and it's my pleasure to introduce our moderator for this evening we're thrilled to welcome Tim Phillips as a journalist Tim has written about business technology economics and Innovation for more than 100 Publishers around the world including the Wall Street Journal Europe the international Herald Tribune the times and Sunday Times The Observer the Telegraph and the independent uh the appendant among others and Tim also hosts the weekly Vox talks economics podcast for the for economic policy research and so with my thanks to everyone for joining us for this evening for this event Tim I'm pleased to hand it over to you to introduce the panel and kick off the session well thank you very much Elizabeth and uh welcome everybody um in the next hour I'm sure we're going to have a pretty good discussion there's some strong opinions going around so uh keep listening don't miss anything now we know advances in AI are going to fund Fally reshaped jobs and industries and even economies as well but just how significant will those changes be for the economies of Europe the Middle East and Africa and how do you actually figure out what impact AI is having now I've got some research here from Goldman Sachs says two-thirds of jobs in Europe and the US are exposed to some degree of AI Automation and a quarter of all work could be completely replaced by AI so how should policy makers companies and workers in AIA countries respond our panel is going to be discussing the impact of AI on the economy and on the firms and workers in it so who are they well first of all we have a veteran of the economic Outlook Series so uh that is Professor Randy Kroszner the Norman R bobbins professor of economics at Chicago Booth and Randy previously served as the deput Dean for executive programs at Booth he was a governor of the Federal Reserve System from 2006 to 2009 and currently he serving as chair of the financial research advisory committee for the US treasury's Office of Financial research he's also a member of the bank of England's Financial policy committee so welcome Randy what's going to happen to interest rates in the UK uh I don't talk about that I he's only kidding I don't expect you to you're very welcome now also we have Julia Lane a professor at the NYU Wagner Graduate School of Public Service and an NYU proval fellow for Innovation analytics she co-founded the CID initiative which aims to transform the way government's access and use data for the social good and Julia previously served on the advisory committee on data for evidence building and the national AI research resources task force Julia welcome as well glad to be here thank you and finally Chad Syverson the George CCH distinguished service professor of economics and a research associate of the National Bureau of economic research and Chad is also co-author of the intermediate level textbook microeconomics and his research on the interactions of firm structure Market structure and productivity have been published in several top journals and has earned multiple National Science Foundation Awards Chad as well thank you Tim it's good to be here now uh we do have uh the Q&A open tonight so you will be able to uh submit questions I would warn you that quite a few of you have submitted questions when you registered as well so we do have a packed agenda I will do my best to reflect all the questions are coming in but I can't commit to asking all the questions that you ask as I say we will do our best to answer all your queries though uh first of all I I'd like to ask each of you on on the panel a sort of introductory question uh Julia can I I come to you first um I quoted some statistics at the beginning when but when we discuss AI it sort of represents a bag of all sorts of different Technologies at the moment do we really know what we mean by AI at the moment and can we easily capture and me me and estimate its impact um so I'll I'll use slightly less colorful language than I used in the opening but I think the polite term is those numbers are I think one might call them bollocks is that the correct pronunciation I think it probably is yeah well uh probably an antip podan point of view at least so the um the reason that I think the they are in a more academic term not robust is um when I served on the national AI research resources task force and you can go and look at the report um we were being charged by the president and Congress there were 11 computer science types and then there was me the social the token social scientist and we were asked to um figure out how much the US uh government should invest in research resources to stimulate AI for uh Workforce Innovation diversity and so on and uh you know everyone's saying well we're going to spend lots of money which is typically what the task force is and I said well wait a minute do we know how much is currently going on in AI how many people are involved in the workforce and if we're going to say we're going to double resources in X where are we starting and how are we going to know when we get there so that's a perfectly reasonable thing to to do right so um the um office of Science and Technology people staff who were policy staff who were really good went off and tried to figure out what numbers they could come up with that would be uh reasonably robust so if you go on to the research resource task force uh we say there have been 2,000 computer science faculty members who have published at least one AI related P paper there are um the share of new computer science PhD recipients spending an AI increased from 19 to 25% between 2019 and 2020 for a total of 442 so I was of course curious to see then where did those numbers come from MH because those are also some of the numbers that feed in to the Stanford AI index and into the