Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Is Software Eating the World? Measuring the Progress and Diffusion of AI

Authors: Filippo Bontadini, Carol A. Corrado, Jonathan Haskel, Cecilia Jona-Lasinio  
Discussant: Martin Beraja  
Video: https://www.youtube.com/watch?v=VdvT0JzwMHU&t=20490s


## Talk (05:41:30 – 06:02:43)

[05:41:39] us whether software is eating the world. >> Thank you. Well, Mark Andre famously said software is eating the world and this paper we ask is that measurable and  
[05:41:53] is the national accounts framework set up to see it? The answer is that the signals and basic framework are there but the measurement is not and that that matters for policy.  
[05:42:06] The vis Oh, now it sort of ran ahead. Is there  
[05:42:18] like a lag? There we go. The signals are visible. Um, software capital expains roughly half of US labor productivity growth since 2012.  
[05:42:32] and the software producing sector that is just 6% of the non-farm economy contributed 43% to a notable acceleration in TFP after 2017.  
[05:42:46] So something real is happening. The three measurement problems systematically hide it. First, the cloud asset boundary hides AI capital income.  
[05:43:00] When a firm in a traditional industry buys compute from AWS to run its AI applications or even to use it in in current production, that's booked as an intermediate consumption rather than an  
[05:43:14] input of AI capital services. Second, there is no quality adjusted price index for AI.  
[05:43:25] Um, I mean there's just there and the official statistics actually have no counterpart uh to what I'm going to be showing you um about the pace of uh  
[05:43:39] price declines in this sector. And third, concentration. The aggregate factor share for AI related assets in the national accounts mostly reflects what's happening in the a handful of of  
[05:43:54] frontier firms. That 6% that I was I was telling you about and it does not necessarily reflect broad diffusion across traditional industries in the economy. All three  
[05:44:07] problems point in the same direction. they understate the true capital deepening especially in traditional industries.  
[05:44:18] So our framework has two sectors. Um what's let me start sec se first there's the upstream sector uh which you can see  
[05:44:30] on your right um uh produces AI capital software train models data assets uh and it conducts R&D crucially it uses AI to  
[05:44:42] produce AI uh creating a recursive loop uh sort of like the turn of a flywheel where each turn. With each turn, the cost is lower. This leads to superior  
[05:44:57] TFP in the upstream sector, which drives the price of capital down. The downstream sector C uses AI capital along with uh labor and other factors  
[05:45:09] and the user cost is what the downstream firms pay for a unit of AI capital for a unit of AI capital services. So as the price of of capital falls uh the  
[05:45:23] user cost falls driving substitution towards AI capital. That's sort of just basic economics. A key parameter governing how much substitution occurs is the elasticity of substitution sigma  
[05:45:37] between AI capital and other factors of production mainly labor. But firms cannot simply swap in AI. organizational capital Z  
[05:45:48] must co co-evolve with AI adoption. So this gives us two conceptually distinct um substitution channels both of which show up in the aggregate factor shares.  
[05:46:01] First there's the downstream diffusion where AI replaces or augments labor and then second is this upstream recursion channel where AI is used to produce more AI and they have very different drivers  
[05:46:16] very different determinants um and so very different sort of policy implications. So to fix ideas um this diagram just shows the mechanism  
[05:46:28] mechanisms just described and as seen in these mechanisms the user of national accounts can focus on two observable diagnostics to follow the diffusion and  
[05:46:40] progress of AI. One is the downstream AI capital factor share and the other is the rate of decline in AI capital costs.  
