The analogy is between AI and the sun. Just as the sun provides free labor and puts candle-makers out of a job, AI will soon provide free labor and put humans working in industry X out of a job. Bastiat's argument is that there will be a net benefit to humanity from such free labor effects.
A ton of jobs have become obsolete. The candlemaker did lose his job to technology. It’s not obvious that other sectors will be there to offer them a new one, so the fear is justified.
I believe that candles weren't cheap. I honestly wonder if we have fewer candlemakers as a percent of the population now as we did, if not in 1845, then in 1345?
I agree, and it’s not even easy to switch careers. Retraining for some other work means learning a lot of new knowledge and skills, with a body and brain that are not nearly as good as they were at 20. The worker also takes a huge loss in income and status by starting again from the bottom. All this assumes that they even succeed at the change.
More generally, the problem is how the profits from the saved labour are shared around society. That distribution need not necessarily be distributed in return for labour.
> the problem is how the profits from the saved labour are shared around society
Not really. You can take an Econ 101 class to see that high profits will lead to more competition which leads to lowering of prices... which kills the excess profits and results in higher consumer surplus.
There are scenarios of market failures, where you might have a monopolistic entity but we have safety checks for that too with anti-trust policy.
> we have safety checks for that too with anti-trust policy.
Yeah I remember when Microsoft came perilously close to dominating the desktop OS market and using that dominance to push their other products, like their suite of Office software. Fortunately the safety checks of anti-trust policy kicked in and they were forced to break up the company to prevent such abuses from happening in the future.
The same thing routinely happens when ISP, TV, and media companies merge:
Ok, perhaps I was not precise in my terminology. The problem is how that consumer surplus will be distributed. And whether those who lose their jobs are able to find an alternative source of income. If consumer goods are cheaper, but people are still unable to afford them because they no longer have an income (or a greatly reduced one) then we still have a problem.
Obviously it’s in his interests to write that. There’s likely some trivial sense where it’s true but there are always jobs being destroyed in some parts and created in others.
Ignoring that sense, I think it’s a reasonable opinion to hold today, even if it turns out to be wrong.
Fun fact: there are more bank tellers in America now than when ATMs were invented.
Yes. At least, that’s ostensibly how things were right-up until the mid-2000s when online banking (and then covid) upended retail banking - but the general gist of the story is:
The introduction of ATMs meant that a branch could be operated by far fewer people than before, which reduces their operating costs - but more importantly: the investment or business risk involved in opening a new branch in the first place.
Where previously a branch could only be opened in a city with a population big enough to supply, say, 10 branch employees, after the introduction of ATMs it becomes possible to open far more branches in more cities and locations as only, say, 3 people, would be needed to run it.
So previously there would be a smaller number of larger branches, now there’s a larger number of smaller branches, but more people employee overall.
To the extent AI fills in for a deficit of labor then the comparison to Bastiat is apt (for this is how candles and lantern oils were always used, for when the 'labor' of the Sun went away). To the extent AI replaces available labor, then I'm not so sure. It may effectively be what Bastiat was sarcastically calling for here: "First, if you shut off as much as possible all access to natural light, and thereby create a need for artificial light, what industry in France will not ultimately be encouraged?"
The parallel does not align. Ludditism and import restriction come from differing motivations, even when they call for the same actions.
If AI is cheaper than human labor, AI will be preferred by industry. When human labor is cheaper than AI (such as free help from a friend or neighbor), then human labor will be preferred. But sunlight is almost always preferred to candlelight or lamp light.
Why are the concerns always 'will this fire people?' instead of 'how much of a multiplier will this be for a person?' Should maintaining the status quo of work be our first concern?
We are concerned with it precisely because when the concern is "how much more productive one person can be", the wealth created by that production is invariably captured by the global elite and used to inflict misery and ruin on everyone else. Rise up and be manipulated and crucified by people like Stalin. Stay your hand and have it chewed off by people like Musk, or have toxic chemicals dumped in your back yard like in East Palestine. It's been that way since the industrial revolution.
