"The less intelligent people have disproportionately been employed in positions with few important decisions to make. They have been laborers, factory workers, workmen, farmers, homemakers. "
You mean that homemakers actually have important decisions to make? Cynically speaking, they can only wreck their own families. Many civil servants today have the power to wreck so much more.
Incidental to your point, but I think the reason standards have gone down in nursing (at least in the US) is that women are now allowed to be doctors.
Women make perfectly good medical professionals, and nursing used to be the highest status medical field generally open to them, so tons of competent, intelligent women went into nursing.
Now, many of those women opt to go into medicine instead. Of course some smart, competent people still go into nursing, but the draw of higher prestige and pay is hard to counteract, so the overall nursing pool suffers as a result.
Good point, I agree this is another reasonable cause of the observed effect. Both can occur at the same time. Trends in each, over time, could be shifting (one cause becoming more significant than the other).
A hundred years ago, yes. And fifty years ago I believe many women became nurses in order to have a flexible job that could complete their husbands' much higher-paying positions. The nurse-wife could follow her husband wherever in the country his job took him and jump back and forth between the care of young children and work. Nowadays, when women are supposed to support themselves, nursing seems like a less attractive option.
This article connects with a lot of interesting issues.
One is that education standards have been increasing. The American production system (e.g. what Henry Ford devised) treated workers as machines and required them to make very few decisions. This worked well for employing illiterate (and sometimes non-English speaking) workers, but decisions about the details of production were made by people who didn't actually do the work. The Japanese system has the advantage that many of those detailed decisions are made by people who understand the work in a very detailed way ... but also, it uses a workforce that is much better educated and (due to the Flynn effect) more intelligent. But there's a tradeoff between moving decisions to more-specialized, more-intelligent people vs. moving them to people who know the action involved more intimately. This relates to the problem that the decisions that go awry when less-intelligent people do them are ones that have a broad scope and ordinary work doesn't give anyone the needed detailed knowledge.
> But because the most intelligent people are hired by the finance sector and IT sector instead, vacating positions in the public sector for the kind of people who would have been factory workers 30 years ago.
Although in Singapore, public sector jobs are the best paying, so they have very competent public servants.
>>This relates to the problem that the decisions that go awry when less-intelligent people do them are ones that have a broad scope and ordinary work doesn't give anyone the needed detailed knowledge.
Exactly! It is good when people can influence their own work, because regardless of how intelligent they are, they are likely to know many things about their own jobs. The bad things happen when less intelligent (or, for that matter, less interested, or less sensible) people make decisions about things they know little about.
On dumb people being "accountable" - it's not accountability if there's basically zero consequences for them messing up. Was anyone fired for that Swedish traffic imbroglio? I'd bet against it.
How often do you ever hear about government bureaucrats being fired for incompetence, or because they were "accountable" and messed up? Essentially never? Yeah, me too.
The "accountability" you're pointing at is entirely false. Arguably, AI will be a step *up,* because people can get mad and replace it with a different / better AI, whereas that never happens for *people.*
You have an important point. "Responsibility" might be a better word than accountability.
Because as you guess, no one was fired. Just like you, I very seldom hear about government bureaucrats being fired for incompetence. So I think you are right that I'm using the wrong word.
In general, the government bureaucrats I have met have believed that they were accountable. When they say that they need to follow the rules because else they will be in trouble, they seem to genuinely believe it. But as you point out, that is mostly an instrumental illusion. An illusion of accountability is not the same thing as accountability.
I'd suggest bureaucrats are held accountable for following rules. They're not generally held accountable for producing results - partly because it's much easier to assign individual fault for the former rather than the latter - but the net result is you get an organization that strictly adheres to rules, which stifles both the application of reasoning and the ability to respond to situations that don't fit those the rules were designed to deal with (basically, anything with complexity and variability rather than repeatability).
I was going to comment on this too. It seems to me that people in Sweden nowadays often deflect accountability by saying "our procedures were correctly followed" instead of looking at what the outcome was. And it seems very difficult to fire someone for doing a poor job, not only in the public sector. I saw an article on the front page of dn.se two days ago about Arbetsförmedlingen wanting to fire one of its employees who had avoided work in very obvoius ways from time to time over several years. My instant reaction was "why is this front page news and why were they not fired years ago?". On the other hand the public sector seems fond of "buying off" inconvenient employees (also those doing a good job) by giving them two years of salary if they resign themselves.
I started working with AI (we called it "machine learning" which I think is a more suitable name) for image analysis twenty years ago and already back then it was possible to develop systems that made better decisions than the average doctor in some specific areas. But it was very difficult for commercial companies to sell such systems, because a human would still need to be accountable for any mistakes. People want to take decisions, not responsibility. I suppose it's also easier for a hospital to say "our super good doctors did not find the problem" than to say "our AI was not good enough". There is an expectation that an AI should be perfect. And if the AI is not perfect then it should not be trusted at all. Which I think is unfortunate, because sometimes experienced doctors just make wild guesses instead of saying to the patient "unfortunately this is outside of my medical competence". It seems to me that most doctors are intelligent, but the more experience they have the less they use their intelligence. People who are never challenged eventually start to think that they know best about everything, even things outside of their expertise (which perhaps is what I'm doing right now, I do see the irony...).
