Ok, phew, yesterday's post by you was great and this is, well, impressive for AI but still slop. I'm dismayed by all the comments here saying the opposite though!
If you had left us to guess which was which, I would have chosen this version as the one you wrote. If this is really from ChatGPT (which I've never used), I'm truly impressed.
On Japan and cleanliness. You've picked the one example that proves too much. Japan is indeed remarkably clean — and Caplan himself used Japan as an example, though he drew the opposite lesson from it. Japanese cleanliness is a cultural norm enforced by the Japanese themselves, including its low-skilled workers. It tells us something about Japanese civic culture. It tells us nothing about whether low-skilled immigrant workers cause filth elsewhere, because you haven't controlled for anything. Are immigrant-heavy cities dirtier than non-immigrant cities of equivalent wealth, governance quality, and density? That's the relevant comparison, and you haven't made it. Correlation between immigration and observable urban problems typically dissolves when you control for poverty rates and municipal governance. Blaming the maid for the mess in a badly-run city is not analysis.
On welfare costs exceeding marginal product. This is the strongest of your three points, and it deserves a serious answer rather than dismissal. But notice what it actually argues: it's a case against the welfare state, not against low-skilled workers. The worker generates real value. The state then mandates transfers on top. If the combination is fiscally negative, the remedy is to fix the transfer policy — not to ban the worker who was contributing positively before the state intervened. Shooting the worker because the bureaucracy is expensive is a strange solution. Caplan has been consistent on this: the fiscal argument is an argument for reforming entitlements, not for excluding workers.
On "enslaving" high-skilled workers through taxation. This rhetoric proves far too much. By the same logic, native-born low-skilled workers "enslave" high-skilled ones just as much — they too consume healthcare, education, and retirement benefits. Do you want to expel them too? If your principle is "no one whose taxes received exceed taxes paid may participate in the economy," you've just argued for abolishing most of the welfare state and excluding a large fraction of the native population. That may be a consistent position, but own it explicitly rather than applying it selectively to immigrants.
- That was all Claude 4.6, of course. to stay in the spirit of todays AI post.
I think this is obviously the AI-written version! Surprising how many people in the comments disagree. The first article was very focused, while this feels full of fluff and flourishes, and doesn't reference economic theory or research as much.
Interesting 2 day exercise. I think it would’ve been more interesting to make this blinded for the reader, then take a poll of which is Caplan and which is GPT.
I think GPT did a pretty good job of capturing the diction and meter of your writing style. And the argument seems to fit as well.
But I think both you and GPT-pretending-to-be-you are conflating “low skilled worker” with “low skilled immigrant”. Both make a good case about the necessity of low skilled workers. I don’t think either have made the case that such necessity justifies “open borders”. I would say that we need flexibility in the workforce such that low skilled jobs are filled, and we can accept enough low skilled immigrants to fulfill that need. But I don’t see your argument justifying accepting low skilled workers as immigrants beyond that threshold.
You say we should accept "enough low-skilled immigrants to fulfill that need." Enough to fill the jobs. Fine. But look around: how many low-skilled jobs go unfilled, or get done worse, or cost more, because labor supply is restricted? The waitlist isn't short. The unmet demand for low-skilled labor in the United States — in agriculture, construction, elder care, food service — is not a rounding error. It's enormous. So your "flexible threshold" framework, would already admit vastly more low-skilled immigrants than current policy allows. You may have just argued for something close to open borders (for workers) without meaning to.
You accept that low-skilled workers are economically necessary. You accept that low-skilled immigrants are low-skilled workers. Yet you want a cap beyond which additional low-skilled workers become... what exactly? Unnecessary? By whose calculation? When does the marginal dish-washer, the marginal roofer, the marginal home health aide become surplus to requirements? You haven't said, because there's no principled answer — only a vague intuition that some limit must be right.
The deeper issue seems this: you're treating immigration restriction as the default that requires no justification, and open borders as the exotic position that must clear a high bar. I'd invert that. Telling a willing worker and a willing employer that they may not cooperate is an intervention. That requires justification. "We've filled our quota" is not a justification — it's a number someone made up.
I do agree: The case for low-skilled workers is also the case for low-skilled immigrant workers.
