WARNING: Summaries are generated by a large language model and may be inaccurate. We suggest that you use the synopsis, short and long summaries only as a loose guide to the topics discussed. The model may attribute to the speaker or other participants views that they do not in fact hold. It may also attribute to the speaker views expressed by other participants, or vice versa. The raw transcript (see the bottom of the page) is likely to be a more accurate representation of the seminar content, except for errors at the level of individual words during transcription.


Rapid technological change has led to questions about the usefulness of GDP as a metric for measuring economic activity. Alternatives have been proposed, such as measuring throughput and value, while AI investments have highlighted a misallocation of capital in this area. These developments raise questions about what GDP is currently measuring and whether it is accurately reflecting the value of different technologies. Ultimately, these changes may lead to a deflationary effect and a sense of unfairness among citizens.

Short Summary

The team discussed the potential disruption to human labor caused by rapid technological change and its implications on existing metrics such as GDP. It was questioned whether GDP would remain useful for the next 50 years and what a competing metric might look like. It was proposed that people may cling to what can be measured, and that current systems may remain in place for some time. GDP is a measure of economic activity, and is only able to illuminate what is visible to it. It is correlated to production and consumption, as what is consumed is usually what is produced. Income and the informal economy are factored into GDP as an indicator, and it was questioned how capital investment would show up in GDP figures.
GDP measures the total value of production, consumption, and income. However, advances in technology, such as DeepMind's Alpha Tensor, could mean that GDP becomes less useful as cost avoidance and opportunity costs become harder to measure. This can lead to deflationary effects, as customers may be unwilling to pay more for the same product due to technological improvements. This highlights the need to consider incremental improvements, as even small increases in efficiency can have a large impact. GDP is becoming increasingly less useful due to deflationary factors such as technology and energy, as it is a measure of how humanity is rearranging the world's matter, energy, and information from less useful configurations into more useful ones.
GDP is a measure of production, consumption and income, and is often used as a proxy for material prosperity. It has its flaws, as it does not capture all forms of production and consumption, and can be gamed by local governments. Keynes argued for government stimulation, but this could lead to busy work. Alternatives such as measuring throughput and value, or micro-transactions, have been proposed, but it is still a reliable way to measure economic growth and compare different periods.
GDP is seen as an indicator of a society's prosperity, though it may underestimate true prosperity due to the rise of technology. This can lead to a sense of unfairness, as people may feel that there is a lot of prosperity, yet their own lives are not improving. AI investments are a sign of optimism, yet there appears to be a misallocation of capital in this area, raising questions about what GDP is measuring. Tesla's demonstration of their bot project highlighted the need for more resources in developing AI, and how GDP may not be accurately reflecting the value of different technologies.

Long Summary

The team discussed the hypothesis that rapid technological change will increase productivity rates and render much of the productivity generated by society less visible to existing metrics such as GDP. They asked what the early signs of this transition might look like and discussed the possibility of using a quasi-experimental approach to answer this question. They also discussed the implications of this hypothesis and the possibility of it happening sooner rather than later.
The transcript discusses the potential for a disruption of human labor due to advances in energy, transportation, agriculture, and intelligence technology. It is suggested that this disruption could be more profound than the first and second industrial revolutions, and that GDP may become less useful as a result. It is asked whether GDP will remain useful for the next 50 years, and whether a competing metric might emerge. It is suggested that people may cling to what can be measured, and that the current system may remain in place for some time.
GDP is a metric used to measure economic activity, and it is only able to illuminate what is visible to it. The digital revolution has caused new forms of value creation to arise, such as YouTube influencers, and traditional economics of scarcity will organize around these new forms. This is a self-fulfilling prophecy, as our ability to measure GDP is a function of centralized states being able to levy taxes and see into economic transactions. Therefore, as new forms of value creation arise, our understanding of the economy will continue to evolve.
GDP is the sum total of all transactions where currency is exchanged, including wages and purchases. Each country has an agency responsible for measuring and reporting on GDP, in order to ensure consistency over time. GDP is estimated based on production, consumption and other factors, such as investment and net exports. Production is the output of different industries visible to the government, while consumption is the value of things purchased by individuals and businesses. These two factors are strongly correlated, as what is consumed is usually what is produced.
The transcript discusses how income and the informal economy are factored into Gross Domestic Product (GDP) as an indicator. It mentions gray and black markets, and how even if someone is paid under the table, the money still contributes to GDP figures. The question posed at the end of the transcript is how capital investment, such as laying down computronium to better serve AI overlords, would show up in GDP. The answer is that it would have a large influence on the economy, as it would require paying people to lay down the computronium and buying the necessary materials.
The transcript discusses GDP and the three main factors that are examined when constructing it: production, consumption and income. It also discusses the potential for GDP to become less useful due to advances in technology, such as DeepMind's Alpha Tensor which is able to do matrix multiplication faster than ever before. The transcript also questions how to account for investments and outgoing expenditure when it comes to GDP, and suggests that there are standard accounting practices for how income is defined in the formal economy.
Deep RL has found a way to multiply matrices faster than the algorithms traditionally used in school. This could potentially lead to trillions of dollars of value, as it could increase the efficiency of computation in data centers. GDP does not measure avoided costs, but there are other measures that can be used to measure cost avoidance and opportunity costs. Deep RL has the potential to revolutionise the way we do computations.
Nvidia can increase the value of its GPU by releasing a software update that doubles its speed. This means that the customer would have been willing to pay twice as much for the GPU if they had waited a day to buy it. However, if the software update is released after the purchase, it results in deflationary effects, as the asset is now worth half as much as it was before. This highlights the importance of incremental improvements, as the five percent of a gigantic number is a gigantic number itself.