Most people who worked with AI were also interested in the stock market, myself included. I have never been employed in the finance sector, but more than once made a single trade in a single day that earned me as much money as working two months in a high salary job. So the number of people being employed in the finance sector might not tell the whole story. I think intelligent people can more easily avoid the need for employment today than they could 100 years ago. I suppose in that way people with capital and an interest in finance resemble aristocracy from 200 years ago. And the capital no longer needs to be inherited, just a few years savings from a regular job can be enough to start with.
I think whether intelligent people can more easily avoid employment today than 100 years ago is not necessarily correct; yes, to the extent it depends on improved class mobility; otherwise, it depends to a large degree on your definition of employment.
I haven't been that lucky, and while I see people who seem to have a high standard of living, I assume that most of they are doing a decent amount of work to get there!
As for a hundred years ago, I picture there were landed gentry, and certain professions that gave a decent standard of living without having a nose to the grindstone, rat race type feel. Like maybe being a lawyer or doctor back then was the equivalent of 300k USD today but without commutes or long hours.
Happy to be educated on this is someone has facts and figures.
I guess there are people making that kind of money in software or finance, as well as the seemingly always lucrative medical and legal professions (those with high stakes and/or high barriers to entry), but I don't picture it's easy to get there and easy to stay there. But I suppose it is easier if you have the right combinations of skills (mostly communications and motivation) and analytical intelligence - but to avoid the need for employment? I can't imagine there's many people who can do that!
I should have clarified what I meant by "high standard of living". I simply meant having a home you feel good about, having access to healthcare if you need it, a car if you want to, being able to buy any food you want. Not having to care about the price of a TV, mobile phone etc. I realize now the phrase "high standard of living" might also be interpreted as "being considered rich", but that's not what I had in mind. And of course having a high standard of living is more expensive in NYC or SF than in rural areas.
According to the UBS Global Wealth Report 5% of the population in Sweden (10% of households if no overlap) had a net worth of one million USD or more in 2022 (down from 6% in 2021). An annual yield of 5% after taxes and inflation would mean at least 50k USD per year, which in my opinion is enough for having a high standard of living in rural areas in Sweden without needing to do any serious work. And someone who is good at playing the stock market can have a high standard of living with a lot less than one million USD in net worth.
Unfortunately I don't know what percentage of the Swedish population was in a similar situation 100 years ago, but I imagine it was fewer than 5% and that most of them had inherited their money and/or land.
You posit " ai can't be held accountable" and thus have to employ more people in controller role.
What is " accountability"? Ai in many ways is way more accountable than people because information systems are built with logs and data measuring systems.
But there is another part ." accountability" is performed by someone who is more powerful that the subject who is held accountable.
In animal world "nature" holds animals accountable. Humans quickly ( by evolutionary standards) got most of the accountability into their own hands. By dominating natural systems with their own agency.
This domination was a product of systems ( language, tools , society) created by slightly more intelligent brain than previously existed.
And now we are on brink of creating intelligence of orders of magnitude more powerful than anything ever. The systems it will create will be similarly more impactful and powerful.
In a world with AI ai will be accountable to higher order systems than human bureaucracy. Humans will control AI no more than ants or chimps control us
Intelligence in this case being machines that have the capacity to weigh inputs and outputs effectively to generalize patterns, and use these generalizations to predict a result. Amazingly well in many cases. I'm not sure that this is better than human intelligence, nor that it will always lead to good things if left unchecked.
Thank you for your thought-provoking article. James Burnham suggests in his "Managerial Revolution" that as businesses grew more complex, a separation occurred between ownership and control. Finance, effectively a managerial class, handles the universal allocation of resources—namely, credit. Consequently, they become a powerful group, despite not owning the resources they manage. The shift described in your post reflects not so much technological innovation as a result of changing power dynamics. Talented individuals gravitate towards the most lucrative positions.
The question of whether AI development will lead to an idiocracy essentially boils down to whether AI will grant more bargaining power to finance managers. After having witnessed so many talented physicists and computer science students enter the finance industry, I see no way for academia to compete with the finance industry in recruiting high-IQ individuals solely based on financial incentives.
The issue is not AI itself, nor is it a matter of socialism versus capitalism. Rather, it is the managerial class that stands to gain the most in the AI era. I would say any tech innovation that enhances a financial manager's capacity to independently manage more credit could worsen the situation.
You could argue that AI could consolidate the financial power in no more than a few hands, and that will release human intelligence to work on other problems. I think it's likely we'll just end up with more and more people having nothing meaningful to do, and filling their lives with mostly meaningless entertainment (TV, music, gaming, porn, social media ...). Perhaps sports and other physical recreation would be the best alternatives to meaningful work. Either way, we need to make sure the AI values human life and liberty above all else. But which takes precedence? I would suggest that life is nothing without liberty.
