While most people are objectively better off with $100K now than in 1959, I think that most people are more after status, and having an annual income of $100K in 1959 definitely had more status than it does now. After basic needs are met, people increasingly focus upon status games. Let's be honest about it, the consumer economy relies upon status games to drive consumer spending.
I agree that status is important, but there are some important qualifiers. A poor country like Haiti has exactly the same proportion of its population in the top 1%, top 10%, top 20%, etc., as does the US. So absolute levels also matter a great deal.
I think that Bryan Caplan pointed out that people often prefer to move to rich neighborhoods (or rich countries), which lowers their status relative to their neighbors . Admittedly, rich neighborhoods have other amenities like lower crime, but the preference of people for rich neighborhoods shows that relative status is not what people care about the most.
> the preference of people for rich neighborhoods shows that relative status is not what people care about the most.
It shows that it's not what people care about most at the moment they are making big life decisions. But it doesn't show much about what people care about in their moments of living life. Daniel Kahneman points out that people care more about the size of a house and yard at the moment of buying a house, but that the commute and walkability make a bigger difference for people's quality of life as measured while they are living in the house.
Agreed. Taking a shift to lower status but getting far higher standard of living and health care would be worth it - to me at least. But once people are somewhat comfortable, the status game rises in importance. And it does not seem that econ can deal with the status game well, it is zero sum and a mix of psychological and social - and perhaps religious. Rich neighborhoods typically have better schools and lower crime - which can make being relatively poor more bearable - but a bit hard on the kids, who definitely feel poorer - I know, I have done it much of my life. But I also am not a status hound.
The fact that people in are equally happy (or unhappy) today as in the past shows that it’s mostly relative and not absolute, I’d argue.
But agree with the overall point that inflation adjustments are overly simplistic. What’s the value in 1959 dollars for a Spotify subscription with access to today’s music?
That is indeed a constantly underrated problem by economists: people care a lot about relativities. I also suppose that the equivalent of $ 100 k in 1959 allowed one to buy more personal services. My connections in India, which I would call (upper)-middle class, naturally rely on servants to do the cooking, cleaning, etc.and have more time to devote to intellectual pursuits.
I'm curious about something. If you think macro is mostly "solved" with only some small scale empirical questions about specific events or the influence of factor x or y to quibble about, do you think there are large scale, important mysteries left in the field of economics aside from pure macro? Something analogous to physics where your typical good physicist would tell you we genuinely don't know how to merge QM and general relativity etc. Or do you think most areas of economics are relatively settled?
I think economics has gone almost as far as it can in areas that are feasible to model. We will probably never be able to predict extremely complex systems due to "butterfly effects".
I wouldn't say that there are no important questions still outstanding, rather I'd say we have the basic tools (models) in place. Take economic development, which is a very complex area. I don't think we know precisely why some countries do better than others, but the major pieces of the puzzle (good institutions, human capital, physical capital, national resources, technology) are in place. There are still important outstanding questions such as the optimal amount of "industrial policy", but the question of North vs. South Korean economic policy is largely settled.
What do you think about the reverse causality problem that is inherent in the "good institutions" hypothesis in development economics? Because good institutions might stem from human capital and vice versa. Or do you find the good institution argument powerful? Are there any new or contrarian ideas that you have about this development topic?
Consider Korea and Malaysia. North and South Korea suggest that institutions matter a lot, independent of human capital. Malaysia suggests that human capital matters a lot, independent of institutions. (Especially the success of ethnic Chinese residents of Malaysia.)
So I'd say causation is complicated, and probably goes in both directions.
Good question. If I were to ask this question to one of my favorite professors, he would answer it by saying, "In economics we care more about the questions rather than the answers."
Are there any old school games that AI hasn't become the best in the world at? I thought bridge, but then I looked it up and AI now beats the world's best bridge players.
So what they do is they search for consensus view on the topic all over the internet and then present to you. But in case of macroeconomics you have 5+ competing theories - so which one should LLM select?