Technology is becoming increasingly more advanced and cost effective, with Moore's law and other advancements leading to a deflationary regime. This could be catastrophic for markets, as why would anyone buy something if they know it will become obsolete soon? Despite this, people are still willing to buy things like phones, as the value and usefulness created by these products is worth it. This is a strange phenomenon, as it would be reckless to do this with other assets like a house. We are imperfectly coping with this situation, and it has led to conversations about the capitalist buying GPU.
GDP is currently used as a proxy for prosperity, but it is becoming increasingly less useful due to deflationary factors such as technology and energy. It is a measure of how much humanity is rearranging the world's matter, energy, and information from less useful configurations into more useful ones. As technology and energy become more efficient, the value of GDP decreases, making it a less reliable measure of success.
GDP is often criticised for not being a perfect measure of human happiness, however it is a useful proxy for measuring material prosperity. The complexity of the globalised economy means it is difficult to understand what is happening and decisions made by well-intentioned people may not be enough to improve the system. GDP can still be used as a measure of prosperity, but it is becoming increasingly difficult to capture all forms of production, consumption and income.
GDP is a way to try and extract some kind of signal from the radical uncertainty and inability to predict or comprehend what's going on, and it is the most terrible system apart from all the others. Despite its flaws and the terrible consequences of it being gamed, China's focus on GDP in the 20th century did lift hundreds of millions of people out of poverty. However, it has had terrible effects on the environment and other things, and local governments continue to try and fake the numbers.
The GDP is a measure of production and consumption, and while it is easy to criticize, it is not obvious that there is a better technique that leads to good outcomes. Alternatives, such as measuring throughput, do not guarantee that the activity is useful. Keynes argued for government stimulation of activity in order to keep people busy and make the economy grow, but this could lead to a lot of busy work such as digging ditches and filling them in.
GDP is a measure of productivity and consumption, which combines throughput and value. Throughput can be measured relatively easily, but measuring value is more difficult. Consumption is a good way to measure value, as it shows what is useful to people. However, it is difficult to think of an alternative to consumption for measuring what is useful. Robots and computronium are churning away at productive activity, so it is important to ensure that this activity is useful and not worthless.
People often discuss Gross Domestic Product (GDP) as a way to measure economic growth, as it is a reliable indicator of how well a country is doing. YouTube, for example, directs part of its ad income to musicians based on how many people watch videos with their music. Jared Lanier has proposed a micro-transaction based ownership model to replace the current ad economy. People often talk about GDP as it is a reliable way to measure how their personal circumstances were in the past and how they are now. It is also a way to compare different periods, such as the 80s, and understand why things were better then.
GDP is useful in many ways, from steering the ship to convincing citizens and guiding investment. At the population level, people understand that GDP growth means more job opportunities and better quality of life. In the West, however, GDP growth is low and not very meaningful. In China, people are able to feel the difference between periods of high and low growth. GDP is also a useful indicator of risk, as it can influence how people and organisations commit resources. It can shape decisions at the level of government, industries, companies and households.
GDP is often seen as an indicator of a society's prosperity, and as a way of reasoning about how likely life outcomes are. When measured GDP vastly overestimates prospects, it can lead to a loss of trust and a lack of optimism. This can lead to a self-fulfilling prophecy, where if people are optimistic then the GDP growth can become real. GDP can also underestimate prospects, leading to a situation where people are willing to take risks and invest in the future, even at the expense of the present.
The transcript discusses the possibility that the GDP figure may underestimate true prosperity in an economy due to technology increasing productivity and rendering it illegible to the formal economy. It is suggested that this could be evidenced by an increasing concern in the distribution of prosperity, as opposed to the mean or median. It is highlighted that people may feel that there is a lot of prosperity, but their own lives are not improving, leading to feelings of unfairness. This could be due to the vast wealth of people such as Jeff Bezos and Elon Musk, while the majority of people are still struggling to make ends meet.
GDP measures the current state of the economy, while risk-taking and investing are based on expectations of future performance. AI investments are a sign of optimism and are reflected in GDP, yet there appears to be a misallocation of capital in this area. Even with the current level of investment, the world is out of sync with what is already known, and the opportunity cost of not investing every available dollar into AI is immense.
Tesla's recent demonstration of their bot project, the Tesla botton, was met with criticism. Despite only having a small budget and team, the demonstration showed the world that more resources should be put into developing AI. This raises the question of what GDP is measuring, and whether it is accurately reflecting the value of different technologies. GDP is currently measuring things like YouTube influencers' ad revenue, which may not be the most valuable use of resources. It is important to consider what should be measured in GDP in order to accurately reflect the value of different technologies.