We could also argue liberty is nothing without opportunity, and that the aim of AI (it's objective function), or indeed of society, should be to maximize a person's life quality by maximizing opportunity and the freedom to choose among them. This promotes extension of life, but allows for adjudicating situations in which either harm to others or self harm (possibly including mental illness or addiction) might prevent future opportunities to exercise liberty.
"This all means that the descendants of production workers, who are no more intelligent than their ancestors, are increasingly working with decision-making. "
I'm not absolutely certain that's true, though.... I seem to remember something about how every 20-40 years, IQ tests have to be 're-normed' to 100 points, because we've gotten better at expecting more children to go further in education and to do a better job of learning how to take IQ tests.
I think the example I saw was that if you take a perfectly normal 16-yr-old out of a modern american high school, with dead-average test scores and a modern-era IQ score of exactly 100....
And then you take him to an English-Speaking country in WW1 that allowed 16-yr-olds to enlist under certain conditions, and gave him a standard WW1 Army IQ test.... He'd most likely score about a 140 or so. There are interesting debates on whether that means that kids are actually GETTING smarter, or if they're only getting better educated, or if they're only getting smarter for certain limited purposes of taking IQ tests. After all, in the British Army in WW1, it was pretty common to enlist full-grown adults who had simply stopped being educated at about the US 6th grade equivalent, age 12. School wasn't free or mandatory past that cutoff point.
Yes! I read a book by James Flynn many years ago, called What is intelligence? In that book I think Flynn somewhat ducked the question in the title. But his research is really fascinating. I think it is a reason to approach the relationship between IQ and intelligence with a open mind.
It is true that there was some Flynn effect between people born in the 1950s and the 1970s. But after that the effect has been much smaller or even reversed in Western countries.
I am reading *The Unaccountability Machine: Why Big Systems Make Terrible Decisions and How the World Lost its Mind.* by Dan Davies. If Davies is correct, then AI is going to make things much better, not worse. Seems the problem is not 'we need smarter decision makers' -- and indeed the chapter on the financial crisis seems to indicate that taking the decision making away from the people on the bottom, and giving it to people higher up in the hierarchy is a large part of problem. No amount of smartness can overcome not knowing how the job is done by the people who do the work. Accountability problems happen because by the time you find somebody who has the agency to actually do something you have also found somebody who hasn't a clue about what should be done, yet often because they are bright greatly underestimate their own ignorance.
Also, most people who are bright are only "bright", that is, they can solve problems similar to those they have encountered before, which almost by definition are not really problems. Skill transfer is vanishingly rare.
Not everything in the AI sphere is hallucinating chatbots. There is a lot of work being done quantifying domain expert knowledge about how equipment fails in the field, for instance, which isn't hallucinating anything.
I'm funding this sort of very practical research, with the goal to give the workers such tools that they can solve many more problems directly. For instance these people https://www.vikinganalytics.se/ are doing great work, and I am happy to be an early investor here. And understanding when your equipment is headed for catastrophic failure before it happens, is a great outcome if in no way glamourous and sexy.
cool. about 15 years ago I did some work for an Italian company that was trying to use relatively primitive ML to predict failures of networking equipment. It was only so so effective
It is much easier for the education system to teach reading, writing and basic math to bright students.
But of course I can't be certain of the measurable psychological properties of people someone complains of in an article in public service media. It is... not a great source. The informants complain that some students expect nursing education to be much easier than it is, which I interpret as a complaint that they are not studious enough.
It could be the students lacking ability. But it could be poor teaching.
Literacy in Japan, for example, is around 99% despite Japan having a complex system of writing that involves two syllaberies AND several thousand chinese characters. ( https://www.theguardian.com/world/2013/oct/08/why-do-japanese-children-lead-world-numeracy-literacy ). Are you telling me that Japanese children are all bright? It is true that Asians seem to have a slightly higher IQ that Europeans but it's not enough of a difference that there aren't plenty of Japanese with IQs under the European average.
I don't know how the education system is in Sweden, but in some countries (e.g. (parts of) the USA - https://thepostmillennial.com/report-baltimore-high-schoolers-due-to-graduate-perform-math-reading-at-grade-school-level ) the education system is so bad that students are effectively graduated from one year to the next solely if they turn up to school more than 50% of the time. It would be easy to imagine that a person who was "educated" in that manner would be unable to handle studying that required actual effort.
I would caution against taking the literacy statistic at face value. Many countries count students as literate if they have completed a certain number of years of education, irrespective of their actual reading abilities.
I trust Japan more than most not to do that, partly because I live there and have personal daily experience there. It is true literacy could be 97% not 99% (or that it's 95% having high school literacy and 99% having elementary school literacy) but for sure it isn't much lower.
I think nursing education might have been a confusing example, because a "nurse" means different things in Sweden and the US. In Sweden a "nurse" is a person with a three year college education. I guess that is what would be called a "registered nurse" in the US.