[05:46:51] uh the latter emanates from technical progress and the former is about about diffusion and the problem is that both of these are mismeasured. So what can we do? Well, before we go any further,  
[05:47:06] there is a I guess I would say a a theoretical caution, certainly a conceptual caution of what we can think of in this framework. We can think of Z which is like organizational capital as  
[05:47:21] playing the same role as what Jones and Tenetti call the weak links constraint uh in task models of production. So this  
[05:47:34] is where even if AI displaces labor completely within a task in every task it enters anyway um the output gains depend upon how easily one task substitutes for another  
[05:47:48] to see how it relates to RZ. Um a point we make in the paper uh is that uh we take a cue from um Mgrim and Roberts 1990 and suggest that  
[05:48:03] the coordination of production itself is a task sort of a meta task if you will uh requiring large investments in new business models that's the Z in our  
[05:48:15] framework and the best example of this is the Amazon logistics revolution.  
[05:48:21] ution uh which is a good example of how new coordination methods ultimately but certainly not initially uh become a driver of profits and um and then eventually downstream TFP not and it's  
[05:48:35] not sort of substitution that that goes on it's just the pure expansion of the po of the production possibility set so a high sigma and a falling user cost is are necessary but not sufficient  
[05:48:50] conditions for diffusion uh and broad-based productivity gains. With that in mind, let's go to the three measurement problems and um and how they  
[05:49:03] are all biasing our diagnostics downward. Um to do that to go through these these issues um we  
[05:49:15] use the tool of the AI production stack. In other words, to understand the measurement problems, we need to understand how the AI supply chain. Um, so this is  
[05:49:28] comp this gets complicated. I don't want to read everything, but it organizes the supply chain into five layers.  
[05:49:36] Applications, you can see familiar applications at the top. Um, followed by the foundation models right underneath them. Uh these two are each about 20 billion dollars today. Well today I mean  
[05:49:50] 2025 when initially put this stuff together. Um platform platforms and toolings is where you know enterprise data sets might sit uh or the vector database that you might  
[05:50:05] have for your your research lab or research project. Um and then we have the hyperscalers but here cru c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c crucially we're just talking about the AI increment  
[05:50:17] um and this total market is 72 we estimate is 72 billion today globally uh compute sits right below it and it's now  
[05:50:29] the largest layer in the stack uh 180 billion two things to take from this table first. Um you can see there's a  
[05:50:42] projection for 2020 2035 uh in this table as well based on industry sources. Um so first the sheer scale of what is coming underscores the urgency of getting the measurement  
[05:50:56] right. Second, at the L4 level, the revenue from the cloud revenue is precisely where this asset boundary problem is  
[05:51:08] most acute. Um, and then so I'm going to talk about that in a minute. And then at L2, these AP API prices L1 and L2 uh are our  
[05:51:23] observable price series that that I'm going to be developing later. So that's that's how it all fits in. Um the paper formalizes the cloud this cloud asset B  
[05:51:35] boundary problem and discusses the issue in some detail um and how the SNA hasn't really solved it. Basically they're they're going after a onesizefitsall  
[05:51:49] solution. And I think I've just convinced you that it's it's not the upstream producers where the problem is.  
[05:51:56] it's the downstream producers. So basically cloud just creates a mirror image error. Too much capital income in the upstream sector, too little in the downstream sector. Um before going any  
[05:52:10] further further, I can show you just an sort of just an example of of what this means. Panel one uh shows observable digital capital income shares for the US  
[05:52:24] uh non-farm business sector that's built by asset in this case ICT hardware software products and software and AI R&D  
[05:52:37] the overall ratio rises gently until uh 2017 and then accelerates rather sharply Um then panel two it's the same top line  
[05:52:51] only the segmentation is by the up by software producers and traditional and then all other traditional industries.  
[05:53:00] And here you can see that what drove that line up was activity in the software producing sector. Um and certainly the  
[05:53:14] the diff the downstream diffusion looks relatively low uh by this standard. And then panel three I've added a very crude  
[05:53:27] uh sort of shadow cloud adjustment that we made for this paper. It's deliberately conservative and you can see how it begins to close the gap. So the bottom line here is that the  
[05:53:40] concentration is genuine. Um a lot of activity is happening upstream, a lot less downstream. But if we close the if we fix this cloud  
[05:53:52] measure cloud asset boundary problem, we we begin to see what's going on and it's a little more of a hopeful um a little more of a hopeful picture.  