oecd index so here's when you go and take a look at the Stanford AI index and I've talked to Eric Bolson and others about this and we're hosting a workshop that recognizes the there these numbers are problematic here's the way in which they come up with them they go and they take a look at full text Publications published on an open science website called archive which people put pre-prints on and they Tagg the papers with whether they AI or not so it's self-identified they estimate a model using the terms used in those papers which can go back a number of years so obviously archaic like chat GPT doesn't show up for till later and if you take a look at the list of ter terms they are also not robust uh we're going to use that as code for bollocks going forward okay so um then they what what they do is they take those terms the model that they've estimated they apply them to titles and abstracts from publication data sets like web of Science and demen and that's where they get the numbers from so the these claims that are being made are then made from feelings or interviews with business managers which kind of feed on other there's no hard numbers other than what we had described before there had been some efforts made by Rob seamons and um and Eric where they've asked firms a question about whether they use Ai and then they've made an assumption about the workforce impact but that's about the the the long and the short of it so there's a lot that needs to be done we've done some work in that area I'm happy to fill you in later but uh that's that it's kind of my uh junest view of of the value of those estimates okay so if the big estimates uh that we're making those big aggregate estimates are um what were you saying maybe not Rob not robust yeah Chad I I come to come to you um you uh do research on productivity and we've all been searching for productivity for a decade and a half now is there evidence that this AI might begin to solve the productivity problem well I hope it will uh as to the evidence you know Julia raises I think some very fair concerns about direct measurement of AI things I've done along with Eric berelson who Julie mentioned and Daniel Rock uh say the problem's even bigger than that a lot of the problem productivity measurement isn't even about the AI Investments themselves it's about the Investments and intangibles that go along with it and we can talk more about that later um but modulo those bar real uh measurement concerns I think there is a data CA datadriven case to be made that maybe we're starting to see you know an emergence from the productivity growth slowdown that's been going on for 15 plus years that you mentioned I think it's worth just quickly setting a context for everyone here that the world has been in a productivity growth slowdown for those 15 plus years now almost every economy uh regardless of income level or Geographic traffic location has seen a productivity growth slow down and that's important because productivity growth is the speed limit on economic growth and we can't get sustained uh growth in material living standards without productivity growth so if that falls uh so does economic growth and to be honest I don't think we really know exactly why productivity growth slowed down 15 or so years ago but we've all been looking for signs of an emergence from that Slowdown and I think like I said there is some data driven case to be made that that might be happening but we're we're far I think I might be on the more optimistic end of things and I still would admit we're far from knowing uh that the productivity growth is slowdown is over so I think that's a really important context for the discussion we're having here today and um Randy we have um we have a lot of excitement certainly about Ai and uh whether or not that is Justified or whether that's hype is it too early to uh be able to think about macroeconomic impacts of AI for example increased demand for investment sure I think it's a very important issue that um that I think hasn't got as much attention as it as it deserves um we certainly don't know exactly what Julia and uh and Chad have said we're not sure about how big a deal this is going to be or um how transformative it's going to be but what we can do is run out a scenario that if it is um you know if you are optimistic and it is going toh to be a big deal I mean certainly we've seen a lot of excitement in the stock market over a number of companies that have been either producing or using uh AI Technologies and that suggests that uh there might be a lot of investment that uh would be going into this area not only by the chipmakers but also by the people who would be buying the chips and uh programmers etc etc that could affect many many Industries so in some sense um you may recall that a few years ago there was a lot of discussion of so-called secular stagnation that um uh that was sort of a way of talking about the productivity slowdown that uh that Chad was was mentioning that uh we've kind of run out of ideas we're kind of uh people are not very excited about investment investment is not having very high payoffs and so we just kind of you know barely kind of chug along obviously this is just the opposite of that now under the secular segmentation hypothesis well you'd see interest rates of move down because demand for investment's not very high because ultimately you think of kind of the the the global Benchmark interest rate is being uh defined by supply and demand no surprise from a University of Chicago Economist you're going to hear about supply and demand and so if there's a fall off in the demand for investment all the things being equal um where supply and demand are going to meet be at a lower lower price that is at a lower interest rate but if