Who is gonna be "validator" in training set or human in the loop to confirm the result? You have them for coding tasks as they hired engineers to check if response is valid and tweak it if necessary during training phase. Even before that they used most upvoted answer on stackoverflow as a guide.
Validator for Macroeconomics ultimately is real world experiment which is not possible to conduct on the whole country scale. That would mean that FED Chairman would wake up one morning and decide to apply NeoFisherian policy in the next 10 years in order to test if that policy is correct.
I'm also skeptical. But AI researchers know far more about this than I do, so I think it's reasonable to at least entertain the possibility that they are correct.
I have the impression - but I may be missing something - that your simple macroeconomic framework does not include the need for fiscal policy to ‘back’ monetary policy and the possibility that unsustainable fiscal policy result in inflation explosion irrespective of what the central bank does.
I view monetary dominance as a reasonable assumption for the US, but that could change if fiscal policy does not improve. In some developing countries you have fiscal dominance. Venezuela, for instance.
Per complexity science, complex systems are often "robust but fragile": robust to "shocks" to the periphery, but "fragile" to shocks in the core.
"Complex" systems frequently are easier to predict outcomes from, in percentage terms, but when the predictions go wrong they go *really* wrong. Less complex systems are more idiosyncratic, but it's also less important to predict their outcomes.
Because "AI" models are out-of-sample statistical predictions trained on a subset of human knowledge, they will represent human-like ways of processing information. Meaning: they will get many things right but some very important things wrong.
We seem to have seriously exaggerated how much syntactical systems can actually replicate semantic thinking. So, Large Language Models as AI might have been a touch oversold.
We need macroeconomists to do the valuable work of convincing politicians to act less stupidly. I’m not sure if AI will ever manage to be more than an assistant in the task of “persuading humans”.
And 1959 was before the "Nixon shock," still on the gold standard, and fiat money had actual backing, then it became a "silver certificate," and now, just a transaction media. Notional vs nominal value - what IS money? As the population increases, the number of units in the price (not cost) of a loaf of bread must increase; the loaf of bread is one of the best currency pegs. So to compare 1959 to now, look at the price of a loaf of bread and the population level and the back-stop for the currency.
You're not a programmer, a mathematician, a computer scientist, or a cryptoghrapher. You know absolutely nothing. Your own colleagues called you a technophobe, according to wikipedia.
I wasn't envisioning an ASI as literally being the central bank, although I suppose that's possible. I presume that markets will be incorporating ASI views, so I'd still favor a market guided policy.
Not LLMs, but AI, can certainly be a valuable improvement to Fed policy, if you view Fed policy as being the transfer function in a servo system (the economy being the "plant"). In this case, AI would be used to determine the fuzzy rules in a fuzzy logic implementation of servo design. You will probably discover that the Fed's desire for credibility is grossly overvalued, and that sudden switches and backtracks are (as in the fuzzy control of a double inverted pendulum) part of what contributes to stability.
While most people are objectively better off with $100K now than in 1959, I think that most people are more after status, and having an annual income of $100K in 1959 definitely had more status than it does now. After basic needs are met, people increasingly focus upon status games. Let's be honest about it, the consumer economy relies upon status games to drive consumer spending.
I agree that status is important, but there are some important qualifiers. A poor country like Haiti has exactly the same proportion of its population in the top 1%, top 10%, top 20%, etc., as does the US. So absolute levels also matter a great deal.
I think that Bryan Caplan pointed out that people often prefer to move to rich neighborhoods (or rich countries), which lowers their status relative to their neighbors . Admittedly, rich neighborhoods have other amenities like lower crime, but the preference of people for rich neighborhoods shows that relative status is not what people care about the most.
> the preference of people for rich neighborhoods shows that relative status is not what people care about the most.
It shows that it's not what people care about most at the moment they are making big life decisions. But it doesn't show much about what people care about in their moments of living life. Daniel Kahneman points out that people care more about the size of a house and yard at the moment of buying a house, but that the commute and walkability make a bigger difference for people's quality of life as measured while they are living in the house.