[05:54:04] And although I don't want to cut into my time, I mean, this is where I would circle back to the um the paper's title and say that that the the actual  
[05:54:16] prophecy that Mark Andre made in 2011 was that softwaredriven companies, I think he said, like Amazon, Netflix, and Google, uh would disrupt and dominate  
[05:54:30] traditional industries. So that hasn't happened yet. Uh but that line is you it's hard to judge the level of the line but um that's sort of where we where we  
[05:54:41] are uh right now. Um so what about prices?  
[05:54:47] Um so here I think we need to have a break from the past. Uh the IT era priced hardware and software separately and it was captured  
[05:55:02] in the era's off-ioned adage, what Intel giveth, Microsoft taketh away. And I can still hear Bob Gordon saying that at one  
[05:55:14] CRW conference or another. Um and that separation was genuine. Uh a chip index could be built independently of the software running on it. But AI breaks  
[05:55:27] this se this separability. Capability gains emerge from the interaction of all the layers of the AI stack that I showed you. Architecture, data, training scale, no single layer can be priced  
[05:55:41] independently. And improvements also are often discontinuous.  
[05:55:48] architectural breakthroughs rather than smooth progress along sort of a fixed attribute space which means that hydonic methods don't help us very much because they need stable measurable characteristics.  
[05:56:01] So the right unit we say is the deployed model uh specifically the price per unit of AI productive output regardless of  
[05:56:13] which layer in the stack produced the gain. So that's the concept that we're going to advocate or that we advocate in this paper. So in contrast with the IT era, I guess we could say I don't think  
[05:56:27] I have it on the slide uh that uh we rather have what the frontier model delivers if you want to keep the verbs  
[05:56:37] the same. So how do we do this? Um, oh right, I to save time I skipped a slide that uh gave you a lot of information on how we delivered a price  
[05:56:52] index, but here's just the results. Um, the paper details the construction. Uh, we build uh two price indexes. One is uh a token price index. You've heard a lot  
[05:57:05] about tokens. That's our unit of productive output for generative AI. Um, and that falls at about 25 and a half 24 and a.5% a year. And second is a  
[05:57:19] capability adjusted index that adjusts token prices by an industry performance metric, the EPIC capabilities index. a metric that incorporates  
[05:57:32] a suite of industry met metrics like reasoning, coding and knowledge are the main main benchmarks uh that are in that and then that price index falls at 38  
[05:57:44] and a half% a year. So um this is uh quite there's no software price that there again there's nothing  
[05:57:58] that compares with this in the national accounts because we're in essence lumping chips and and uh and software and model and R&D and all kinds of other  
[05:58:11] things together um and saying what's coming out what's coming out the And um I think it's noteworthy that there's a 14 percentage point, you know, difference between what is essentially a unit cost price index and a quality  
[05:58:26] adjusted in uh price index. But uh we could talk about that another day. So um I can see I'm running short on time. Um so the proof of concept in con in  
[05:58:41] context that means I only regard this price index as a proof of concept concept because it has very few observations really this is the whole software market  
[05:58:53] um at roughly 72 billion today AI is 17% of it and again what this tells us is that uh if we look forward it's going to  
[05:59:06] be 79% of the market by 2035. Just industry guesses. Um, and if we don't treat this these measurement challenges  