I would like to turn the question around: Should 99 percent of Japanese people become registered nurses? Or should teachers start complaining if they get a certain number of students from a certain percentile?
It is true that the education system could be much better. Still, it is implicitly assumed that everyone should not pursue higher education. I can't be certain in any sense that the nursing teachers behind the article thought that such a limit had been reached. It was just my impression.
My post explains that the current LLM AI systems cannot be trusted to not hallucinate a wrong answer and that this is inherent in them. You cannot trust them, it is impossible to do so. If the wrong answer is for something that involves, say hazardous materials, then that wrong answer could lead to a fire or other disaster.
A smart well educated person would be likely to spot the LLM hallucination before something bad happens. A not so smart/well educated one probably won't.
Yes. AIs hallucinate. But so do social workers. And building inspectors. And traffic planners.
When it comes to really important systems, relying on AI as we know it is foolish. The importance of human vigilance should never be underestimated. And my impression is that in systems that are known to be hazardous, there are still comparatively high IQ people overseeing them. I'm more worried about not-so-clever people being employed in not-so-important positions. So few people notice, until those positions suddenly get important (like when it started to snow).
I think the role of AI would be to create fallible, statistics-based systems. The point is not to create perfect outcomes, but to create less disastrous outcomes. I think social work is an area where such a system could improve things. Currently, on a scientific level, it is a very well-known fact that foster care is hazardous for the outcomes of children. Still, children are placed in foster care for banal reasons and because psychological problems are misinterpreted as social problem. An AI would not make perfect decisions. But it is difficult to make worse decisions than some current groups of human social workers. Only its lack of bias is a good start.
I am happily reading your substack since at least 2 years, usually getting a fresh and spirited perspective on things I care about. Thank you for that! However, your classist understanding of intelligence ("...production workers, who are no more intelligent than their ancestors") defies every fiber of my being. Let me tell you that I come from a real socialist country, one of which ceased to exist in '89. The success those countries had in recruiting the brightest minds from the literal working class can not possibly have escaped you entirely?! Especially as you live in Sweden, which from my understanding is much more egalitarian than the US for example.
Of course there are bright minds among working class people. A hundred years ago, before educational opportunities were spread over the population, there were uncountable of them. My only point is that on aggregate, the generation of boomer production workers did not have children that were more intelligent than themselves. (Two of those workers were my parents, by the way. So I'm included in the category “descendants of production workers” myself.)
I live in Sweden too, and one of the secrets of our success is that, wherever possible, you try to get the problems solved at the bottom, and not at the top. Usually management tells people what they want accomplished, but leaves it to them to figure out how to do it. The number one reason why disasters happen when US or worse British companies decide to purchase a Swedish one is the cultural clash between the business cultures. From a British perspective, Swedes aren't capable of being managed at all. If you try to manage them, and take away their autonomy in ways that are common in the UK, they will get terribly insulted, and then quit. I used to teach courses on the cultural difference back when fixing industrial problems was my full time job.
But I think that Tove's greatest experience is with the hospital/medical bureaucracy in Sweden. And it's very much not run along the same lines as things are usually done in Sweden, for reasons historical and having to do with the fact that it is the government and not the patients who are paying for the health care. Which you sort of need, unless you are willing to let the poor do without. But it means that a health care system can end up being accountable only to the people who count beans in the government accounting offices, or entirely captured by the healthcare workers, rather than accountable to the patients who often would prefer if things were done in another way.
Also: the problem with nurses who cannot read is mostly an effect of having a significant number of people who have moved to Sweden to get a nursing position from overseas, and who have only been in the country for a few months. They're very much not stupid people -- they are bright and ambitious and once they learn Swedish things go much better for them. This is a problem in how nursing education is done, and a problem in how you really need the new immigrants to start their training soon after they arrive. (Otherwise, you are stuck with the people who have no intention of doing the training but stick around forever, claiming to be learning Swedish and even going to the classes, but never reaching any acceptable level of proficiency. It's all a mess, and the tradeoffs all have severe defects, whatever you decide to do. )
This is a great point. Living in Sweden I probably take a certain degree of worker autonomy too much for granted. Taking away power from people on the floor and giving it to people higher up who know less about reality down there is a potentially disastrous idea. If nothing else my mother, who spent almost her entire working life in a factory has taught me that.
>>Also: the problem with nurses who cannot read is mostly an effect of having a significant number of people who have moved to Sweden to get a nursing position from overseas, and who have only been in the country for a few months.
There certainly are a lot of immigrants in the Swedish care sector. But the article in question was not about recent immigrants but about new students.
>>But I think that Tove's greatest experience is with the hospital/medical bureaucracy in Sweden.
I can't complain of decision-makers with moderate IQs in the health care sector, because doctors are very much selected on IQ-adjacent tests and (less, but still) IQ adjacent high school grades. I could complain about the health care system for a day or two, but not in that sense. When it comes to more and more not-so-bright people making important decisions, I'm much more worried about positions requiring engineering degrees, teaching degrees, and, especially, degrees in social work.