Raw Transcript

okay so he posed two questions last time and then Matt added a third so I'll just go through them briefly and then maybe we'll just take them in order so the first if I remembered it correctly was that um taking for granted the the hypothesis that rapid technological change is going to increase productivity rates perhaps not seen for um well maybe a century or more over the next few decades uh will GDP still be a useful measure for making decisions about how to allocate resources or just to get an idea of what's going on yeah maybe one thing that I want to dig into is um useful for what I guess um and it's also not the case that you know GDP is a perfect measure even without rapid technological growth um the second question was I think this is yeah probably the most interesting part uh can we see hints of this already like what would this look like in practice yeah that was the thing that really struck me so that the the hypothesis that that moving towards the post scarcity uh condition um into abundance would would render less and less of the product productivity um uh Society generates legible or visible to our existing metrics like GDP that's not my idea I don't think that's all that new and I think this is certainly something that folks have talked about if not completely explicitly then certainly in a roundabout way um in the past and then it didn't that idea didn't come for me that idea is something that came from my team from the other members of my team what what occurred to me and that I I think is that quite a bit more interesting is a question um is what what would what would the early signs of that transition or that shift look like um because if we can hypothesize what we look like what they might look like then we can ask the question you know in a sort of quasi-experimental uh manner you know are we are we seeing what we might expect to see um yet in in the real world right which I think is quite an interesting question and if even if we don't see any of the evidence now it's something to keep an eye out for yeah it actually it reminds me of some of the kinds of questions that Peter Thiel likes to ask so maybe I can briefly recapitulate some of that that might inform the context for question two so question three is uh as I said one that Matt raised at the end of the session and that is uh taking for granted questions one and two uh well rather that question one has a positive answer and that we will see this happen uh perhaps sooner rather than later
what do we expect to happen will people continue pursuing GDP just because well there's many reasons you can imagine that happening I guess we could talk about them or do we think that there actually will naturally emerge a competing metric which is more appropriate and I clarified with Matt on Discord afterwards that he does in 10 question three is as descriptive rather than normative so he's not uh I suppose suggesting we dig into the somewhat pointless question of is GDP good or not um and why is it bad because you know we I think we've all had that conversation 20 times already uh uh privately but rather well what do we think will happen um as a kind of to resolve this issue that's raised in questions one on two okay are there any high level things you want to introduce before we go back to question one and dig into it nothing at this level I think that's this is a great start okay okay um I suppose we've talked about it numerous times but maybe uh I'll say briefly what we mean by the scenario in which productivity increases you can clarify that if you want to and and then maybe I'll ask you what useful means in your mind so we have in mind that there are uh as Adam has outlined in published work and also here there's a Confluence of multiple disruptions in energy and transportation in agriculture and perhaps most profoundly in intelligence in Computing and how we solve our problems and the tools we use to do it in Ai and and its impacts and all of those I like to think of this as a potential disruption of human labor writ large right yeah and this seems quite likely to be much more profound than say the first or so-called Second Industrial revolutions where um yeah really the nature of civilization shifted quite a bit and their expectations for what Society looked like and how rich we would be also changed dramatically and we could easily imagine more profound changes coming okay so past some point uh it's sort of clear that the answer is yes to question one I suppose right um so yeah is it is it is it a question of like when we think GDP will become less useful or should we examine first do we do we assign any probability mass at all to the idea that GDP is currently construed will remain useful uh for 50 years is there any chance of that yeah I I scenario that I've been the name of as I've mulled this over is that we will um we will cling to what can be measured and in the same way that we might that that you know the attractor basins for for organizing
uh uh uh economy will center around the things that remain scarce and the things that remain visible or met or measurable or legible to the you know the the um to the metrics that we can bring to bear like GDP and other metrics is sort of the spotlight or not the spotlight sorry the street light effect where you're where you only see the things that that are illuminated by the tools of Illumination that you have um and so you end up only only observing you know in a small area I think that there's that one could could make the case that um uh just as we've sort of seen a rise in a service sector of the economy and we've seen new jobs and new occupations um arise it's just since the digital Revolution and the disruption of information and Communications lots of new um forms of value creation that didn't exist before I mean the quintessential one is YouTube influencer this this job um this is sort of the the the punching bag the straw man the stereo The Stereotype that's held up as an example of this is sort of this an occupation that was unimaginable prior to the internet um but it's now you know people are making a lot of money at it and and so that the the argument is that um new new forms of uh new forms of value creation will emerge and traditional economics of scarcity will organize around those and so you'll have sort of this this this natural tendency for um for uh the the way that we organize our society around scarcity to to essentially gravitate towards whatever new forms of um uh new forms of employment and other occupations can be organized around what remains scarce maybe let's take it so this is sort of a self-fulfilling it's a sort of a self-fulfilling prophecy there right it's it as it were is this again this is all sort of I'm steel Manning yeah a position that I don't really believe in let's step this is sort of one scenario that I can imagine let's take a step back sorry yeah go ahead go ahead yeah go ahead yeah I was just going to say that maybe it's worth uh attempting to define a little bit what GDP is and how it's measured I don't think I have a good grasp on it really but I guess I imagine the history of being something like the following that I mean until until you had centralized states that were able to Levy taxes and collect them and and sort of see into economic transactions well enough to tax them there probably wasn't a good Central understanding of what was happening in an economy and our ability to measure GDP is is kind of a function