[05:59:16] as urgent, uh, we're going to, uh, have a problem with our with our fundamental statistics. The window to act is actually very narrow. Um and actually  
[05:59:29] that's what the rest of the paper um um addresses because we connect the price index to factor shares and to do  
[05:59:42] that we estimate a translog cost function on the downstream uh on downstream industries using a cross-country uh data set. So we have  
[05:59:54] time, country and industries. Um and we report Morishima elasticities here which is the conceptually appropriate measure for a multi-input case. And note that  
[06:00:05] every every entry is below one and that means that well I think we all know that what that means that uh firms are going to substitute toward cheaper inputs but  
[06:00:17] not buy enough to raise the input share. Um and I think this is primmaaccia evidence for um the Jones and Tenetti weak links constraint that I mentioned  
[06:00:30] earlier. [snorts] Uh we do some simulations um that are very labor friendly. Uh not surprisingly um I I could you know it's it takes a  
[06:00:43] bit to sort of interpret them but let me just say they're labor friendly. Um and um but there's limits to to what those simulations and the estimation  
[06:00:57] shows. Um and um I think that crucially we need to say that these estimates embed the existing sort of meta technology of the  
[06:01:10] coordinating tasks and we're holding that constant and unless AI sort of with the passage of time and better and updated estimates if AI shifts the  
[06:01:22] morishima elasticities above unity uh then the labor friendly results will will go away but they haven't yet. Um so  
[06:01:34] the bottom line is that uh you can't um is is that it's hard to capture the structural change in any sort of production possibility set or talk about  
[06:01:48] it. Um, and that's precisely why the measurement agenda is not just a technical exercise but a prerequisite for understanding whether AI's arrival ultimately complements or displaces uh  
[06:02:02] labor at the aggregate level. And um I think it's clear that uh the papers covered um a lot of ground and the future agenda seems pretty clear and  
[06:02:17] financial signals from announcements could give us leading indic indicators to work with and measuring all these dimensions correctly is a prerequisite for the policy analysis that AI's  
[06:02:31] arrival makes urgent. So thank you. [applause] >> Okay, our next paper is the micro  

## Discussant remarks (06:47:54 – 07:02:47)

*Shared across the three papers in this session; Martin Beraja discussed all three.*

[06:48:03] >> Hi everyone. Yes. Uh thanks Eric and Karen for inviting me to discuss these uh three papers on AI productivity and economic value. And uh so know discussing three papers is kind of hard in only 15  
[06:48:18] minutes. I'm not going to go deep. I saw my role as a discussant as know framing a bit what these how these papers fit together and uh and in fact and also sort of giving you a uh know hitting the call of Matt Jackson's this this morning  
[06:48:32] on you know there's one shallow view of this p of this session and also I think it applies to other sessions which is oh these are just three papers that measure AI use and value uh at different layers national accounts firms and households  
[06:48:47] and in a bunch of other sessions we saw the same idea A that no people are just trying to paint pictures of what is going on with AI use adoption uh value and I think the call of MAT which I wholeart wholeheartedly agree on is that  
[06:49:01] we need a bit more theory to kind of see what like what are we learning from all these different pieces of evidence and the theory that I'm going to put on the table right now is that I think these papers are pointing to something deeper.  