Given how they mock monopoly, If the tallow makers of France saw what we had done to capitalism they might fear our barons of industry would charge for sunlight or blot it out.
Universal human good appears on no financial report nor ballot sheet or any other objective of the mass of human work on the face of the earth (except perhaps a byline in a shrinking church or fraternal order).
If new suns are in store, how exactly are we to organize its distribution and protect its fertility? What commons has not been spoiled? What resource has not been extracted and hoarded?
The black gold of AI will be pumped from ground to mingle with the water of art and truth and community and education and as sure as such extraction has polluted the land it will sully our minds and squeeze our wallets.
Because the losses are concentrated and the gains are dispersed.
The total economy will be drastically better off with AI and robots to do more work for us, but a small overall fraction will see the greatest losses in employment and employability.
It's the illegal immigration debate all over again. Economically, they are a boon to the US economy over the long term. If you are low skilled or in other industries where you are competing with those migrants, it sucks for you. For everyone else, it is great. (Speaking purely from the standpoint of bringing in workers in their prime working years, ignoring cultural or other issues that arise with concentrations of out groups)
Regarding how the overall economy will fare, observing the current trends I don’t see how AI won’t just futher enshittify more.
I watch TV with obviously auto-generated subtitles. Where a human cares enough to backspace and correct typos, the broadcaster is perfectly content paying cheap rate for 80% okay subtitles, which can’t even type “oh goodness that’s TMI” without mangling it.
My walmart shopping is self-checkout, and you’re in for a treat if the fruit you grabbed is missing a barcode so you stand around waiting on an attendant. But others would just “steal” it out of impatience, causing some corpo MBA to complain about “high shrinkage”.
Disney has setup a fake screen behind the actors to display the background for shooting, how long until an AI generates fake cityscapes for movies, and people stop getting a sense for what the real world looks like? But hey, paying workers for foreign set locations and B roll is pricey!
There are also immediately concentrated gains of stuff like this into the hands of those who own it vs the people they currently have to pay for labor.
A venture capitalist, of course, is highly aware of this and being disingenuous to not mention it.
If AI pans out, it will be a huge multiplier of human progress. It may also deprive huge numbers of people of their livelihoods and dignity.
It's worth pointing out that the (ideal) inexpensive AI tech would remove basically every remaining advantage that human beings have over machines, at least if we continue to live in a society where human beings must work to survive and find meaning. AI is fundamentally different from the steam engine or cars or machine tools, in the sense that those inventions still left room for human operators. To the extent that AI leaves room for humans, it may only be due to the fact that "ideal" AI isn't achievable and what we build is (perhaps temporarily) some approximation of the more powerful thing people are worried about.
> It may also deprive huge numbers of people of their livelihoods and dignity.
Though historically the John Henry story has been bull (machines didn't replace but augmented human work), how great if it became true in this case.
There are other sources of dignity beyond compensated work. And as the marginal cost of production approaches zero, why not just pay people — money may lose its “store of value” property but retain its signaling (prioritization) function.
How much better to go spend time hanging out with your aging parent than be changing their nappies. How much better to be writing shitty poetry for your enjoyment and that of your friends.
I'd like to see an AI snake out a plugged toilet. Or do literally anything else from the Dirty Jobs TV show. We're making some progress on the software side, but hardware lags far behind.
It should be noted that this is not an inevitable outcome, but merely the one that is caused by our existing socioeconomic arrangements. We have a choice whether it will deprive huge number of people of their livelihoods and dignity or not. If we go with the existing model, AI - like any capital - will end up concentrated in very few hands. Except, crucially, this is the kind of capital that - at least in the long term, if not quite just yet - does not require labor to produce value, which is normally the limiting factor on how inhumane capitalism can get.
Ultimately because of the Protestant Work Ethic, the belief that labour is inherently good and so if people aren't working this is a moral failing.