of the government's ability to kind of be interested in economic transactions in order to tax them and perhaps things that aren't taxable or the government isn't interested in taxing maybe are less likely to be contributing to GDP just by virtue of the way it's defined and you know it's not only the thing you're in I mean it's not like people set out to say this is what counts in the economy and then let's go measure it and then that's GDP I have to imagine that it's more like what are we already measuring like what is convenient to add up to get a number and then we call that GDP I don't know if that's correct or not but I can imagine that um for the purposes of our current discussion do we do we mean by GDP like the sum total of all transactions where some sort of currency that is exchanged like the transactions in the market in the economy like all the wages all the purchases so my understanding is that that's that's broadly correct the the GDP is sort of triangulated upon and I don't I don't know what formal methods the um government agencies that are responsible for this bar but they're they're they're each country will have an agency that has a formal methodology for how it evaluates and then reports on GDP and the idea is that that method is consistent over time um in order to be reliable as as an indicator you know if you if you can sort of fudge it and always show growth or or show growth when it's convenient too then it wouldn't be very useful measure it so there's there's I think some some institutional pressure to kind of have this as an indicator for decision making be reasonably consistent in how it's done I don't think that necessarily means every country does it identically but my understanding is that it's it's is that GDP is sort of it's estimated based on uh measuring a few different things production is one of them so the output of of all of the different Industries in an economy that are visible to the um to you know the tax uh to the government basically and and specifically to taxation um the expenditures so these are again the sort of the value of things that that um individuals businesses and so on purchase so the the the the the that's consumption so basically it's production and consumption and those aren't necessarily identical but they're super super strongly correlated of course because basically what's what's consumed is what is produced and vice versa and then there's the other thing and I think you mentioned this Dan is that is that it is
I think income is also factored in like what is the income what what what do what what do individuals and what do businesses report as income and the thing that's tricky I think in all of this is that this is all uh that is all the formal economy but there's a there's an informal economy as well um some of which is sort of roughly speaking um uh well you sort of have you sort of have gray markets and then you have black markets right so you have gray markets that aren't necessarily horrifically illegal or nefarious um but they're nevertheless just not all that visible to to um you know into our public institutions so if you get paid under the table and you don't report cash and you don't report it then then that's part of the informal economy it's a part of the Great economy um but it's that's a little bit different than slave trading human trafficking um you know arms dealing that sort of thing and so the I don't know to the what extent uh countries try to include the informal economy both the gray and black markets in GDP as a measure I don't know that it's even in there at all in most of the methodologies but one could easily imagine that those those gray and black markets are pretty big it's probably relative to the two probably relevant to that discussion that uh you could right you could imagine that uh even if you get paid under the table you then go and buy Starbucks or a loaf of bread or something and that that money turns up somewhere so that transaction I suppose contributes to um to GDP figures right right so again it's it's this I think I think there's just a method for there's just some method for for combining these different specific measures into some into GDP as an indicator um and then obviously you want to try to apply that consistently from one year to the next so that you can make you know meaningful inferences about about you know about things so the the the so yeah so that's that's the method then the real question and I think this wraps down into question three is is you know what are you really measuring there why are you and why are you making decisions around that yeah here's a question sir uh when the cone Fields uh ripped up and covered with computronium to better serve our AI overlords uh how does that show up in GDP I mean I'm talking about capital investment I guess so uh a lot of capital investment I mean it has influences on the economy because you have to you know pay the meat bags to lay down the computernium and you know buy the
tractors to dig up the corn and all that uh but you know maybe 90 of that money is just going to tsmc right so you pay 50 billion to tsmc and they give you a bunch of computronium and then you you put it somewhere I mean those flows that are like foreign purchases and then locally just kind of capital Investments do they are they sort of uh they counted over multiple years or they just sort of show up in GDP figures in the year that the investment has made do you have any idea I mean it might be highly relevant you know I I really don't and that may be where the the that may be a reason why the methods as I've looked at them why uh production consumption and then also income are the three main things that are that are examined as part of constructing GDP as an indicator and income how how households how businesses you know individuals how they report what's coming in that's different than you know um what's being produced and what's being consumed and maybe what you're getting at which is you know how if you if you if you build some or invest or commit to some you know productive capital of some kind how do you account for that like where and when and how do you account for for you know that out that outgoing expenditure and then whatever comes in from that and when and over what period and so forth and I I suppose there are standard accounting practices for how that income is defined again assuming you're in the formal economy and you're reporting on it um yeah and not in the black market and and hiding it so yeah I think also it seems to me a key reason to believe that GDP is going to become less useful is is the following kind of scenario and uh uh just this morning I don't know if you saw deep Minds latest um assault on human dignity uh so they they have a paper Alpha tensor which um which does matrix multiplication much better than we figured out how to do it so there are uh I say assault on human dignity and jest but uh you know we people were quite proud that we figured out some clever way of doing matrix multiplication um which is sort of the basis of how gpus work for example and now deepminders through deep RL figured out algorithms for doing it that are 10 to 20 faster if I understand the headline claims which I don't know that it directly translates how the hell how is that even isn't that fairly fundamental mathematics is that even possible well uh there's actually not there's actually multiple different algorithms for multiplying two integers
so if you think about it really yeah if you think about multiplying a 10 digit number with another 10 digit number you know one way of doing it right that you learned in school oh yeah because there's actually there's actually a different algorithm um and it's faster so that's called current cyber multiplication so it's uh you don't tend to think about it but these basic operations are algorithms that you perform with the digits in that case uh and for matrix multiplication there is a kind of clever way to do it that we know in some cases but um yeah deep RL has found more or less essentially in the style of alphago right I just found good moves that you can do to multiply matrices Jesus now that that is amazing I don't know if that translates directly to improvements on existing Hardware right because existing Hardware may have baked in the old algorithm and I'm not an expert enough to think about that right now but and I haven't read the paper yet but you could imagine that either this or some future Advanced like overnight essentially increases the efficiency of computation in all the world's data centers doing GPU calculations okay putting aside like sorry for the noise yeah sure putting aside they like uh uh queasy feeling at the bottom of your stomach hearing that uh you could also just sort of say well that's like a small team of people sitting in London somewhere have suddenly produced potentially trillions of dollars of value right I mean it's it's it's not an exaggeration to imagine that uh that deep mind just paid off its entire budget you know like they they siphon like half a billion dollars a year or something from the parent company um and maybe they just you know maybe that counts as now they're in the black you know uh so I yeah I don't I don't think this is I don't think this is a stretch at all and I mean you think about what so so one of the things that I that I have on a list of observations to share um and there are a few things here but one of them was that I wanted to bring up and it's relevant here is uh GDP as a metric one of the things that it doesn't measure is avoided cost so if you figure out how to do something more efficiently and then you then you start doing it more efficiently um that is not something that GDP there are other measures of course for you know for avoiding um you know cost avoidance and and opportunity costs and you know things like Net Present Value and there's also there are no other member other number of indicators and