[06:49:14] It's not just different layers. They're pointing uh to thinking about AI not just as an as a technology that uh will automate production tasks like a fancy robot but more so that is a technology  
[06:49:28] that can accelerate learning across different agents. So firms, workers and households when they learn and codify and act on on information. Okay. And and so the the the question the organizing question for the discussion is not going  
[06:49:42] to be is AI being adopted which can be you know read as a bit shallow but does AI change how agents learn and use uh information and uh and I would say the the different papers measure different  
[06:49:55] parts of this learning process as it happens through the economy starting with households searching uh ideas and jobs and learning skills and then turning into firms and workers who then qualify some sort of knowledge into organizational capital that then  
[06:50:09] aggregates into something that we can measure in the in the in the national accounts and so know starting with the last paper that here comes first is the household's paper uh where there you know their language is that no judgment is mostly used for productive online  
[06:50:23] tasks and and I like to translate that into being about pre-market entry learning in the in the following way. So one which they talk about quite a bit in the paper and another one that I which they talk more about. I think you have  
[06:50:37] the the data there to to do so. So the the first part is is is learning uh by workers or potential workers both through education and job search different informationational activities  
[06:50:49] that end up bu being built into what we will call just human capital matches across firms and workers and and know in the in the paper they they go the they they go a long way uh trying to to look at those type of activities. There's a  
[06:51:02] separate activity that also happens at home which is someone just ideating a new project, a new firm, a startup and uh and in that process no they're doing a lot of informationational research and discarding ideas that may be good  
[06:51:17] discarding bad ideas and and focusing on on on better ideas and that all fits into entrepreneurial pre-entry signals that eventually become firms. So I would really like I I'll show you later. I think this second part is actually a big  
[06:51:31] deal. it could be a big deal quantitatively in in how it fits into aggregate GDP. Uh the paper then that goes and now know we're moving within firms and uh and here I really liked in  
[06:51:45] this paper how they separated the the functions and not just saying oh these are the tasks that are performed within the firm versus not based on own it but functions that really uh align in my mind better with what is organizational  
[06:51:58] learning. So the bottom up channel that that that was discussed this idea that there's workers that are experimenting with both the tool and the task and are learning in the process of performing those tasks how to uh performing better  
[06:52:10] and which eventually leads to some codification coming from the top of those uh activities as being firm policies or broader functional uh deployment. So  
[06:52:23] again this this this this chain from individual learning to organizational capital is kind of what is what is being uh shown in in in the in the in the paper on firms and you know an initial  
[06:52:36] reaction in that paper is well if you see some of the numbers actually they look small in terms of adoption uh so far so narrow use but I'm I think that the way I read this is just early evidence for organizational learning the firms are still figuring it out exactly  
[06:52:51] how is it that we're going to uh this uh these new AI tools and uh and again I would commend the authors to really like push this idea of how is it that we move from one layer to the to to the other layer. I thought that was the most interesting part of of the paper,  
[06:53:06] not just oh here's another survey where we're measuring AI use within firms. Um lastly the the paper on on intangibles or AI organizational capital in the argate. I love this phrase. I didn't  
[06:53:20] know it from Mor it was from Mgrim and Roberts that that one way to think of organizational capital is as the accumulated stock of solutions to the meta task coordination problem. I think that's a fantastic phrase. It's a fantastic way of thinking about this and  
[06:53:33] that what AI can do here is really reduce the unique costs of coordinating activities uh within the company that then leads to organizational capital and things that we measure eventually as higher uh productivity. And so this is  
[06:53:47] really the bridge from task exposure measures to organizational learning. How how we how we gonna uh coordinate those uh those tasks. Now for those of you who have never thought about intangibles but  
[06:54:00] know a bit about physical capital and the perpetual inventory method, let me spend like two sec two minutes on on on why is it what Carl was doing hard and and and why I give her a lot of credit for trying to do this. So if you think  
[06:54:13] about physical capital, how is it that we measure it? don't go out and count machines. That's not what we do. What we do is we count the flows. So we look at say machine purchases the flows we some sort of deflator for for those prices  