you got a lot of excitement about investment in in Ai and either in kind of the fundamentals of uh creating the um and training the those models or in using them that could lead to a substantial increase in investment over the next few years and all other things being equal that will lead to uh to interest rates moving up and this is a um you know very big debate right now about well after central banks start to uh normalize normalize interest rates after the uh after inflation comes down where will normalization be will it go back to where things were um over the last decade or so at very very low levels will go back to uh what it was uh prior to the the global financial crisis in this scenario it means that it'll go back something to closer that's uh the the long run level be closer to what it was before before the financial crisis rather than the the very low levels that we've seen recently and then obviously that has an impact in thinking about what's the valuation of startups uh who might be coming into the AI space etc etc so it's uh I think a very important and interesting topic to think about we don't know the answers yet but it's important to work out these scenarios particularly if you're thinking about investment in in this area we got to think about what the Benchmark interest rate is going to be over the next over the next decade thank you very much now we're going to be looking at the impacts as I mentioned in the introduction on workers and firms and on the economy so we start with workers first of all there is a live debate at the moment that's going on about the impact of AI at that uh micro level in the workplace and whether it is going to replace tasks um augment workers or whether it will replace jobs entirely uh Chad what what do you think about that what uh what's the research that you're doing telling you so it's early but I I think the kind of case studies that have been done so far where we really best think we can measure things seems to indicate it's more going to be more task-based than job based uh and therefore push to complimentary rather than substitutable um applications um and that it tends to actually help workers I guess economists would say lower skilled workers but lower lower at the lower end of the wage scale within an occupation so if anything and this is part of the big debate that you're talking about is you know is this it 2.0 where we're going to see even more skill bias technological change and in even further increases in inequality or might we see uh an actual compression in the in the earnings distribution it's very early we only have a few good case studies but I think those so far seem to indicate it's the latter that they that AI Technologies at least in the applications they've been used in seem to help the workers at the lower end of the of the skill distribution and therefore um bring up um productivity most among among those workers relative to higher skill workers uh will tell we need a lot more work obviously and we've hardly seen the entire scope of potential applications of AI so it's very early but I think the early returns are at least uh suggestive along that Dimension as I read them Julia what do you think because we have a question come in to say how would you even begin to measure the productivity effects of something like generative AI well obviously I've been thinking about that quite a lot since uh having been on that n Tas force and I don't think it's going to come as any surprise to Chad and and to Randy that I've gone back to thinking about what are the underlying Micro Data that you can work with with which you can work um so here's uh the way in which we are doing it um the the basic idea here is in some sense the challenges is that AI is neither an industry nor a scientific field so we don't have a statistical infrastructure that deals with it you know what happened was a 100 years ago we had sic codes which captured manufacturing and agriculture because that was what we did and we clustered firms in in those by that grouping and then about 40 years ago we started to say well it's a service economy we need to describe how firms produce services and so we clustered firms by how goods were produced how things were produced which is the service economy and NES code so Financial Services health services and so on but the problem we have now is that we're not really producing physical goods or services as much we're producing ideas it's very much in line with what Paul Roma highlighted endogenous technological change so the way in which uh growth is occurring is not so much investing in capital labor energy material Services Clems as the way in which they're being put together and I love Heidi Williams's quote that what Innovation does is it breaks the shackles of scarcity so what you're doing is you're combining things so how is that happening well we've got this massive industrial policy in which science agencies and other agencies have been asked to invest in ideas that's people and it's largely been at universities so the AI winter you know changed to um in NSF in particular but other agencies like doe and I'm using American examples apologies but we're in the United States um so they invested substantially in AI research and there was a 100-year report which Eric co-authored as well which was AI research is what AI researchers do so when I was on the N task force I was going how are we going to measure this F Fe Lee said um who developed imag and who helped pull us out of the N of the AI winter she said well the way in which you're going to find AI research is just just go to the conferences and the base AI researchers are the ones who go to those conferences that's how they communicate that's how they transmit ideas because it's the creation