You will see this in government programmes that assert the goal is that people have jobs, any assistance on offer is temporary, until they get jobs, and the programmes are very, very reluctant to accept that some people aren't able to work, let alone that many people don't want to work.
The PWE suits many of the wealthy and powerful, who would prefer that the poor must do what they say or starve - and so it isn't questioned. But as the actual value of more human labour falls to zero, this whole idea doesn't work. Countries should have been preparing for this for decades, but like climate change it'll start to have horrible effects before they actually take serious steps to tackle it, and like climate change it'll hurt the poorest first and most of all.
There's a lot of money to be made, shovelling back the tide.
There is no evidence that the value of human labor will fall to zero. This is not an eventuality that countries should waste effort preparing for.
Some few occupations will become obsolete, just as has happened occasionally since the dawn of civilization. Countries should provide unemployment insurance and job training for the impacted workers.
I am happy to pay taxes and contribute donations to support people who are unable to work or pursuing education. As for people who are able to work but don't want to work, fuck 'em. I have zero interest in subsidizing them and will vote accordingly.
The Luddites did not want to ban progress - they wanted to be compensated with a portion of the increased revenues from automation. They weren't crazy, the world is.
Since this is presented with minimal commentary, if you read French, you might prefer to read the original version. It loses a little in translation, IMO.
During the 2020 election cycle, the obvious inevitability of massive AI-driven unemployment was being presented as a reason that everyone with a brain had to support UBI.
Everyone is interpreting this to be about ChatGPT, so one comment:
As automation starts to do more and more of the work of today’s laptop and professional class, we’re gonna hear a lot of fancy whining about it. These people can’t write as fast as ChatGPT, but they sure can write.
50 comments
[ 346 ms ] story [ 1215 ms ] threadIn his other recent newsletter post, Marc Andreessen actually claims that AI will not cause unemployment [0].
[0]: https://pmarca.substack.com/p/why-ai-wont-cause-unemployment
Not really. You can take an Econ 101 class to see that high profits will lead to more competition which leads to lowering of prices... which kills the excess profits and results in higher consumer surplus.
There are scenarios of market failures, where you might have a monopolistic entity but we have safety checks for that too with anti-trust policy.
Yeah I remember when Microsoft came perilously close to dominating the desktop OS market and using that dominance to push their other products, like their suite of Office software. Fortunately the safety checks of anti-trust policy kicked in and they were forced to break up the company to prevent such abuses from happening in the future.
The same thing routinely happens when ISP, TV, and media companies merge:
https://en.wikipedia.org/wiki/Zero-rating#Existing_programs
https://fortune.com/longform/media-company-ownership-consoli...
Ignoring that sense, I think it’s a reasonable opinion to hold today, even if it turns out to be wrong.
Fun fact: there are more bank tellers in America now than when ATMs were invented.
Proportionate to the population?
The introduction of ATMs meant that a branch could be operated by far fewer people than before, which reduces their operating costs - but more importantly: the investment or business risk involved in opening a new branch in the first place.
Where previously a branch could only be opened in a city with a population big enough to supply, say, 10 branch employees, after the introduction of ATMs it becomes possible to open far more branches in more cities and locations as only, say, 3 people, would be needed to run it.
So previously there would be a smaller number of larger branches, now there’s a larger number of smaller branches, but more people employee overall.
The parallel does not align. Ludditism and import restriction come from differing motivations, even when they call for the same actions.
If AI is cheaper than human labor, AI will be preferred by industry. When human labor is cheaper than AI (such as free help from a friend or neighbor), then human labor will be preferred. But sunlight is almost always preferred to candlelight or lamp light.
Universal human good appears on no financial report nor ballot sheet or any other objective of the mass of human work on the face of the earth (except perhaps a byline in a shrinking church or fraternal order).
If new suns are in store, how exactly are we to organize its distribution and protect its fertility? What commons has not been spoiled? What resource has not been extracted and hoarded?
The black gold of AI will be pumped from ground to mingle with the water of art and truth and community and education and as sure as such extraction has polluted the land it will sully our minds and squeeze our wallets.