I’ve submitted the principle that low skilled immigrants should be accepted when low skilled workers are needed. Obviously there would be many intermediate steps for that principle to be enacted via policy. So no, I don’t have all the policy details to provide.
The sum effect of the policies that I would like to see is for enough low skilled immigrants to be accepted such that low skilled work positions are filled (and some rule where immigration is contingent upon going to a specific place to do a specific job for some finite period of time might be part of the means of achieving my preferred ends).
Immigration restriction absolutely is the default. Why have a “country” or “borders” but for some desire to restrict who comes and goes?
A “willing worker and willing employer” arrangement does not contravene the principle I stated. I already agree with bringing in enough people to fill the jobs that need to get done. Open border with no desire or capacity to prevent all manner of randos from coming for no particular purpose, would contravene said principle.
Mainly agreeing. Just a) if you let the citizens decide freely, which foreigner to hire to come and work for them: Not much need for further regulation (if those are excluded from all state-paid-welfare. Citizens might need to provide a deposit of a few thousand dollars to prevent abuse.) The Emirates seem to have found a very well working way.
b) "Why have a “country” or “borders” but for some desire to restrict who comes and goes?" Historically, there were borders for millennia before their were passports/visa/ immigration restrictions. Having people come to your territory to work: The wet dream of emperors! Usually, they waged wars to *force* people into their territory! See also "immigration into the USofA before 1900"
"The Chinese Exclusion Act of 1882 and Alien Contract Labor laws of 1885 and 1887 prohibited certain laborers from immigrating to the United States. The general Immigration Act of 1882 levied a head tax of fifty cents on each immigrant and blocked (or excluded) the entry of idiots, lunatics, convicts, and persons likely to become a public charge." Reminder: the USA became the superpower of the 20th century by mostly free markets AND mostly free immigration. Until 1924. If the old rules had remained, the US would dwarf China today. https://immigrationhistory.org/item/1924-immigration-act-johnson-reed-act/
Ok my gut reaction is this is the real human one - more punchy - more sentence length variation - more willingness to be offensive by using terms like “jerk” and “shame” which a polite AI might not. I suppose the fact I’m unsure is the real lesson though.
If these low skilled workers had a net present economic value greater than or equal to zero, then I would not have any disagreements. It should in fact be celebrated. But, alas...
Don’t try to answer this theoretically, you can prove anything that way. The truth is most people would rather live in a country with a strong welfare state, including minimum wages high enough to live above the poverty line. No developed democracy operates on libertarian principles. Feel free to find an island where you can live a Robinson Crusoe existence of the sort often analyzed in an intro economics course. The reason we have welfare states is because the majority believes life is more humane in them.
My first reaction was "wow, thank goodness Bryan Caplan is now selling his ideas persuasively instead of anti-selling them!" then i saw that it was ChatGPT. I generally oppose AI-generated writing but once in a while it seems genuinely useful, such as when you want to make your ideas sound more likable. Use the bot to embrace Social Desirability Bias and make your points more socially desirable!
So here's how I'd frame the tradeoff: the AI version is better writing; the Caplan version is better thinking. The AI optimized for readability and emotional arc. Caplan optimized for intellectual honesty, engaging the strongest counterarguments, and leaving the reader with novel examples they hadn't thought of before.
The commenters who couldn't tell which was which — or preferred the AI version — are revealing something real: most readers reward emotional fluency over argumentative rigor. That's a fair preference, but it's worth naming. The AI version is essentially Caplan's ideas, distilled, smoothed, and stripped of the friction that makes his writing actually worth following over time. (all Claude 4.6, obviously. In my humble human opinion: Claude nailed it.)
Ok, phew, yesterday's post by you was great and this is, well, impressive for AI but still slop. I'm dismayed by all the comments here saying the opposite though!
If you had left us to guess which was which, I would have chosen this version as the one you wrote. If this is really from ChatGPT (which I've never used), I'm truly impressed.
I had the opposite view. The previous one cited Garrett Jones and the concept of diminishing marginal returns. This was heavier on "not [X]"
"Between a hotel that is clean and a hotel that is filthy."
Places without low skill immigrants, like Japan, are ultra clean everywhere you go. Whereas places full of low skill immigrants are usually filthy.