other tools that we can use to capture these things but certainly one of the things that's missing in GDP is anything to do with improving the efficiency uh um against the counter factual right so you're creating a new uh real scenario um uh relative to whatever some counterfactual feature might have been without that progress or Improvement and if you if you if you were to account for that well and you just raise this perfect example well yeah I mean if Deep Mind figures out a trick that can save five percent of the energy consumption at the data centers that Google uses it pays for itself it paid for itself of course it did right because of the because those those incremental improvements are so valuable because you know the five percent of a gigantic number is a gigantic number so um yeah uh anyway and and I don't really know how we capture anyway so anyway yeah sorry I'm just let's violently agree let me think this through briefly from first principles a bit so if you have a fixed GPU there it's not it doesn't have like an intrinsic value right you can't eat it uh but it it has a value in terms of say the number of computations per second you can do and then Nvidia is sitting there and saying ah you should give us a thousand dollars for this thing and then you think to yourself well you know I need to optimize ads on Twitter or Facebook and you know I think to myself well I'll increase my sales by this amount so sure I'll give you the thousand dollars now if if you bought that GPU and then the next day they sent you a software update that made it twice as fast well you'd actually be willing to pay twice as much for it you know uh or you would have been willing to pay twice as much for it right if if you had waited a day to buy it I mean Nvidia could jack up their prices by 2x assuming nobody else had that technology and you'd probably pay 2x right because actually what you care about the equilibrium you were you found that you know made you think you were interested to buy it was okay I value this number of computations per second at this price and they're going to sell it to me um but once you've already bought it and then they send you the software update and maybe they don't you don't have to pay because you know some crazy people just publish it for free um then quite cool it's kind of deflationary right it's like your asset actually became half as like maybe you're happy but maybe you're also pretty pissed because like on your balance sheet your assets now worth half as much according
to market prices well it in in in in a more abstract sense technological progress is especially that we've seen in Computing you know Moore's law has been um uh running for quite a while um and and you know things adjacent to Moore's Law like the you know Improvement in costs and memory and bandwidth and those sorts of things um those are all part of a massively deflationary regime right where every year or or two years what is 18 months 22 months I forget what the exact number is for for uh price performance of computing but basically I mean yeah if you wait if you wait around a year and a half or two um the the value of of any asset that you've got any computational asset that you've got um uh yeah I mean it it drops because of the progress that's being made and so there's this massive deflation deflationary Dynamic that's happening there and in in principle you would imagine that this would be this would be or could be catastrophic for for uh forward-looking markets and producers like oh boy well why why would anybody buy this thing if they could buy something better next year why won't they just wait and and you know if you if you if you Domino that across in many years you could naively reach the Assumption well why would you why would you ever step in and buy anything and then consumers face this problem like when's the right time to upgrade because you know it's really hard every because as soon as you buy something it's obsolete uh but what we've seen which I guess you know again it it bucks against that naive assumption is that the um it's so valuable to use these things in the meantime and so much additional value is being and usefulness is being created by these these products and services that it's still worth it despite the fact that they rapidly lose value um so people will still buy a 600 phone knowing that it'll be worthless three years from now um so uh and for lots of other assets that's not true like you wouldn't it would be it would be crazy and Reckless to do that with a house or or something like that but for for these technology assets yeah or tools or whatever you want to think however you want to think of them it is a strange thing already that we're that we're I don't know if we're I think we're probably imperfectly coping with that um right now maybe it's with contrasting you know the situation of your consumer buying an iPhone with the capitalist buying GPU uh this goes back to conversations we had many years ago reading uh that's capital I guess where the the
capitalist has to upgrade the technology right even if they don't expect you know they expect to be deflationary because they're in competition with other producers that's right and you can't especially for tech companies I mean you can't afford to fall behind that curve because catching up is exponentially harder than staying there and so there's just like the fear of missing out is so extreme that you know you gotta buy the thing even if it's going to depreciate in value super fast you you know you've got to buy all the gpus so you can throw some graduates at it and pray to God that you'll have some sort of deep learning capability in a few years before you know before deepmind eats your industry or whatever um so yeah I think it's it's very different to being a consumer just kind of eyeing and waiting and thinking hey I can get more for my buck later if you're also a producer you don't you don't have that luxury that's a good point yeah yeah let's come back to the question of utility So when you say maybe GDP will come yes less useful I guess we've now covered a few examples why that might be the case we could have also mentioned energy as a deflationary factor right I think that's an important one in GDP as well so what it would have I mean so what GDP becomes kind of useless uh most people I think well it's a kind of General perception in society that GDP is kind of an evil thing anyway and optimizing it is kind of the devil's work uh at least in American and Western Society in China they think quite differently but so yay GDP is dead uh what's wrong with that yeah I mean so right now we're sort of locked into GDP because it's the it is for all its faults it has this it has the the Merit of um being a pretty decent proxy for for very high level process to a first proximation uh prosperity and I I I I again it's sort of it's sort of the highest level and looking down I think ultimately that's what it's been in the past useful for and it of course what it does is it's looking at production and consumption and income um and maybe a few other bits and pieces in the methodology to kind of Cobble together and try as I said triangulate in on some some some combined indicator uh but but basically what though what each of those things are a reflection of is how busy humanity is at rearranging the stuff of the world the matter the energy the information from less useful configurations into more useful configurations and the more that we do that the more we Prosper to a first