[06:54:28] and then we come up with a measure of the quantity of machines that were uh added to the stock and then we accumulate those and hope that over time know the old machines depreciated enough that they don't matter anymore and we get a measure of the stock. That is the  
[06:54:41] perpetual inventory method. This works well because what you're measuring on the investment side machines is exactly that what goes into the stock. Okay. The problem with organizational capital  
[06:54:55] intangibles now the new version being AI based organizational capital is that what you're seeing say use of AI or expenditures in AI in cloud compute whatever it is like give me your your  
[06:55:08] preferred measure is not really what goes into the stock. It's an input into what goes into stock. So you're not measuring the the investment flow. Well, in particular, you can think for example that the way to build the stock of organizational capital is not just with  
[06:55:23] compute or tokens or AI use expenditures. You need to combine that somehow with say managerial and worker time. Okay? And that by combining those two things they figured out how to say produce better, coordinate those meta  
[06:55:36] tasks. That's the knowledge that is being produced. And what you would like to measure is this knowledge flow not the use of AI and that is hard because we don't observe that other part we don't observe the managerial time we don't observe the the worker time or any  
[06:55:50] other resource that goes into the the investment of uh uh that leads to organizational capital accumulation and and so what are the risks uh or when is it that this works well so suppose that you know the imagine that the investment technology is just CRS. So here you can  
[06:56:04] see very clearly that that if you have and no they did a great job at trying to measure the the decline in in the price of of of uh of AI. If you have AI that is getting cheaper then what's going to  
[06:56:18] happen within the companies that no we're going to use a bit more AI and we're going to substitute away from other things that we were using before to build organizational capital. So if we just count and accumulate all the expenditures of AI, we're gonna be  
[06:56:33] overstating the actual increase in investment because we're not seeing all the substitution that is happening say with manager time, worker time, h whatever it is. Uh so no, this only works it works perfectly when the two are perfect compliments. So if the  
[06:56:47] managerial time and AI are perfect complements, let's say to build organizational capital, then you're getting the right measure. But otherwise, you're you're overstating it.  
[06:56:55] uh so one idea here is to use you know how do you fix this so you could use substitution elasticities and I don't know if you try this for other form other forms of IT related organizational investment that we have seen in the past  
[06:57:09] and no the extrapolation would be well maybe the institution are similar between computers and managial time than they were than they are now for AI and and managial time and then we then we move we move forward uh now the second  
[06:57:23] one I this one really I don't know what to do. I think it's a it's a it's not just a problem here. It's just a more general problem that when you're measuring AI use or when you're measuring AI investments uh and  
[06:57:36] expenditures, what you're seeing is not really the part that goes into that is really investment. You're seeing both the production and the investment side.  
[06:57:47] So imagine that you know you're spending X number of tokens and that's what we see. That's what you see for each firm.  
[06:57:53] We don't know where those tokens were used for writing emails which would be like a production task or they were used for you know ideating new products or codifying information anything that looks more like a like an investment. So  
[06:58:06] again, I think this is a call for uh perfect this is a call for you know kind of trying to separate better what is investment what what part of what we're seeing as being AI use and AI expenditure is really an investment  
[06:58:20] versus something that goes into uh production and and it's maybe automating or augmenting uh those tasks. So here I think the two papers the firm and the intangibles paper can kind of like talk to each other a bit and see whether from  
[06:58:34] the surveys we can get some idea of how much of the use is really something that looks more like investment versus something that looks more like production and with that we can do maybe some some adjustments. Um uh  
[06:58:49] good so let me just close you know again going back to to theory uh a bit on what are the no I talked about learning but what are really the two mechanism of learning that can be important here. So one that the firm paper and the national  
[06:59:02] accounts paper point to is really uh learning through firm organizational capital accumulation and uh and and that may that that that may an important part of the story might only be half the story. A more important part or as  
[06:59:17] important part that I think you know receives much less attention is whether AI can help you know potential entrepreneurs or even companies select better which projects are more likely to  