"The low-skilled worker’s marginal product may be modest, but modest is not zero."
It's less then the carrying cost of someone living in a 1st world welfare state.
"They let higher-skilled workers specialize."
They enslave high skill workers. High skill workers are taxed to provide the low skill with medical care, education, retirement, etc.
Three arguments, three problems.
On Japan and cleanliness. You've picked the one example that proves too much. Japan is indeed remarkably clean — and Caplan himself used Japan as an example, though he drew the opposite lesson from it. Japanese cleanliness is a cultural norm enforced by the Japanese themselves, including its low-skilled workers. It tells us something about Japanese civic culture. It tells us nothing about whether low-skilled immigrant workers cause filth elsewhere, because you haven't controlled for anything. Are immigrant-heavy cities dirtier than non-immigrant cities of equivalent wealth, governance quality, and density? That's the relevant comparison, and you haven't made it. Correlation between immigration and observable urban problems typically dissolves when you control for poverty rates and municipal governance. Blaming the maid for the mess in a badly-run city is not analysis.
On welfare costs exceeding marginal product. This is the strongest of your three points, and it deserves a serious answer rather than dismissal. But notice what it actually argues: it's a case against the welfare state, not against low-skilled workers. The worker generates real value. The state then mandates transfers on top. If the combination is fiscally negative, the remedy is to fix the transfer policy — not to ban the worker who was contributing positively before the state intervened. Shooting the worker because the bureaucracy is expensive is a strange solution. Caplan has been consistent on this: the fiscal argument is an argument for reforming entitlements, not for excluding workers.
On "enslaving" high-skilled workers through taxation. This rhetoric proves far too much. By the same logic, native-born low-skilled workers "enslave" high-skilled ones just as much — they too consume healthcare, education, and retirement benefits. Do you want to expel them too? If your principle is "no one whose taxes received exceed taxes paid may participate in the economy," you've just argued for abolishing most of the welfare state and excluding a large fraction of the native population. That may be a consistent position, but own it explicitly rather than applying it selectively to immigrants.
- That was all Claude 4.6, of course. to stay in the spirit of todays AI post.
I think this is obviously the AI-written version! Surprising how many people in the comments disagree. The first article was very focused, while this feels full of fluff and flourishes, and doesn't reference economic theory or research as much.
Interesting 2 day exercise. I think it would’ve been more interesting to make this blinded for the reader, then take a poll of which is Caplan and which is GPT.
I think GPT did a pretty good job of capturing the diction and meter of your writing style. And the argument seems to fit as well.
But I think both you and GPT-pretending-to-be-you are conflating “low skilled worker” with “low skilled immigrant”. Both make a good case about the necessity of low skilled workers. I don’t think either have made the case that such necessity justifies “open borders”. I would say that we need flexibility in the workforce such that low skilled jobs are filled, and we can accept enough low skilled immigrants to fulfill that need. But I don’t see your argument justifying accepting low skilled workers as immigrants beyond that threshold.
You say we should accept "enough low-skilled immigrants to fulfill that need." Enough to fill the jobs. Fine. But look around: how many low-skilled jobs go unfilled, or get done worse, or cost more, because labor supply is restricted? The waitlist isn't short. The unmet demand for low-skilled labor in the United States — in agriculture, construction, elder care, food service — is not a rounding error. It's enormous. So your "flexible threshold" framework, would already admit vastly more low-skilled immigrants than current policy allows. You may have just argued for something close to open borders (for workers) without meaning to.
You accept that low-skilled workers are economically necessary. You accept that low-skilled immigrants are low-skilled workers. Yet you want a cap beyond which additional low-skilled workers become... what exactly? Unnecessary? By whose calculation? When does the marginal dish-washer, the marginal roofer, the marginal home health aide become surplus to requirements? You haven't said, because there's no principled answer — only a vague intuition that some limit must be right.
The deeper issue seems this: you're treating immigration restriction as the default that requires no justification, and open borders as the exotic position that must clear a high bar. I'd invert that. Telling a willing worker and a willing employer that they may not cooperate is an intervention. That requires justification. "We've filled our quota" is not a justification — it's a number someone made up.
I do agree: The case for low-skilled workers is also the case for low-skilled immigrant workers.