Ah, the "nurse-teaching-has-gone-to-hell" cohort around here are mostly complaining about the immigrants whose Swedish is lousy. And they are trying
to change the way they teach the undernurses as a result of this. (And with Sverige Demokraterna demanding a certain level of Swedish proficiency in the care takers that interact with the old, I think they are going to have to change their course of study.) I assumed you were reading the same sort of papers, but your complainers have the larger claim that Swedish childhood education in general has gone to hell?
I read something a few days ago which said that when the Anglosphere learned about the Swedish idea of work/life balance and thought that they should try it instead of aiming for all-work ... they fed these ideas back to the Swedish english-reading youth who got the idea that they were working too hard as well. But the people making the claim were members of the 'Swedish children have too long a summer vacation' crowd, which makes me discount anything they say (and want to throw my rotting vegetables at them, too :) ) But maybe something like that is going on? I'll ask my 18-23 year old relatives and see what they think.
ah, I think this misunderstands how so many of them are taught to see their role: they are there to execute the process dutifully, not to achieve a specific result. For accountability you might only be talking about the higher-level folks, or the electeds who provide their marching orders.
I am not saying I 100% love this situation. But I am saying it is a better mental model for managing your expectations of bureaucrats.
I have long believed that a major factor in the 2008 Global Financial Crisis was the desire to employ more college graduates who couldn't otherwise get the prestigious-sounding job they wanted.
"The less intelligent people have disproportionately been employed in positions with few important decisions to make. They have been laborers, factory workers, workmen, farmers, homemakers. "
Homemakers? Really?
You mean that homemakers actually have important decisions to make? Cynically speaking, they can only wreck their own families. Many civil servants today have the power to wreck so much more.
That and the implication that high IQ women would not be attracted to being homemakers.
Incidental to your point, but I think the reason standards have gone down in nursing (at least in the US) is that women are now allowed to be doctors.
Women make perfectly good medical professionals, and nursing used to be the highest status medical field generally open to them, so tons of competent, intelligent women went into nursing.
Now, many of those women opt to go into medicine instead. Of course some smart, competent people still go into nursing, but the draw of higher prestige and pay is hard to counteract, so the overall nursing pool suffers as a result.
More men are in the nursing field than ever before. That's one reason for the rise of a lack of intelligence and competency in the field.
Good point, I agree this is another reasonable cause of the observed effect. Both can occur at the same time. Trends in each, over time, could be shifting (one cause becoming more significant than the other).
A hundred years ago, yes. And fifty years ago I believe many women became nurses in order to have a flexible job that could complete their husbands' much higher-paying positions. The nurse-wife could follow her husband wherever in the country his job took him and jump back and forth between the care of young children and work. Nowadays, when women are supposed to support themselves, nursing seems like a less attractive option.
Nursing is a high stress job. Going back and forth between that and childcare at home is a nightmare. When do these women ever get to rest and relax?
When I grew up I had the impression that mothers did not rest and relax.
Maybe that's one of the reasons why many more women these days are consciously choosing not to become mothers. They value their health and wellbeing.
Absolutely. I'm sure many women today have no children because they don't want to become like their mothers.
This article connects with a lot of interesting issues.
One is that education standards have been increasing. The American production system (e.g. what Henry Ford devised) treated workers as machines and required them to make very few decisions. This worked well for employing illiterate (and sometimes non-English speaking) workers, but decisions about the details of production were made by people who didn't actually do the work. The Japanese system has the advantage that many of those detailed decisions are made by people who understand the work in a very detailed way ... but also, it uses a workforce that is much better educated and (due to the Flynn effect) more intelligent. But there's a tradeoff between moving decisions to more-specialized, more-intelligent people vs. moving them to people who know the action involved more intimately. This relates to the problem that the decisions that go awry when less-intelligent people do them are ones that have a broad scope and ordinary work doesn't give anyone the needed detailed knowledge.
> But because the most intelligent people are hired by the finance sector and IT sector instead, vacating positions in the public sector for the kind of people who would have been factory workers 30 years ago.
Although in Singapore, public sector jobs are the best paying, so they have very competent public servants.
>>This relates to the problem that the decisions that go awry when less-intelligent people do them are ones that have a broad scope and ordinary work doesn't give anyone the needed detailed knowledge.
Exactly! It is good when people can influence their own work, because regardless of how intelligent they are, they are likely to know many things about their own jobs. The bad things happen when less intelligent (or, for that matter, less interested, or less sensible) people make decisions about things they know little about.
On dumb people being "accountable" - it's not accountability if there's basically zero consequences for them messing up. Was anyone fired for that Swedish traffic imbroglio? I'd bet against it.
How often do you ever hear about government bureaucrats being fired for incompetence, or because they were "accountable" and messed up? Essentially never? Yeah, me too.