[06:59:30] be successful and households the household paper is going some way into trying to measure that. I would push you to do way more on this because I think there's we we need much more work on figuring out uh this part so that AI may not only matter by speeding up our  
[06:59:45] capital accumulation but also by improving uh selection of projects in the in the economy and and so let me just you know give you you might say well why do we care about all these things is this a big deal or not and another way of asking that question is  
[06:59:59] the question that we asked in this paper with with Edward Talamas on if we had a technology that could accelerate learning like AI high then what would the aggregate uh GDP gains uh be are those potentially large or not if because if they're not large then we  
[07:00:13] could just know close shop and just think about AI as a robot h but if they are large then I think it requires much more careful thinking and measurement about uh those two those two channels so in this paper we know we came up with with this metric vault which is the value of organizational learning  
[07:00:27] technologies that and we show that in a very large class of firm dynamics models where there is learning happening you can come up with uh the gains just by looking at two sufficient statistics that I try to capture how much learning  
[07:00:41] is going on in the economy. So one is if you see an economy where mature firms are much larger than than than young firms that is an economy where organizational capital accumulation is really slow. It takes time and and it's  
[07:00:53] important and and and another the other stat statistic is if you're seeing an economy where young firms exit much more frequently than all than older firms that is an economy where selection at entry before entry is not very good. So,  
[07:01:08] we're not doing a pretty a very good job at the economy is not doing a very good job at selecting uh uh the the right projects uh pre-entry and and when you compute those statistics in in the in the in census data, you come up with numbers that are you know older firms  
[07:01:23] are three times as large as younger firms and they die uh no seven times less than than uh than uh than younger firms. And so that gives you a a gain in terms of aggregate GD potential gain of  
[07:01:36] accelerating learning if you go to the very top firms. So this accelerating a lot 40 years. So we think about this as as an upper bound of two meaning that no accelerated learning could double GDP in the US. And um now what we don't know in  
[07:01:50] this paper with the dart is well how much does AI accelerate learning? We can look at different scenarios. Imagine you can accelerate learning by 40 years, 30, 10, two and then the gains are very different whether you have you know  
[07:02:04] whether a accelerates learning by many many years or not and also depending on the mechanism that you have in mind whether it works through mainly through organizational capital or through the better selection and and prediction of projects. So then the call for all these  
[07:02:18] you know the the papers that are working on measurement is try to frame the discussion more in terms of these mechanisms and and and and the efficiency gains in each of these mechanism in terms of for examples years accelerated uh of learning because I  
[07:02:32] think then that's going to create much more value for the rest of us that are trying to quantify uh sag gains or or macro implications of this evidence to try to map them better to to to the models there. Okay. Thank you very much.  

## General Q&A (07:02:47 – 07:10:06)

*Shared across the three papers in this session.*

[07:02:47] >> [applause] >> We've got our microphone up there if people want to line up if you have questions or comments.  
[07:03:00] >> All right. Yeah, please go ahead. >> Yeah, thank you. Um, so thank you um all for coming and presenting uh to us. I think these were amazing papers. Um, I guess I have a question maybe for the first two especially about AI diffusion.  
[07:03:15] Um so there was a lot of focus on the AI diffusion maybe within workers workers adopting AI and within organizations where um either from the top down or bottom up it's used more but if we think  
[07:03:28] about AI diffusion across like an entire industry or across the economy there's also maybe similar to the idea of a selection effect uh mentioned by the discussant um you could have new firms who pop up who are early adopters of AI  
[07:03:42] technology. I mean, I'm I'm just struck by we're um in Silicon Valley right now, the home of startups, and there are so many uh AI native firms hoping to disrupt, you know, existing incumbents and existing uh industries um with uh  
[07:03:55] some sort of solution that's fully AI uh based. Um so I was wondering um if there are any thoughts on like metrics uh that could be used to track this to see if it's important or not. And I guess if  
[07:04:07] you have any personal opinion about um this uh avenue being important for AI diffusion within the economy. Thank you.  
[07:04:18] >> Do any of the Yeah. Uh could you get a micro There's a microphone right there.  
[07:04:27] >> Is it like this maybe? Uh okay, it works. So yeah, that's a great question.  