transmission and Adoption of ideas that we're talking about so that gives you your seed if you think ideas are embodied in people then you need to trace people so what you can do is you can see who's being doing AI work at the universities you can track them through the research grants not just them but The Graduate students postdocs undergraduates and so on who by the way we know have value not by ad hoc assertions but we're seeing them flow in to the into the labor market getting paid 300 sorry um uh 750,000 a year plus stock options so the firms are valuing those ideas so we create created something about 15 years ago called star metrics which is now um metrics which captures the HR and finance information that is spent on the grants at those universities all the people not just the principal investigators and then you can track them into the wage records of the firms in the private sector so you can see what their earnings are how many of them are being are clustered how many businesses they're starting up you can then identify the AI touched firms or the EV or any of the other critical and emerging Technologies like Quantum Computing or um synthetic biology so it's not it's technology agnostic approach that says okay what we're really looking at is not just manufacturing Industries or service Industries but what we're calling the industry of ideas that's how you capture what's going on and what happens to the other workers in those firms and what skill needs there are because you can link the administrative records in yeah thank you Randy I want to ask you about the impact of this industry of ideas on the workers within the firms there are uh uh several conceptions of how this is going to play out we see we have some research uh some a paper out from David otor who's uh speculating that AI May rebuild middle class jobs is how he he describes it um Don s moglu notably is um believes that the the direction of innovation uh driven by big Tech in AI may not might not be worker favorable what do you have an opinion on who winners and losers may be well exactly as my colleagues have said you know it's a little bit early to to to speak definitively about that but one thing one might do is is go back to history uh because we've had big revolutions before in u the the teens and 20s was an enormous revolution in terms of electrification of uh of plants the development of um of small uh um uh small machines uh the way plants used to work was that there was kind of one big power source was like a conveyor belt that kind of ran everything through the the plant and uh it was either the plant was working or not working if something went wrong with that conveyor belt but then of course with um the modern technologies that came in to be used in like auto manufacturers that you still had a a production line but people were using individual machines in that production line and electrification allowed for plug-in machines for that so there was a lot of debate in that period about what would happen to to jobs and the concern initially was that well it would be the um the lower skill jobs that would be eliminated by these machines but obviously that hasn't been the case and so I think you could make a more optimistic case that this is not going to be eliminating those those sorts of jobs exactly as Chad said it could be complimentary to many types of jobs and tasks simply making people more productive or asking people do things that are slightly different than what they were doing but not necessarily require them to have a PhD in order to be able to do it and uh and certainly we've seen these technological revolutions before not end in um you know with no jobs but you know with lots of jobs in the the postwar us period so I'm recently optimistic that this can be uh used in a way that would be jobs complimentary and increase productivity I mean for the optimists it may be that well people you know we've talked about well you know in the early part of the 20th century um uh most most countries had a Six-Day work week then we moved to a five-day work week we're sort of seem to be on the verge of a 4-day work week uh but if you're super optimistic maybe it'll be down to uh maybe it be down to three um but it still seems that at least for the foreseeable future there's still going to be a lot of stuff that people will have to do Chad one way in which work has definitely changed over the past uh decade has been the rise of gig work non-traditional work structures uh how do you see AI impacting that that's a good question I we have seen how just you know Communications Technologies change where work has taken place and sort of the flexibility of of location of workers and um and that's tied in part to the gig economy it's also a separate phenomenon um I'm not sure how things are how AI is going to either you know accelerate that further or reverse some of the changes we've seen over the past years I'm a little bit you know speaking we're all Chicago people here I'm a real Kian on this and Ronald Co had this idea basically he said the idea that has applied to this case is okay who who wins from remote work for example is it the workers get to stay at home because they don't want to commute anymore even if there's a cost to the companies maybe in productivity and I know that's up to debate but let's just Suppose there is or can the companies actually Force the workers to come in for this that or the other reason and my view the sort of kium view is look wages will adjust to make whatever the most efficient thing is happen so if it's really efficient and the productivity gains to a company or from having the workers all together in an office they're going to have to compensate the workers to come in and and commute now in a way they didn't before