approximation now the criticisms of GDP of course are that production and consumption income are not perfect proxies for human well-being and happiness and quality of life and so on and that's that is a way of saying that material Prosperity is not uh a perfect proxy for um it does not correlate perfectly to human happiness and that's perfect that's fair and then that's you know that's the basis of a lot of the criticisms of GDP however you have to be you have to be pretty silly I think um to put it kindly to say that Prosperity it doesn't matter at all or that it is Meaningful at all it's pretty damn meaningful and and um so I think that that's been at its if I if I had to put my finger on it I would say that is probably the the basis of gdp's usefulness historically is that it's basically been a decent imperfect but decent um way to quantify uh material prosperity and um uh the thing I think that that again my colleagues raised concern about it as it as GDP ceases to be able to capture all of uh you know the things that are happening is that it is it's basically it is basically going to become a worse and worse proxy for that thing it's supposed to be measuring which is prosperity um because too many too many forms of production too many forms of consumption too many forms of income and utility are just just going to cease being uh visible to it does that make sense yeah I think it's useful to clear the air a little bit by pointing at an assumption that a lot of people have talking about GDP and just deconstructing it which is I guess people have in mind uh I mean unless you've tried to run a large company or a country or involved and kind of thinking hard about that maybe people have a kind of naive idea that the system is simple enough that if you just put well-intentioned moral people in charge they'll look at the system and be like oh yeah we should arrange it like this that and that and then everything will work better um and that really is just naive because the complexity I mean it was naive 150 200 years ago it's certainly astonishingly naive today with the complexity of the system that exists in the globalized economy like you can't even tell what's happening nobody knows what's happening it's so inscrutable and so complex when people I mean when people on the TV or whatever or you read an article talking about what's happening in the economy it's about as dumb as people talking about what's happening in the stock market where they're obviously confabulating stories about like
post-talk justifications oh it's astrology yeah I mean it's it's silliness so if you if you buy into that impression you might think that okay if you just put good people in charge and they can see what's happening they'll arrange things in a smart way like as if like that might be possible but once you see through the veil and realize nobody really knows what the hell is going on um then the idea like the question is well what else do you do I mean okay how are you supposed to have some sort of idea of what choices to make in the light of that radical uncertainty and radical inability to predict or even comprehend what's going on you so GDP represents it seems to me the attempt to try and extract from that some kind of signal which you can then attempt to nudge in some direction or another like from a position of radical humility about how little you understand what the hell is going on to then like just try and push the system in some direction [Music] um and then the question becomes it's a bit like democracy right like yeah it's the the most terrible system apart from all the others uh I guess that's supposed to be Churchill's quote right uh I mean I I completely agree with the criticisms of GDP and I think it's related to but different than the the terrible practice of just kind of trying to promote science by counting citations or something so it's it's kind of easily gamed but then everything else is more easily gamed in I mean GDP I'm talking about right and so it's you kind of have to argue for in a positive way for what you think the alternative is uh rather than just um yeah I mean everybody sees the problems with GDP they're obvious right it's obvious the distortions that happen how it's gamed but I think like China in the 20th century is just this great illustration of okay despite all its flaws and the terrible consequences of it being gamed and the fact it can be gamed like grabbing that number and trying to dial it up in in a fairly like direct way literally did lift hundreds of millions of people out of poverty that terrible effects like on the environment on many other things but you know in some total I think it was a good thing for Humanity that you know hundreds of millions of people who were previously struggling to eat no longer struggle to eat so it's and they did that by this like mono maniacal focus on GDP because it's hard to fake even though local governments in provinces tried like mad to fake it and continue to fake it like at some level the the best way to fake
it was just to make the economy grow and they they did it you know so it's like it's easy to criticize GDP but at least the general approach it's not so obvious to me that I can see it like a a better technique that actually leads to an algorithm for making actions in the world that lead to good outcomes yeah I mean the the uh I guess in my mind there are two things that are two things that are happening one I I okay so set aside so I mentioned three things and I'm just looking here at for example on the UK's office for National statistics what GDP is comprised of How It's constructed and production consumption and income are listed and there are other similar sites like that set aside income for a moment in the role that it plays just think about production and consumption people make stuff and people buy stuff and consume it use it utilize it um why would that be a good measure or a useful measure as a quantity of something to maximize the the and and what could be an alternative you might this is so so the the when you combine those two things together production consumption you you you could imagine trying to measure something like throughput like what is the energy throughput what is the material throughput what is the information throughput um in this system if it were if it were if you were to think of it as an organism or a machine or an engine or something like that what is it what is it what is its metabolism look like what is its throughput it's taking in you know sort of um you know these inputs and uh it's it's somehow exporting entropy locally and it's creating greater order or usefulness or something or something something analogous at any rate to to entropy I suppose and then it's exporting this you know this this waste or or less useful stuff and and uh the result of that is prosperity from the from a human perspective the thing the thing with the thing there and I think we could so for example I think we could uh imagine alternative ways to measure throughput the problem is that is that throughput alone doesn't care it doesn't necessarily um guarantee that the activity is useful so you could imagine and I think GDP has been criticized for this too that that um uh you could have a lot of busy work you could dig ditches and fill them in to use the proverbial example right um You can pay people to dig ditches and fill them in and keep them busy and um I think that was Keynes um argument for just something different arguing for government stimulation of