The total economy will be drastically better off with AI and robots to do more work for us, but a small overall fraction will see the greatest losses in employment and employability.
It's the illegal immigration debate all over again. Economically, they are a boon to the US economy over the long term. If you are low skilled or in other industries where you are competing with those migrants, it sucks for you. For everyone else, it is great. (Speaking purely from the standpoint of bringing in workers in their prime working years, ignoring cultural or other issues that arise with concentrations of out groups)
I watch TV with obviously auto-generated subtitles. Where a human cares enough to backspace and correct typos, the broadcaster is perfectly content paying cheap rate for 80% okay subtitles, which can’t even type “oh goodness that’s TMI” without mangling it.
My walmart shopping is self-checkout, and you’re in for a treat if the fruit you grabbed is missing a barcode so you stand around waiting on an attendant. But others would just “steal” it out of impatience, causing some corpo MBA to complain about “high shrinkage”.
Disney has setup a fake screen behind the actors to display the background for shooting, how long until an AI generates fake cityscapes for movies, and people stop getting a sense for what the real world looks like? But hey, paying workers for foreign set locations and B roll is pricey!
‘Look up item’.
You don’t need an attendant for that.
Sorry to interrupt your rant, ranting is important.
Hacker news is not the place for hot take-downs and ignorant dismissals. There are other places for that.
No shade meant, and clearly no benefit of the doubt given to someone trying to just pass along a helpful hint either.
A growing illicit workforce is in no-one’s long term interest.
https://www.reuters.com/investigates/special-report/usa-immi...
A venture capitalist, of course, is highly aware of this and being disingenuous to not mention it.
It's worth pointing out that the (ideal) inexpensive AI tech would remove basically every remaining advantage that human beings have over machines, at least if we continue to live in a society where human beings must work to survive and find meaning. AI is fundamentally different from the steam engine or cars or machine tools, in the sense that those inventions still left room for human operators. To the extent that AI leaves room for humans, it may only be due to the fact that "ideal" AI isn't achievable and what we build is (perhaps temporarily) some approximation of the more powerful thing people are worried about.
Though historically the John Henry story has been bull (machines didn't replace but augmented human work), how great if it became true in this case.
There are other sources of dignity beyond compensated work. And as the marginal cost of production approaches zero, why not just pay people — money may lose its “store of value” property but retain its signaling (prioritization) function.
How much better to go spend time hanging out with your aging parent than be changing their nappies. How much better to be writing shitty poetry for your enjoyment and that of your friends.
You will see this in government programmes that assert the goal is that people have jobs, any assistance on offer is temporary, until they get jobs, and the programmes are very, very reluctant to accept that some people aren't able to work, let alone that many people don't want to work.
The PWE suits many of the wealthy and powerful, who would prefer that the poor must do what they say or starve - and so it isn't questioned. But as the actual value of more human labour falls to zero, this whole idea doesn't work. Countries should have been preparing for this for decades, but like climate change it'll start to have horrible effects before they actually take serious steps to tackle it, and like climate change it'll hurt the poorest first and most of all.
There's a lot of money to be made, shovelling back the tide.
Some few occupations will become obsolete, just as has happened occasionally since the dawn of civilization. Countries should provide unemployment insurance and job training for the impacted workers.
I am happy to pay taxes and contribute donations to support people who are unable to work or pursuing education. As for people who are able to work but don't want to work, fuck 'em. I have zero interest in subsidizing them and will vote accordingly.
The sun is owned and operated by corporations for their own benefit?
http://bastiat.org/fr/petition.html
Nevertheless, it’s a great read in its own context.
(AI risk is an entirely separate issue)
As automation starts to do more and more of the work of today’s laptop and professional class, we’re gonna hear a lot of fancy whining about it. These people can’t write as fast as ChatGPT, but they sure can write.
Plato's allegory of the cave, but the cave wall is Stable Diffusion.
Moby Dick, but the whale is AGI.