I’ve submitted the principle that low skilled immigrants should be accepted when low skilled workers are needed. Obviously there would be many intermediate steps for that principle to be enacted via policy. So no, I don’t have all the policy details to provide.
The sum effect of the policies that I would like to see is for enough low skilled immigrants to be accepted such that low skilled work positions are filled (and some rule where immigration is contingent upon going to a specific place to do a specific job for some finite period of time might be part of the means of achieving my preferred ends).
Immigration restriction absolutely is the default. Why have a “country” or “borders” but for some desire to restrict who comes and goes?
A “willing worker and willing employer” arrangement does not contravene the principle I stated. I already agree with bringing in enough people to fill the jobs that need to get done. Open border with no desire or capacity to prevent all manner of randos from coming for no particular purpose, would contravene said principle.
Mainly agreeing. Just a) if you let the citizens decide freely, which foreigner to hire to come and work for them: Not much need for further regulation (if those are excluded from all state-paid-welfare. Citizens might need to provide a deposit of a few thousand dollars to prevent abuse.) The Emirates seem to have found a very well working way.
b) "Why have a “country” or “borders” but for some desire to restrict who comes and goes?" Historically, there were borders for millennia before their were passports/visa/ immigration restrictions. Having people come to your territory to work: The wet dream of emperors! Usually, they waged wars to *force* people into their territory! See also "immigration into the USofA before 1900"
"The Chinese Exclusion Act of 1882 and Alien Contract Labor laws of 1885 and 1887 prohibited certain laborers from immigrating to the United States. The general Immigration Act of 1882 levied a head tax of fifty cents on each immigrant and blocked (or excluded) the entry of idiots, lunatics, convicts, and persons likely to become a public charge." Reminder: the USA became the superpower of the 20th century by mostly free markets AND mostly free immigration. Until 1924. If the old rules had remained, the US would dwarf China today. https://immigrationhistory.org/item/1924-immigration-act-johnson-reed-act/
Ok my gut reaction is this is the real human one - more punchy - more sentence length variation - more willingness to be offensive by using terms like “jerk” and “shame” which a polite AI might not. I suppose the fact I’m unsure is the real lesson though.
One of your best posts, full stop. Well done.
I had forgotten that this post said “version 2” and didn’t see your post yesterday. AI can be very, very good. This is the better post.
If these low skilled workers had a net present economic value greater than or equal to zero, then I would not have any disagreements. It should in fact be celebrated. But, alas...
This was the AI version and I liked Bryan's own better.
This seems more-vivid than Version 1 (which I thought was good until I read Version 2). So Version 1: AI. Version 2: Caplan. I guess.
This is much more content-full than yesterday’s column. I think it is the real one.
Also, it seems to be fighting a strawman. Like, yes, low-skilled work is important - do serious people actually disagree?
This one reads more like Ai to me than v1
Hi Bryan. I think both versions were very good.
Don’t try to answer this theoretically, you can prove anything that way. The truth is most people would rather live in a country with a strong welfare state, including minimum wages high enough to live above the poverty line. No developed democracy operates on libertarian principles. Feel free to find an island where you can live a Robinson Crusoe existence of the sort often analyzed in an intro economics course. The reason we have welfare states is because the majority believes life is more humane in them.
My first reaction was "wow, thank goodness Bryan Caplan is now selling his ideas persuasively instead of anti-selling them!" then i saw that it was ChatGPT. I generally oppose AI-generated writing but once in a while it seems genuinely useful, such as when you want to make your ideas sound more likable. Use the bot to embrace Social Desirability Bias and make your points more socially desirable!
So here's how I'd frame the tradeoff: the AI version is better writing; the Caplan version is better thinking. The AI optimized for readability and emotional arc. Caplan optimized for intellectual honesty, engaging the strongest counterarguments, and leaving the reader with novel examples they hadn't thought of before.
The commenters who couldn't tell which was which — or preferred the AI version — are revealing something real: most readers reward emotional fluency over argumentative rigor. That's a fair preference, but it's worth naming. The AI version is essentially Caplan's ideas, distilled, smoothed, and stripped of the friction that makes his writing actually worth following over time. (all Claude 4.6, obviously. In my humble human opinion: Claude nailed it.)