the economy but um but to keep the throughput going to keep the blood pumping through the system as it were and the problem with that is that is that there's no guarantee that throughput alone is useful to product or or yeah well that is useful that it's valuable that it's creating value you can imagine doing a lot of useless busy work um and so if we were going to the the I think the power maybe the luck of stumbling across GDP as a measure is that it combines those two things it it measures productivity which is which in his sense is the is the throughput or the busyness of the system and um but it also looks at consumption how and how consumption is paired to that because if you I think this is one of the arguments again for the marginal uh theory of value and against the labor Theory value is that things are only as valuable as as as consumers are you know willing to um enact right so it a pair of shoes that's crap that nobody wants to buy they have no value doesn't matter how much labor we put into making them if nobody will buy them and everybody thinks they're rubbish then they're not useful and so you the consumption is the is the the part of GDP that that um ground truths the reality of the usefulness and if and and if people won't buy things if people don't want things if they refuse to consume them well then they're not useful so you you would have to pair some measure of throughput with some measure of useful of usefulness or utility or or other otherwise some something about meaning value that's meaningful to to the beneficiaries us individuals families communities whatever of all of that activity that throughput that productivity and so the the second part there is what I think is hard measuring throughput I don't think is all that difficult I think we could measure how much energy and material and is going through the system maybe even how much information um uh you're welcome to Splash bonds um but I can't I I can't think of a trivial super easy way other than consumption a better alternative to consumption for measuring what is useful to people and and so if it were me that would be the thing I would think about uh as you know when we're trying to imagine alternatives to gdps how do you how do you measure whether all of this productive activity that robots are and are are churning away at and the computronium is is is puzzling away at um figuring out how do we how do we ensure that that's useful and not not worthless monkey work busy I guess the
way that YouTube works and maybe also to some extent roadblocks is is maybe a hint of that right where YouTube directs part of the ad income to musicians whose music shows up in videos uh on the basis of how many people watch videos that have the music in them so that's a that's kind of a much more direct measure in some way right it's sort of like the idea that uh you could almost quantify the degree to which having a track from I don't know Beyonce on your on your video increases the probability people will watch it for 10 seconds longer uh that turns up and just the the raw amount of time people watch it and value can be fed back somehow you can imagine systems like that I know that Jared Lanier has written extensively about this um I forget the title of his book but he was proposing that the the ad economy online is so pernicious in its incentives that it would be better to move to a kind of micro transaction based ownership model where people are sort of compensated in tiny increments for the degree to which their data is used and their Creations are viewed or interacted with by other people you can imagine a future economy that that works in some way like that right it would be quite an effort to somehow shift I mean right now it feels a bit like those parts of the economy exist but they're a bit peripheral right I mean YouTube certainly dominates uh culture and entertainment but it's not dominating the economy right its income is still a relatively small percentage of Google's overall income if I when that restaurant um yeah that was my comment on that maybe let's let's zero in still more on the on the useful part um like okay well where does where does this I mean obviously the technocrats want to know GDP because they need to set interest rates or make decisions about budgets but let's let's sort of zoom out of that bubble and ask from the point of view of an average citizen like why why do people talk about GDP I mean here's a sort of story maybe um so we have a memory uh as voters that you know maybe stretches back to our teenage years and we sort of remember different periods and the recessions and how the labor market was and how our personal circumstances were and we hear our parents talking about similar things from their own experience and we kind of know how that tracks with GDP growth so I think a lot of people have the sense that you know maybe things were better in the 80s you know and and one of the reasons was that maybe GDP growth actually was higher
than and people understand I think at a sort of mass population level that in those periods where GDP was growing that meant that it was easier to find a new job you didn't have to stress so much about paying for daycare for your kid maybe you know people really just did kind of go and get a job after school and they didn't have to worry for 10 years beforehand which job they were going to get because maybe they were just a lot of jobs so people I think have a good there's a sort of good population level understanding of how to translate the politicians speak about GDP into like our personal histories and the histories of people around us and that seems to me one of the main ways in which we consume the idea of GDP right where we cash it out in quality of life quality of life improvements um and that's become less true in the west because we just have you know very low growth or kind of growth that's not real um but in China it's definitely true that you know the the periods of 10 growth and the periods of three or five percent growth everybody knows the difference between how that felt what it meant for people's careers and their prospects it's so obvious to people what the connection is between GDP growth and everything they care about in their lives um so that's clearly a way in which it's broadly useful and not just from a technocratic steer the ship kind of perspective so that's two ways it's useful can you think of any others like steering the ship and like convincing voters or citizens to like not overthrow you uh I think the other way it might be useful um perhaps at all levels uh is in terms of guiding investment and I mean this sort of like like you know what is your confidence in the future and how much risk are you willing to take how much investment and what kind of Investments are you willing to make are you are you know would you want short-term very you know um low risk high security Investments because the future is extremely uncertain and and and and frightening and potentially uh very bad or are you confident enough in the future that you're willing to take risks or make longer bets or whatever it might be and I think that this can happen at the level of government at the level of entire Industries at the level of individual companies and then of course the level of even down an individual households and so GDP may be a good uh uh uh indicator of the near future in terms of risk and and that may shape how we choose to commit resources you know
um um make outlays to purchase a shot at more opportunity or more growth or more value Creation in the future at the expense of the present right and and if if you think that things are going to be roaring um in a couple of years or if you think they're good enough now that they they can Coast for a few years and and it's a pretty good bet well then maybe you're willing to to um take those kinds of risks uh put that skin in the game and so that's another that's one other thing that I can work way in which I can think of sort of GDP as a general barometer for uh the the prosperity of society is useful because if you think if you think Society is very prosperous you're probably will need to take more risk if you think things are not prosperous or they're heading in the wrong direction or whatever then that's more of a batten down the hatches um kind of uh uh situation and and yeah and then change your posture towards risk and towards the future yeah okay so that's kind of gdps coordination signal in part yeah all right so maybe let's we've only got 10 minutes or so so maybe let's come back to uh if what would it what would it look like for okay suppose GDP is no longer a good measure of those things so what does that mean in those three scenarios so GDP is a means of reasoning about uh uh maybe GDP as a figure leads to the idea of GDP growth and being able to measure GTP growth and understand how that correlates to your likely life outcomes and that's also a way of making decisions kind of personally similar to what you were saying about maybe at the more like Capital level um if if GDP is no longer a good indicator of How likely you are to get a job or to be able to switch jobs or what the payoff for your education will be or the prospects of your children um I suppose it could be one of two ways right so maybe let's think through both scenarios so suppose the measured GDP vastly overestimates your prospects so GDP growth seems to be 10 a year but in fact it's zero or you know uh I mean Chinese people kind of find themselves in this position that the advertised GDP growth maybe is five but if you ask people they kind of all think it's zero or negative maybe currently um so what does that mean for you personally well it means that it's kind of a loss of Faith somehow like a loss of trust right you just kind of stop uh that that the ability of that number to kind of coordinate everybody to be optimistic and in some ways a self-fulfilling prophecy right if everybody is kind of optimistic then