Agreed. Taking a shift to lower status but getting far higher standard of living and health care would be worth it - to me at least. But once people are somewhat comfortable, the status game rises in importance. And it does not seem that econ can deal with the status game well, it is zero sum and a mix of psychological and social - and perhaps religious. Rich neighborhoods typically have better schools and lower crime - which can make being relatively poor more bearable - but a bit hard on the kids, who definitely feel poorer - I know, I have done it much of my life. But I also am not a status hound.
The fact that people in are equally happy (or unhappy) today as in the past shows that it’s mostly relative and not absolute, I’d argue.
But agree with the overall point that inflation adjustments are overly simplistic. What’s the value in 1959 dollars for a Spotify subscription with access to today’s music?
That is indeed a constantly underrated problem by economists: people care a lot about relativities. I also suppose that the equivalent of $ 100 k in 1959 allowed one to buy more personal services. My connections in India, which I would call (upper)-middle class, naturally rely on servants to do the cooking, cleaning, etc.and have more time to devote to intellectual pursuits.
Correct, but relative status should not be a part of aggregate inflation calculations, as it's a zero sum game.
When Biden's prostate cancer was announced, I felt status over the fact that I seem to have better health care providers than the President does.
I'm curious about something. If you think macro is mostly "solved" with only some small scale empirical questions about specific events or the influence of factor x or y to quibble about, do you think there are large scale, important mysteries left in the field of economics aside from pure macro? Something analogous to physics where your typical good physicist would tell you we genuinely don't know how to merge QM and general relativity etc. Or do you think most areas of economics are relatively settled?
I think economics has gone almost as far as it can in areas that are feasible to model. We will probably never be able to predict extremely complex systems due to "butterfly effects".
I wouldn't say that there are no important questions still outstanding, rather I'd say we have the basic tools (models) in place. Take economic development, which is a very complex area. I don't think we know precisely why some countries do better than others, but the major pieces of the puzzle (good institutions, human capital, physical capital, national resources, technology) are in place. There are still important outstanding questions such as the optimal amount of "industrial policy", but the question of North vs. South Korean economic policy is largely settled.
What do you think about the reverse causality problem that is inherent in the "good institutions" hypothesis in development economics? Because good institutions might stem from human capital and vice versa. Or do you find the good institution argument powerful? Are there any new or contrarian ideas that you have about this development topic?
Consider Korea and Malaysia. North and South Korea suggest that institutions matter a lot, independent of human capital. Malaysia suggests that human capital matters a lot, independent of institutions. (Especially the success of ethnic Chinese residents of Malaysia.)
So I'd say causation is complicated, and probably goes in both directions.
Good question. If I were to ask this question to one of my favorite professors, he would answer it by saying, "In economics we care more about the questions rather than the answers."
I recall a professor telling me that the hardest task is thinking up good questions, not good answers.
Simple and inspirational!
Chaos theory.
Are there any old school games that AI hasn't become the best in the world at? I thought bridge, but then I looked it up and AI now beats the world's best bridge players.
The issue with LLMs is that they lack grounding.
So what they do is they search for consensus view on the topic all over the internet and then present to you. But in case of macroeconomics you have 5+ competing theories - so which one should LLM select?
Who is gonna be "validator" in training set or human in the loop to confirm the result? You have them for coding tasks as they hired engineers to check if response is valid and tweak it if necessary during training phase. Even before that they used most upvoted answer on stackoverflow as a guide.
Validator for Macroeconomics ultimately is real world experiment which is not possible to conduct on the whole country scale. That would mean that FED Chairman would wake up one morning and decide to apply NeoFisherian policy in the next 10 years in order to test if that policy is correct.
I agree, but this post was not about current LLMs, rather I was discussing a hypothetical future ASI that is far smarter than humans.
I see, thanks. But I just don't see how can they get there riding on LLM wave.