because that was the only way you could you could do it before but now with work at home that's an option but if if companies find that too costly they will pay the workers they'll compensate them for the commute okay if if their the productivity gains aren't that large certainly relative to the cost of commuting for workers then the workers will be able to stay at home they're not going to get compensated for it in terms of wages but they'll get compensated for it in terms of not having to commute and having more flexible schedules perhaps I sort of think whatever way the technology pushes us whether it's more or less work from home more or less gig work uh we've got this price out there that we call the wage that can move around uh to compensate the parties in a way that the parties can share in uh amongst themselves to let the most productive thing happen now that can take time and and there's plenty of frictions that slow that process down but I sort of the in the long run that's kind of how I think about your question absolutely before we move on to uh talk about firms and and economies in general then I we do have a question in something that we haven't discussed at all there has been one test of uh workers rights one large test of workers rights in the face of AI and of course that's the Screen Actors Guild strike and that led to a resolution of sorts uh does anyone have an opinion anyone on our panel have an opinion about that and what that tells us about the future path adoption of AI I'm not going to say I'm much of an expert in the particular Technologies at hand but I actually I think that's an example of the kind of Coan bargaining I was just talking about you have the producers and Studios on one side you have the actors the writers on the other there's this new technology that's going to influence how output in that industry is made uh each party is trying to figure out what the contribution of each is Visa the technology and they came to an agreement it wasn't completely friction frictionless that agreement there was a strike but they came to an agreement to share those gains uh in a particular way at least for the time uh for the next several years to come um and that's kind of the process I have I have in mind when I was talking about that Coan bargaining um we'll see whether it holds stable and whe the technology has the effects that all the parties thought but uh that's kind of kind of what I would expect more of as we see this diffusion the diffusion of the technology into other areas thanks Chad and uh thanks for the question as well let's move on to talking about firms Julia um in this new economy where different skills are going to be demanded uh yeah are firms going to have to look for a new type of worker that's a great question um I think again I I'm a data guy right so I'm I think we're going to be um seeing a different style of education and we're going to be able to observe that in the data so um we've always thought about education the K through 20 education system is providing the skills for particular uh workers but one of the things we're expecting to see in the um in the data that we're working with which is linking these records up together with education records is the increased importance of different types of certification so you're going to have to have a m we've always needed a flexible labor market now it's much easier to acquire the different types of skills so capturing ing um certifications that are going to make sense to firms is going to be a really important activity so credential engine estimates there are about a million different credentials in the US so talking about uh signaling Chad they're going to the firms are not are being overwhelmed by uh the the the certifications and they don't know what the signals are telling them so I think there'll be a lot more interest in understanding what are high quality signals what are low quality signals and what skills can workers acquire and how do firms understand how to hire the workers with the requisite skills uh Randy Julie mentioned flexible labor markets we are dealing specifically with uh Europe Middle East and Africa uh today um sometimes the labor markets in some of those countries are criticized for being too rigid too little flexibility how do you think that the impact of AI if it demands flexible labor markets will play out in these countries compared to the US experience perhaps I think that's a very important point and I think that has been one of the U the points about us labor market versus other labor markets is the flexibility to deal with change um uh and one of the key things will be uh there tends to be much higher unionization in Europe Middle East Africa relative to the US so uh the U uh the unions will have to make sure that they U they take this into account exactly as um as Chad had said there are a lot of frictions in the um uh in the strike that and leading to a strike that happened in the the US with the screenwriters and I think because of the uh the potential for large change and the uncertainty around it there may be more um friction that is associated with that so you may see some uh labor market disruptions because of that um but it's going to be very important to take this into account we can't live in the past countries that try to just keep everything exactly the same as it was and try to preserve all jobs all tasks because sometimes uh labor um uh labor contracts focus on particular tasks actually the unions in in Europe midle Africa tend to be a little less focused on that than in the US uh but to the extent that they are that's that's problematic because the tasks are definitely going to be changing and so um in thinking about the labor market consequences it is very important to think