maybe growth actually happens that kind of fails and people become cynical and they just start to not trust that number and absent that number maybe they just kind of batten down the hatches expect the worst I don't know maybe that's once or no right I mean yeah the willingness to accept risk is certainly a part of the dynamic of that of the self-fulfilling prophecy there for sure right because that's if the GDP figure uh overestimates I mean it becomes decoupled from real measurement of the economy and is an overstatement if it's a vast understatement uh so the GDP stays flat but actually there's an insane amount of growth uh or just a high level of productivity that it matters how it's distributed right I mean maybe as an individual it's actually still rat maybe like your life is actually described by that that GDP figure like maybe you actually have very poor prospects and the the you know the the stuff that's not being measured by GDP is is over there in California and doesn't benefit you really um so this this seems like worth disaggregating this case right where the measured GDP is an underestimate well I think we're saying that that's what we that's what our hypothesis our working hypothesis here is is here is that it if anything the technology increasing productivity and rendering things illegible to normal the formal economy um that that that will that GDP will therefore under report uh the true productivity and consent and true consumption or they're true and by extension the true prosperity of an economy I think it's really interesting this point you break you bring up Dan about it about if if that were the case one could imagine that a sign of uh of that might be um an increasing concern in the distribution of prosperity as opposed to the mean the per capita new GDP GDP and GDP per capita are are you know it's typically it's a very aggregate measure of course that's one of its flaws and points of criticism is that it doesn't you know it it looks at the mean and perhaps the median it doesn't really capture the distribution um of prosperity and if at least my just anecdotally and my intuition that is what a lot of people are really pissed about is man there's a lot of prosperity but don't tell me the world's all getting better don't tell me about that Prosperity my life still sucks it's not fair you know you have you have Jeff Bezos and Elon Musk around worth 200 billion dollars but you know I'm still struggling to make ends meet this sucks and yeah that might be a
a a an interesting Harbinger a a a sign of of this perhaps yeah I wonder if we should wrap up but I'm wondering if um to some extent well it's not really true GDP right I mean GDP measures or tries to measure what it's like real like what what it's like right now whereas when we were talking about your propensity to take risks or invest that's about a that's a question of what you expect the GDP to kind of do over the future right so it's a little bit separate from believing that the current number really represents it's kind of more like the trajectory right um and I think I think one example of okay you could say that there's a there's already a gap somehow in okay I would say my estimates of future GDP are much larger than most peoples as a result of knowledge I have and conjectures I have or beliefs about what's likely to happen in AI in The Fairly short term right so that's a kind of knowledge that you know I will make personal decisions based on that is a bit I mean it's of a similar kind to let me what I'm getting at is that some of these things in GDP growth uh it I mean if it if it measures investment it's also partly a summary of the level of optimism of other people right it's kind of like there's a you can look at other people's Investments and be like well they know something I don't I guess but they certainly think things are going to be good in the future so who am I to disagree let me take that risk and start a new business to the extent that investment is part of GDP it also tracks that to some degree right right and you could say that well people are making large investments in say Ai and you know new companies based on cheap energy and to the degree those Investments exist um then that's also tracked by GDP but I think we have even even with the wild investment in AI a massive misallocation of capital based on what we already understand like actually there should probably be a hundred times more Capital going into these things right um yeah even even based on fairly conservative I mean really really the world is quite out of sync with what we already know um so in that sense GDP I agree is a vast underestimate because it that investment doesn't exist because you know in a sense there's like any dollar that's not invested in these kinds of places is just going to right it's just going to go away any dollars that's right yes exactly it's just such a there's such a massive opportunity cost to not putting every death spare dollar we've got into these things and and I I I I I
would add on top of that AI I think AI is obviously the main the main thing because it's the software but um uh I think I think the hardware is is is got to be part of the picture so for example people were mocking Tesla's recent uh uh demonstration of their bot project the Tesla botton and some of that criticisms may be justified some of it's a bit you know a bit harsh because they've only been working on it for a little while but the main thing that that the main thing that I I uh took from it was just disappointment that the entire world isn't already seeing Oh my gosh we should be throwing hundreds of billions of dollars at this why is it only Tesla with the with so far a peanuts budget and the 20-person team and they're getting mocked because because of it I mean the the joke is on all the first person to come out of a you know they dig a mine and it's a gold rush and they come they come out of the mine and they're they're hauling their gold out in buckets and you're like these buckets buckets buckets are so lame you know like who cares right right exactly exactly okay we should we should leave it at that right I know we're a couple of things in here that we can think about next time um the uh the I think I think this will find it falls under the idea of the composition of GDP I think one of the things that we ought to think about this is for question two can we see hints of this already I think the composition of GDP is revealing there like what are we measuring and here's a fun dog experiment wind the clock back 100 or 200 years and and ask a policy maker or a senior person in decision making capacity then um if if they think the things GDP is measuring are silly are are just just ridiculous or not now if if we're not measuring things like you know being able to safely navigate your way across the city effortlessly with a magical device in your hands you know okay the map function on your phone that's that's free it's completely invisible to UDP if the GDP is not measuring that but it's measuring the millions of dollars of AD revenue on a YouTube influencers uh channel that seems ridiculous so if the composition of GDP is becoming increasingly silly maybe that's a sign like we're measuring why are we measuring these things well because we're that's what we can measure still with this tool so if you're a got a hammer and all you see all you can do or see everything with nails and then you've nailed everything and there's only a few things left they don't really