruling in favor of large language models sometimes basically they're saying, yeah, it's fair use sometimes they're saying, that's a copyright violation. And specifically that so that the New York Times versus open AI being kind of the most arguably the most prominent of these cases, am I kidding that right? Exactly, right. So there's that is an anthropic case. There's lots of these cases now and I don't think we have a clear cut regulation on like what is considered copyright violation what's considered fair use. So I feel that they should be something along that line coming up soon. A third kind of regulation that I think will happen soon or later is the effect of inadvertent effect of AI on consumers like what I mean, but that is so let's say a brand launches advertising campaign created entirely by AI right all the ads are by AI generated. Is it mandatory for the brand today to disclose that this is a generated. But I don't think so it's not mandatory today it's optional. Should it be mandatory. Maybe maybe not it's not clear to me, but I think there's a question there that should consumers be have complete transparency on how companies are creating content using Jenny and maybe it's a simple question, maybe it's not but I think I don't think it's a simple question. They answer is quite complicated because one of our most research projects recent research project showed that AI generated ads are actually outperforming human generated ads. But that's when people are not told it's an AI ad when people are told that this is an AI generated ad then they actually respond less favorably to that compared to human ads. Yeah, it's a really complicated question, right? I think like the the writer and creator and human in me my knee jerk reaction is of course people should know if it's humans or AI created that, but yeah, I don't know if I'm if I'm in a driving a car. And I do I need to know what component of the car a human design is certain grid on the I'm not a car guy some making all of this stuff right but like the steering wheel fabric was what a human design where the fabric goes and where was the computer or AI used to like help design the material goes into the car right. If AI is useful at creating this good or service for me, doesn't matter if a human or AI created it right and so it's a great area advertising right I think then again if I saw if I was reading a book a novel, I would sure as hell want to know to the human being right this novel or the AI with this novel. If I'm watching a movie and all of the actors, we're where every single actor is an AI actor it's all computer generated. I probably want to know that so would I want to know for advertising it's I could I could see it argued both ways it's a pretty fascinating topic. Exactly I think that's why it's not a trivial question you know regulators at some point may have to wait and it's not that you know like think back to like sponsors or chads on Google 20 years from now right 20 years before they were initially there was no regulation then they were told that very clearly you have to disclose if something is an organic listing on sponsor that right. Yeah so I think I think there's there's potential for that sort of regulation to come and then maybe just on the topic regulation one last thing I was going to mention is you know this is a more geopolitical global question which is should American companies be prevented from working with open source large language model companies that are not American. And it needs to be explicit much of that is happening in China right now there's a great deal. So when I was in Riyadh last week and I mentioned this earlier before we started talking is the Saudis are telling me that look we have a very interesting choice we can work with you know all American Western companies getting 100% quality at a very high price. Yeah or and this is literally their words we can work with one of the open source companies paying only 20% of the cost but getting 90% of the value and they're referring to the Chinese open source companies. The Saudis are investing billions in this space like the Emirates in UER I feel like the world is a very interesting junction with juncture which is do they go and partner with the West or do they look East we don't know. Yeah no it's it it's hard that's one reason why I find this topic so fastening is all the elements it touches and ranging from kind of like the philosophical to geo politics and world power so as we wind down there's one book that no was written by a human and that's and that's your book tell us a little bit about because you still are book tour tell us what you about the book and what you want listeners to take away from it. Yeah thank you for asking so this is a book I co-authored with another professor Ravi Bob now who's a professor Minnesota the two have been of us have been tacked in in various engagements for 20 years he finally decided to write right a book together. So the book is actually. Sorry it's called thrive correct yeah the book is called thrive yeah thrive exercising the well being and if I don't mind I can show like this what it looks like. Yeah the book is about telling or not hitting the positive side of the eyes you know so three years ago when we started writing the book we saw a lot of negative it around AI right it's going to take all our jobs away it's going to be unfair it's not transparent it's not fair no it's sure there are some negative we talked about the negatives too but it's not all negative there's a lot of upside to it and so Ravi and I decided you know based on our collective experience we wanted to tell the world. What are some of the positive examples the upside of using the AI and so we talk about not just work a career of finance but also like relationships. Health mental health physical health each chapter goes deep into each of these purely research and science driven examples and the other thing is about re-skilling up. So one of my favorite chapters in the book I mean they all my favorite but one of my favorite is chapter one where we tell the audience in this AI transformation journey here's how you retool rescale yourself using this framework called the house of AI okay so the house. - House of AI is a very simple, yet a very powerful, compelling framework that can help individuals and organizations navigate this AI transformation. I will, I will spare you the details, anybody interested in reading the book, but I think that based on what I've seen so far, at House of AI has really resonated with both individuals and organizations. - You have an interesting role as amongst the many hats you wear, you also run the AI program at NYU. So you have a unique vantage point of hearing what students are thinking about AI. I'm curious, what's the temperature right now of from your vantage point of students and excitement versus anxiety level of AI? How has that shifted in the last few years? - So it's been extremely positive, right? So our program, it's a master's of science and business analytics and AI. So it's the acronym is MSB AI. It's running the business school, and as you mentioned, I run the program. I've been running it for almost eight, nine years now. You know, I, I'm really passionate about this program because I feel like off the different hats, maybe this is one hat where I can make the most difference or the most, you know, in terms of giving back to the society. And I'm really, really passionate about making sure that every course, every skill they get out of the program is relevant based on my real world expertise and experience working with companies. So, you know, when I traveled this summer and talked around 25 companies, I learned five things. I already implemented them in the program using the projects or the coursework or additional lessons. So I feel, I feel we are in a great shape. I see the students very excited. You know, in the short run, in the job seen in the US, in general, is not the best. So in the short run, there's gonna be some hiccups, but overall, the feedback from alumni in the past and the current court is very, very positive. So, you know, because I feel so passionate about this program, I just wanna make sure that I, whatever knowledge I get from wherever I implemented in the program. - Awesome. And final question for you. Last bit of advice you give listeners on how to embrace AI in a way that makes them more human and helps them flourish, as opposed to being replaced. What would you, kind of finally, - Yeah. - Take away from listeners. - So, yeah, read my book or read our book. (laughing) But, more importantly, immerse yourself in all these Gen AI applications and products that are being released. Don't just stick with like one, you know, product to one application. So if you're using charity, you also try Gemini and GROC, like try multiple things, look for differences. Like, ask the same, give the same task to five different LLMs and look for where they are similar, where they're differences. Because, like I mentioned, the most important skill for the next generation is figuring out when Gen AI continues to hallucinate and confidently give wrong answers. - And why? - Okay. - Yeah. - Because the more you can position yourself in the company like, I actually, everybody knows Gen AI reasonably well, but very few people actually know when Gen AI fails and why it fails. If you can be that one person, your job is far more stable and secure than everybody else. Amazing advice, love it. And Indio, thank you so much. Really enjoyed talking to you, and congrats on the book launch, and really enjoyed. Thanks again. - Thanks so much for having me, Jeff. Really enjoyed chatting with you. - Well, here you have it. Thanks again to Professor Enno Gosh. Thank you to you, dear listener. Happy new year. If this is your first time, AI curious, please consider subscribing, rate it five stars. So do a friends, all the good stuff. Thanks again, and see you soon. (upbeat music) (upbeat music)