about um the labor market institutions the uh the structure of the unions the flexibility that's there there and so I think uh one of the key things that this will drive is greater flexibility in those those markets uh because I think people who try to just do things as they were in 2000 when we're in 2025 or 2020 uh 2030 um it's going to be not good for the workers not good for the economy um just trying to Prov preserve that from the past I stay with you for the moment Randy what um because we are discussing uh Africa as well here um a country that techn technologically starts at a disadvantage do you see globally this uh playing out as AI giving a growth advantage to more advanced economies how do you think countries within Africa will be able to make the best use of AI to generate growth so it's interesting um because it depends how you think about uh the starting point so for example if you looked 20 years ago you would say that well telecommunications was was much less developed in in Africa than in other countries it was very difficult for people to get a landline but in some sense precisely because of that that constraint there was much more rapid development of uh of cellular technology in Mobile phones and then mobile phone banking became much more uh widespread and is much more widespread in Africa than it is in in other countries so sometimes um what might seem to be a negative can be turned into a positive when um the when you have the the basics so you obviously will need uh the um uh the the internet connections you will need the the mobile uh mobile technology but that's there in Africa and so in certain ways uh it may be that Africa will be poised to be able to take advantage of some of these things because you don't have the same rigidities that you have in some of the other uh other countries so I guess my plea would be to make sure that um African countries don't put some of those think we've lost Randy for the second there let's just wait to see if he comes back I don't I I don't think so we'll try and unfreeze him Chad I wanted to ask you a question that one of the questions the Pres submitted questions that came in um asking about um potential Black Swan events in particular industries that will uh be associated with the impact of AI do you foresee these radical changes in particular Industries I think you know any technology that's a general purpose technology in other words has such huge scope of potential application that's basically used everywhere and I think AI is one of those it's going to have really different effects and some of those will be radical uh there's going to be certain markets where um and probably in an unforeseen way because the more you foresee something the more you pre- respond to it and therefore the less actually radical uh on the ground effect it tends to have so these these I think the the the things that surprise us most are the ones that end up being the most radical I guess that's the blacks one idea so yeah I I bet that will happen now I can't tell you which industry it's going to be but I think it will happen you know I was just thinking this morning about rewinding the clock about 30 years where it1 did actually really affect in industry and it was travel agency and the way travel agents used to make their money was over half of the revenue in the industry was on airline ticket commissions okay you you buy an airline ticket it's usually from a travel agent there'd be a $50 Commission on there in the US and commensurate amounts and in in other places and over the course of six years six years the airlines figured out with new technologies they could just sell tickets directly to Consumers and sort of and consumers trained themselves to think they could buy tickets directly from Airlines rather than go through a travel agency and so you had half of the industry's revenue disappear over six years okay um that was not how it affected most Industries most industries did not see half their revenue decline because of it but for that one particular industry there was a radical shift now travel agents are still around they do different things there's still plenty of f it actually made the travel agencies that survived even larger I've got work on that so it's not like it completely destroyed the industry or something but yeah you know any kind of general purpose technology is going to have effects all over and some of those are going to be larger than others and some of those are going to be extreme I'm I'm almost sure of that but I can't tell you where that's going to be well I I guess I take your point if you could tell us exactly where it was going to be then they wouldn't be Black Swamp I suppose but absolutely so let's move in the in the the last section of what we're talking about today to look at the impacts on economies on policy on society uh Julia you've said one of the problems at the moment is that we have an inadequate data however policy makers are under pressure to act at the moment from all sides what do they do about that well part of the thing that I think is going to be super important is to build high quality data right so um thinking about where those data source is going to come from and where you're going to have the trust in the data so a big push is not just trustworthy AI from the point of do I trust the AI but do I understand the impact and I think Randy you uh um cited David 's work we got it badly wrong with NAFTA and the impact of NAFTA uh on labor markets and you know at least there's some evidence that that's caused a lot of deaths of Despair so I think uh understanding and building a labor market information system that is local and actionable And Timely is going to be critical and we don't have that statistical infrastructure right now