I'm also skeptical. But AI researchers know far more about this than I do, so I think it's reasonable to at least entertain the possibility that they are correct.
You seem to have convinced Grok: https://x.com/i/grok/share/4RCm8Ee3hhe8IfHix4F8LGMR4
Is that the correct link?
Their linking is a bit clumsy. If you scroll to the bottom, you'll see that it praises NGDPLT and name checks you!
Thanks.
I have the impression - but I may be missing something - that your simple macroeconomic framework does not include the need for fiscal policy to ‘back’ monetary policy and the possibility that unsustainable fiscal policy result in inflation explosion irrespective of what the central bank does.
I view monetary dominance as a reasonable assumption for the US, but that could change if fiscal policy does not improve. In some developing countries you have fiscal dominance. Venezuela, for instance.
Per complexity science, complex systems are often "robust but fragile": robust to "shocks" to the periphery, but "fragile" to shocks in the core.
"Complex" systems frequently are easier to predict outcomes from, in percentage terms, but when the predictions go wrong they go *really* wrong. Less complex systems are more idiosyncratic, but it's also less important to predict their outcomes.
Because "AI" models are out-of-sample statistical predictions trained on a subset of human knowledge, they will represent human-like ways of processing information. Meaning: they will get many things right but some very important things wrong.
Good points.
We seem to have seriously exaggerated how much syntactical systems can actually replicate semantic thinking. So, Large Language Models as AI might have been a touch oversold.
https://hollymathnerd.substack.com/p/the-golden-retriever-of-software
https://hollymathnerd.substack.com/p/training-the-golden-retriever
We need macroeconomists to do the valuable work of convincing politicians to act less stupidly. I’m not sure if AI will ever manage to be more than an assistant in the task of “persuading humans”.
You are irreplaceable.
And 1959 was before the "Nixon shock," still on the gold standard, and fiat money had actual backing, then it became a "silver certificate," and now, just a transaction media. Notional vs nominal value - what IS money? As the population increases, the number of units in the price (not cost) of a loaf of bread must increase; the loaf of bread is one of the best currency pegs. So to compare 1959 to now, look at the price of a loaf of bread and the population level and the back-stop for the currency.
Large Language Models as AI might have been a touch oversold.
https://hollymathnerd.substack.com/p/the-golden-retriever-of-software
https://hollymathnerd.substack.com/p/training-the-golden-retriever
But we are discussing a theoretical possibility here, so it’s fine.
$100k in 1959 means I could have more time, either in retirement or by choosing to work less. Much better than anything else that could be bought.
Not twice this day
Inch time, foot gem
Please stop talking about AI.
You're not a programmer, a mathematician, a computer scientist, or a cryptoghrapher. You know absolutely nothing. Your own colleagues called you a technophobe, according to wikipedia.
The future is simply this:
Marcoeconomics will cease to exist.
Blockchain will replace you.
Mises will be taught in classrooms.
The funny thing is that even Mises wouldn't agree with you.
It's all too appropriate, and aggravating, that this post in particular is getting trolled by a Scott "Sunmner" doppelganger bot.
I'm flattered.
There are three different cases, with different implications for NGDPLT guardrails.
1. ASI is the central bank, trading with lesser intelligences
2. Human central bank is trading against ASIs
3. ASI central bank trading with other ASI counterparties
Is the scenario described in your post closer to the 1. or 3. ?
I wasn't envisioning an ASI as literally being the central bank, although I suppose that's possible. I presume that markets will be incorporating ASI views, so I'd still favor a market guided policy.
Not LLMs, but AI, can certainly be a valuable improvement to Fed policy, if you view Fed policy as being the transfer function in a servo system (the economy being the "plant"). In this case, AI would be used to determine the fuzzy rules in a fuzzy logic implementation of servo design. You will probably discover that the Fed's desire for credibility is grossly overvalued, and that sudden switches and backtracks are (as in the fuzzy control of a double inverted pendulum) part of what contributes to stability.