The "accountability" you're pointing at is entirely false. Arguably, AI will be a step *up,* because people can get mad and replace it with a different / better AI, whereas that never happens for *people.*
You have an important point. "Responsibility" might be a better word than accountability.
Because as you guess, no one was fired. Just like you, I very seldom hear about government bureaucrats being fired for incompetence. So I think you are right that I'm using the wrong word.
In general, the government bureaucrats I have met have believed that they were accountable. When they say that they need to follow the rules because else they will be in trouble, they seem to genuinely believe it. But as you point out, that is mostly an instrumental illusion. An illusion of accountability is not the same thing as accountability.
I need to keep this in mind and think it over.
I'd suggest bureaucrats are held accountable for following rules. They're not generally held accountable for producing results - partly because it's much easier to assign individual fault for the former rather than the latter - but the net result is you get an organization that strictly adheres to rules, which stifles both the application of reasoning and the ability to respond to situations that don't fit those the rules were designed to deal with (basically, anything with complexity and variability rather than repeatability).
I was going to comment on this too. It seems to me that people in Sweden nowadays often deflect accountability by saying "our procedures were correctly followed" instead of looking at what the outcome was. And it seems very difficult to fire someone for doing a poor job, not only in the public sector. I saw an article on the front page of dn.se two days ago about Arbetsförmedlingen wanting to fire one of its employees who had avoided work in very obvoius ways from time to time over several years. My instant reaction was "why is this front page news and why were they not fired years ago?". On the other hand the public sector seems fond of "buying off" inconvenient employees (also those doing a good job) by giving them two years of salary if they resign themselves.
I started working with AI (we called it "machine learning" which I think is a more suitable name) for image analysis twenty years ago and already back then it was possible to develop systems that made better decisions than the average doctor in some specific areas. But it was very difficult for commercial companies to sell such systems, because a human would still need to be accountable for any mistakes. People want to take decisions, not responsibility. I suppose it's also easier for a hospital to say "our super good doctors did not find the problem" than to say "our AI was not good enough". There is an expectation that an AI should be perfect. And if the AI is not perfect then it should not be trusted at all. Which I think is unfortunate, because sometimes experienced doctors just make wild guesses instead of saying to the patient "unfortunately this is outside of my medical competence". It seems to me that most doctors are intelligent, but the more experience they have the less they use their intelligence. People who are never challenged eventually start to think that they know best about everything, even things outside of their expertise (which perhaps is what I'm doing right now, I do see the irony...).
Most people who worked with AI were also interested in the stock market, myself included. I have never been employed in the finance sector, but more than once made a single trade in a single day that earned me as much money as working two months in a high salary job. So the number of people being employed in the finance sector might not tell the whole story. I think intelligent people can more easily avoid the need for employment today than they could 100 years ago. I suppose in that way people with capital and an interest in finance resemble aristocracy from 200 years ago. And the capital no longer needs to be inherited, just a few years savings from a regular job can be enough to start with.
I think whether intelligent people can more easily avoid employment today than 100 years ago is not necessarily correct; yes, to the extent it depends on improved class mobility; otherwise, it depends to a large degree on your definition of employment.
Would you agree "having a high standard of living without needing to do any serious amount of work" is easier today than 100 years ago?
I haven't been that lucky, and while I see people who seem to have a high standard of living, I assume that most of they are doing a decent amount of work to get there!
As for a hundred years ago, I picture there were landed gentry, and certain professions that gave a decent standard of living without having a nose to the grindstone, rat race type feel. Like maybe being a lawyer or doctor back then was the equivalent of 300k USD today but without commutes or long hours.
Happy to be educated on this is someone has facts and figures.
I guess there are people making that kind of money in software or finance, as well as the seemingly always lucrative medical and legal professions (those with high stakes and/or high barriers to entry), but I don't picture it's easy to get there and easy to stay there. But I suppose it is easier if you have the right combinations of skills (mostly communications and motivation) and analytical intelligence - but to avoid the need for employment? I can't imagine there's many people who can do that!
I should have clarified what I meant by "high standard of living". I simply meant having a home you feel good about, having access to healthcare if you need it, a car if you want to, being able to buy any food you want. Not having to care about the price of a TV, mobile phone etc. I realize now the phrase "high standard of living" might also be interpreted as "being considered rich", but that's not what I had in mind. And of course having a high standard of living is more expensive in NYC or SF than in rural areas.
According to the UBS Global Wealth Report 5% of the population in Sweden (10% of households if no overlap) had a net worth of one million USD or more in 2022 (down from 6% in 2021). An annual yield of 5% after taxes and inflation would mean at least 50k USD per year, which in my opinion is enough for having a high standard of living in rural areas in Sweden without needing to do any serious work. And someone who is good at playing the stock market can have a high standard of living with a lot less than one million USD in net worth.
Unfortunately I don't know what percentage of the Swedish population was in a similar situation 100 years ago, but I imagine it was fewer than 5% and that most of them had inherited their money and/or land.
You posit " ai can't be held accountable" and thus have to employ more people in controller role.