so part of the thing that I put in the chat was how do we build that and and how do we respond to it so part of the work that's being done to inform policy makers the other committee I was on was the advisory committee for date of evidence building how do we get that information to Governors and to local areas because there there isn't a National Labor Market there's lots of local labor markets AI is going to be as different in different parts of the country as agriculture is now so I do think we need to rethink the data system and and push towards uh faster better um data otherwise the policies are going to be really bad that's what worries me now uh what you did reference is uh your paper isn't it which is the industry of ideas measuring how artificial intelligence changes labor markets is that correct that's right yes and and we do have uh so NSF as you know is putting a lot of money into their Regional Innovation engines a lot of the new industrial policy is local Regional interventions and so part of what NSF is doing is funding this work that I the second thing I put in the chat with at the University of Michigan The Institute for research on Innovation and science which does precisely what that infrastructure um should could look like Randy I'm delighted you made it back uh I have a question for you on these Regional effects because uh perhaps it will happen that AI will benefit Industries will benefit firms in places that are already doing pretty well and will be damaging to firms in places that are doing pretty badly is there a justification for more place-based policies as a response to AI well as I I said hopefully you heard the the response on on Africa that um that some countries and Africa may be well suited to be able to take advantage of uh of some of the um the benefits that could come from AI so I think the key is that this is a new and and rapidly evolving technology it's going to be very difficult to uh to figure out how to regulate it and I think attempts to do so are probably going to end up uh leaving creating more problems for the uh the local economies or the particular countries than uh than helping so I think we really have to see how it's it's going to evolve certainly you may want to provide some support in transition for for workers uh this has come up exactly as Juliet said in the context of thinking about uh International competition and so here is a just a different type of of competition so you know having some sort of safety net to allow for uh some sort of sport and transition but I think anything that tries to like lock in the past and make new technology adapt to where we have been is not going to be successful because the new technology will be adopted and will just blow through whatever types of uh of rules you try to put into place Place best to allow it to evolve and but have a support mechanism there as um as some of the uh unintended consequences of it may affect different groups or different areas differently CH if you were a policy maker in anywhere in Europe in the Middle East in Africa what would you be thinking about with regards to reforming or improving education uh considering the impact of AI That's a great question I mean there's a couple there's a couple layers to that one is what do you do for um workers are already in the workforce to allow flexibility if their job is substantially affected by the new technology what's the structure that uh makes it um as practicable as possible for them to be able to retrain to something uh uh more amendable to the new technology how do you deal with you know we were talking about job security and things like that I I've always been you know protect the worker not the job try to keep the jobs flexible but and then help the workers who are hurt when technology changes the nature of jobs so I think education and retraining is part of that and then in terms of sort of you know education before PR pre-work Force um of course I'm a math kind of a math head I I I think everyone should learn more math now maybe people disagree with that but I I I can't see how it would would hurt to sort of buff up uh math education and actually maybe you know move a little bit the types of math education more towards uh you know statistics and data analysis a little bit away from trigonometry even as a engineer myself you know I think trig came from a history of engineering but I think what people are more apt to use these days are stats and data analysis um but anyway that that maybe a little into the weeds I think the the short answer is we'll have to think about education at all ages and just build building in flexibility we it's hard hey it's hard to predict the future we all know that so you really want to have a system that offers options and can react to the different realizations that might happen as as we go forward Dad I think you're as likely to find a math positive audience here as anywhere so I can you're okay on that one Julia just a quick follow-up question on that one have can we learn from the past uh when we've made these adjustments badly and the data has shown that we made these adjustments badly well let me flip the uh question and say what have we done well right so um the I I been very interested in the success of American Agriculture and investments in American Agricultural R&D which uh even as a New Zealander I'm very impressed what by what the Americans did a and so why was that so successful so part of the reason was was that farmers you know when they came to this country had many different problems so Wisconsin apples and Texas Longhorn or whatever and what they did was they built Farmers institutes be to learn from each other and they that evolved into the egg extension program in the United States so they had a set of