[07:04:33] One thing we haven't so far accomplished in this paper is that we didn't match the BTOS data to LBD which is the ne next task and it's almost done and then we're going to be able to bring in the  
[07:04:45] age variable for firms right so then we can really look at the diffusion patterns for young firms versus old firms and you know basically uh you know chart out the discrepancies there and  
[07:04:58] see if there's a um extra propensity for younger firms to come in with already AI technologies for example and I do have a related paper that looks at um business formation micro data from the business  
[07:05:11] formation statistics of census bureau where uh using text analysis I'm able to sort of like fuzzily identify uh which firms are AI related uh which uh business applications are AI related and  
[07:05:26] aiming to start AI related businesses. So we have seen an acceleration uh in AI rel related business applications and formations in the last uh eight years since the you know deep learning revolution trying to commercialize these  
[07:05:41] technologies. So definitely there is some evidence of that you know the diffusion through entry and new firm formation is definitely there and maybe we could do more to quantify that. I  
[07:05:53] agree. Yeah. >> Carol, do you have anything to No, I don't have anything to add. I I to that I think uh the the linking to the age  
[07:06:06] variable uh is you know answers the question the gentle the the gentleman had. Um I I think that one does have to bear in mind if somebody is using AI  
[07:06:17] technology to disrupt a traditional business as opposed to an AI firm that's operating in that AI stack. I mean they're just serving very different roles in the economy. Both both are  
[07:06:32] good. Um but um it's the diffusion is the the first example that I used.  
[07:06:41] >> Terrific. Eric, are you go? >> Yeah, it's a question about the com score and maybe I missed it, but the adoption was, you know, if they'd ever they visited chat the website and then there were adopters there ever after,  
[07:06:55] but it seems like the the power of the comore data if you have this second by second. I think most people use or a lot of people anyways use chatpt through the website and wouldn't that give you a very detailed information about  
[07:07:09] extensive use and people who use it for a while and then stop use it. I've seen some of the adoption surveys see falling usage by by some people or or just level of intensity and I don't know if you could do more with that or if I missed a  
[07:07:22] reason why you weren't >> Yeah. No, I mean I think it's a very welltaken point. Um we have the intensive margin based measure of just you know in a given quarter how many seconds or what share of your browsing  
[07:07:35] seconds did you spend on chatbt or openai.com?  
[07:07:40] uh but it's not something that we use to kind of uh extend any of the the insights of of our analyses. It was more just kind of as a you know robustness check for the indogenous variable. But I agree with your point.  
[07:07:52] >> You can say I mean honestly that would have been the first thing I would have looked at. But but what what what is is there anything you can say about what what that showed? I mean >> yeah I mean we find similar treatment effects essentially when we scale it by uh that alternative endogenous variable.  
[07:08:04] But I think kind of to your you know more to what you're asking I think that you know there's a lot more that we can push on in terms of things like is there a generative AI divide when it comes to if you start using it do you keep using it like is the divide we see the  
[07:08:19] >> John Hartley found some evidence in their their survey that that usage was going down I don't know whether it was like a blip or noise or whether it was real did you see anything like that >> um I honestly do not think that we have  
[07:08:33] No. So we see that the intensive margin is >> usually end before when when does your data end? 2020.  
[07:08:39] >> So it's about to be extended up until the end of 2025, but currently it ends in December 24.  
[07:08:46] >> Well, no, we haven't seen any evidence of a dip so far, but I think within person that could, you know, it could be masking within person heterogeneity. So yeah, >> Eric, that's great. Just to add to that um so we are also exploiting the  
[07:09:00] substitution between computer versus mobile because uh when we talk to the open AI I think you know there's an important fact folks are here um so uh there is a kind of a massive rise of  
[07:09:13] mobile use at the end of our sample I think it's a June of 2025 I believe so that's that's another thing I think you're mentioning we're we're really excited to look into. Yeah.  
[07:09:27] >> Okay. Since we have about a minute left, I just want to ask if any of the paper authors have anything to say in reaction to Martin's excellent discussion.  
[07:09:45] [laughter] >> All right.  
[07:09:49] >> I just want to thank you. It was really nice and you gave us some new ideas to exploit entrepreneurship. Um, yeah, that's cool.  
[07:09:57] >> Okay, we have a 15minute break, a little more networking time, and then uh please be back here promptly at uh 4 o'clock and we have an amazing panel. See you then.