What is " accountability"? Ai in many ways is way more accountable than people because information systems are built with logs and data measuring systems.
But there is another part ." accountability" is performed by someone who is more powerful that the subject who is held accountable.
In animal world "nature" holds animals accountable. Humans quickly ( by evolutionary standards) got most of the accountability into their own hands. By dominating natural systems with their own agency.
This domination was a product of systems ( language, tools , society) created by slightly more intelligent brain than previously existed.
And now we are on brink of creating intelligence of orders of magnitude more powerful than anything ever. The systems it will create will be similarly more impactful and powerful.
In a world with AI ai will be accountable to higher order systems than human bureaucracy. Humans will control AI no more than ants or chimps control us
Intelligence in this case being machines that have the capacity to weigh inputs and outputs effectively to generalize patterns, and use these generalizations to predict a result. Amazingly well in many cases. I'm not sure that this is better than human intelligence, nor that it will always lead to good things if left unchecked.
Thank you for your thought-provoking article. James Burnham suggests in his "Managerial Revolution" that as businesses grew more complex, a separation occurred between ownership and control. Finance, effectively a managerial class, handles the universal allocation of resources—namely, credit. Consequently, they become a powerful group, despite not owning the resources they manage. The shift described in your post reflects not so much technological innovation as a result of changing power dynamics. Talented individuals gravitate towards the most lucrative positions.
The question of whether AI development will lead to an idiocracy essentially boils down to whether AI will grant more bargaining power to finance managers. After having witnessed so many talented physicists and computer science students enter the finance industry, I see no way for academia to compete with the finance industry in recruiting high-IQ individuals solely based on financial incentives.
The issue is not AI itself, nor is it a matter of socialism versus capitalism. Rather, it is the managerial class that stands to gain the most in the AI era. I would say any tech innovation that enhances a financial manager's capacity to independently manage more credit could worsen the situation.
You could argue that AI could consolidate the financial power in no more than a few hands, and that will release human intelligence to work on other problems. I think it's likely we'll just end up with more and more people having nothing meaningful to do, and filling their lives with mostly meaningless entertainment (TV, music, gaming, porn, social media ...). Perhaps sports and other physical recreation would be the best alternatives to meaningful work. Either way, we need to make sure the AI values human life and liberty above all else. But which takes precedence? I would suggest that life is nothing without liberty.
We could also argue liberty is nothing without opportunity, and that the aim of AI (it's objective function), or indeed of society, should be to maximize a person's life quality by maximizing opportunity and the freedom to choose among them. This promotes extension of life, but allows for adjudicating situations in which either harm to others or self harm (possibly including mental illness or addiction) might prevent future opportunities to exercise liberty.
"This all means that the descendants of production workers, who are no more intelligent than their ancestors, are increasingly working with decision-making. "
I'm not absolutely certain that's true, though.... I seem to remember something about how every 20-40 years, IQ tests have to be 're-normed' to 100 points, because we've gotten better at expecting more children to go further in education and to do a better job of learning how to take IQ tests.
I think the example I saw was that if you take a perfectly normal 16-yr-old out of a modern american high school, with dead-average test scores and a modern-era IQ score of exactly 100....
And then you take him to an English-Speaking country in WW1 that allowed 16-yr-olds to enlist under certain conditions, and gave him a standard WW1 Army IQ test.... He'd most likely score about a 140 or so. There are interesting debates on whether that means that kids are actually GETTING smarter, or if they're only getting better educated, or if they're only getting smarter for certain limited purposes of taking IQ tests. After all, in the British Army in WW1, it was pretty common to enlist full-grown adults who had simply stopped being educated at about the US 6th grade equivalent, age 12. School wasn't free or mandatory past that cutoff point.
Yes! I read a book by James Flynn many years ago, called What is intelligence? In that book I think Flynn somewhat ducked the question in the title. But his research is really fascinating. I think it is a reason to approach the relationship between IQ and intelligence with a open mind.
It is true that there was some Flynn effect between people born in the 1950s and the 1970s. But after that the effect has been much smaller or even reversed in Western countries.
I am reading *The Unaccountability Machine: Why Big Systems Make Terrible Decisions and How the World Lost its Mind.* by Dan Davies. If Davies is correct, then AI is going to make things much better, not worse. Seems the problem is not 'we need smarter decision makers' -- and indeed the chapter on the financial crisis seems to indicate that taking the decision making away from the people on the bottom, and giving it to people higher up in the hierarchy is a large part of problem. No amount of smartness can overcome not knowing how the job is done by the people who do the work. Accountability problems happen because by the time you find somebody who has the agency to actually do something you have also found somebody who hasn't a clue about what should be done, yet often because they are bright greatly underestimate their own ignorance.
Also, most people who are bright are only "bright", that is, they can solve problems similar to those they have encountered before, which almost by definition are not really problems. Skill transfer is vanishingly rare.
No. See my comment below(above? nearby). AI will lead you into a catastrophe with total confidence.