landr Institutions where you had researchers working on problems that the farmers came up with and then they transferred that information via the EG extension system and that was codified by in 1862 by the moral act with the land Grunts and then with hbcus I think in the 1890s I think that kind of system where you provide local information about how to respond and how to react for local needs in local labor markets is going to uh enable or avoid the mistakes of the past it worked spectacularly well with a very robust um uh a extension information system we could do the same for local labor market policy local extension services and local industrial policy and if if I could just really quickly add to that and tie it to what Randy said about we don't want to freeze today's economy yeah going forward you know in 18 62 when the moral Act was passed that established the land grant system probably about 60% of the workforce were farmers now does anyone think it was the same policy to try to keep 60% of us farming for the next hund 60 years of course not and but increases in agricultural productivity meant we didn't need hardly as many farmers as we did then but that was not a bad thing that we all moved to do something else that we allowed the economy to shift its shape as productivity growth happened in part because of the system that Julia was mentioned and then just to jump in but you need data you said Education and Training Department of Labor uh partnering like the employment and training Administration partnering uh with the agencies that are spending the money on the resources and building the data infrastructure so that the training providers which very often are community colleges or these certificate ins institutes that is the way to be build flexibility into the economy now before we finish I can't let a Chicago webinar about uh Europe pass without talking about the EU AI act and whether regulation how big a role regulation of this style is going to have in the impact of AI on the economy Randy do you have an opinion on that so as I was saying before and hopefully you were able to to hear that that um it's you know this is something that's rapidly evolving it's something that it's also somewhat difficult to to to precisely Define and and so I think it is important to uh to to think about whether there are potential guard rails think about some of the the downsides but I think we have to be very very careful in sort of rushing to to regulate U this has been one of the challenges that um that I think the EU EU has has faced B in general in Innovation um so you know the big Tech firms tend not to be in Europe there are some uh there are some that are there but uh but most of them are in the us where we haven't had as much sort of rush to regulate in some of these areas and so um I'm not sure it's in Europe's best interest to do that I think along the lines of what Chad and I were describing as well as Julia make sure that there's a safety net that's that's there for um transition because we don't know exactly what the consequences are going to be but you want to make sure that um workers and people in the economy overall and investors uh can take advantage of these uh uh these great new things that are happening and so um make sure that there's a safety net that's there but I would be wary of jumping very rapidly into regulating an area where we really don't quite understand it yet Chad do you agree very much so I I mean I study this stuff and I have very diffuse Notions of how AI is going to be used in various kinds of settings and the effects it's going to have and to try to I think anticipate all the potential Nega negative effects of that and codify it now is really too early uh we need to be paying attention to what effects it has but um I think it's you know any anything reasonably strict I think it's is jumping the gun Judah I'm gonna give you the last word um that's very wise you started off so well so um anything anything that we uh anything that we're saying about AI at the moment is of course speculative so even though you are the data guy as as you say Do you think that we're going to be in a couple of years coming back here and talking about the negative impacts on um inequality on the uh on the welfare of the most vulnerable people in society because of AI That's the fear that dominates the public discourse at the moment well we just heard from two Chicago people now you're going to hear from a NYU person so you're covering the Spectrum very nicely there of course you've got to worry about the the people who are going to be most vulnerable um I think uh gadal said we value an economy by how we treat the least Among Us and I I think that is a true statement you have to be sensible unlike uh many of the previous policy programs which have have exacerbated problems not solve them and I do think that what Chad and Randy have said which is figure out who's getting negatively affected and I think you use data and evidence to build policies that don't have the unintended consequences that we have had in the past which have broken up families and had all kinds of uh bad pieces data and evidence I still believe will make a difference so I hope that we're suful and don't make the mistakes of the past it's a good place to finish and we are out of time so uh thank you very much to all of our panelists thank you Julia and Chad and Randy thank you to all of you who uh asked questions we did manage to squeeze a decent number of them in and so I hope you think that we represented the things that you were interested in I think it was a high quality debate and it's not going to be the last one we have about this so uh panelists thank you very much from me Tim Phillips we will see you soon but goodbye for now
2024-03-07