Not everything in the AI sphere is hallucinating chatbots. There is a lot of work being done quantifying domain expert knowledge about how equipment fails in the field, for instance, which isn't hallucinating anything.
well good. I guess I should have been more precise. Given the way LLM AI works I don't see how you can ever trust it not to hallucinate.
Other kinds of AI may work differently, AI isn;t really my field
I'm funding this sort of very practical research, with the goal to give the workers such tools that they can solve many more problems directly. For instance these people https://www.vikinganalytics.se/ are doing great work, and I am happy to be an early investor here. And understanding when your equipment is headed for catastrophic failure before it happens, is a great outcome if in no way glamourous and sexy.
cool. about 15 years ago I did some work for an Italian company that was trying to use relatively primitive ML to predict failures of networking equipment. It was only so so effective
Things are a lot better now. YAY!
BTW poor education is quite different from low intelligence. Inability to read and write (or do basic sums etc.) is the fault of the education system.
It is much easier for the education system to teach reading, writing and basic math to bright students.
But of course I can't be certain of the measurable psychological properties of people someone complains of in an article in public service media. It is... not a great source. The informants complain that some students expect nursing education to be much easier than it is, which I interpret as a complaint that they are not studious enough.
It could be the students lacking ability. But it could be poor teaching.
Literacy in Japan, for example, is around 99% despite Japan having a complex system of writing that involves two syllaberies AND several thousand chinese characters. ( https://www.theguardian.com/world/2013/oct/08/why-do-japanese-children-lead-world-numeracy-literacy ). Are you telling me that Japanese children are all bright? It is true that Asians seem to have a slightly higher IQ that Europeans but it's not enough of a difference that there aren't plenty of Japanese with IQs under the European average.
I don't know how the education system is in Sweden, but in some countries (e.g. (parts of) the USA - https://thepostmillennial.com/report-baltimore-high-schoolers-due-to-graduate-perform-math-reading-at-grade-school-level ) the education system is so bad that students are effectively graduated from one year to the next solely if they turn up to school more than 50% of the time. It would be easy to imagine that a person who was "educated" in that manner would be unable to handle studying that required actual effort.
I would caution against taking the literacy statistic at face value. Many countries count students as literate if they have completed a certain number of years of education, irrespective of their actual reading abilities.
I trust Japan more than most not to do that, partly because I live there and have personal daily experience there. It is true literacy could be 97% not 99% (or that it's 95% having high school literacy and 99% having elementary school literacy) but for sure it isn't much lower.
I think nursing education might have been a confusing example, because a "nurse" means different things in Sweden and the US. In Sweden a "nurse" is a person with a three year college education. I guess that is what would be called a "registered nurse" in the US.
I would like to turn the question around: Should 99 percent of Japanese people become registered nurses? Or should teachers start complaining if they get a certain number of students from a certain percentile?
It is true that the education system could be much better. Still, it is implicitly assumed that everyone should not pursue higher education. I can't be certain in any sense that the nursing teachers behind the article thought that such a limit had been reached. It was just my impression.
Your idea that less intelligent / less educated people oversee AI to take responsibility takes a post I made on AI to make for lots of nightmare - https://ombreolivier.substack.com/p/llm-considered-harmful
My post explains that the current LLM AI systems cannot be trusted to not hallucinate a wrong answer and that this is inherent in them. You cannot trust them, it is impossible to do so. If the wrong answer is for something that involves, say hazardous materials, then that wrong answer could lead to a fire or other disaster.
A smart well educated person would be likely to spot the LLM hallucination before something bad happens. A not so smart/well educated one probably won't.
Yes. AIs hallucinate. But so do social workers. And building inspectors. And traffic planners.
When it comes to really important systems, relying on AI as we know it is foolish. The importance of human vigilance should never be underestimated. And my impression is that in systems that are known to be hazardous, there are still comparatively high IQ people overseeing them. I'm more worried about not-so-clever people being employed in not-so-important positions. So few people notice, until those positions suddenly get important (like when it started to snow).
I think the role of AI would be to create fallible, statistics-based systems. The point is not to create perfect outcomes, but to create less disastrous outcomes. I think social work is an area where such a system could improve things. Currently, on a scientific level, it is a very well-known fact that foster care is hazardous for the outcomes of children. Still, children are placed in foster care for banal reasons and because psychological problems are misinterpreted as social problem. An AI would not make perfect decisions. But it is difficult to make worse decisions than some current groups of human social workers. Only its lack of bias is a good start.
I am happily reading your substack since at least 2 years, usually getting a fresh and spirited perspective on things I care about. Thank you for that! However, your classist understanding of intelligence ("...production workers, who are no more intelligent than their ancestors") defies every fiber of my being. Let me tell you that I come from a real socialist country, one of which ceased to exist in '89. The success those countries had in recruiting the brightest minds from the literal working class can not possibly have escaped you entirely?! Especially as you live in Sweden, which from my understanding is much more egalitarian than the US for example.
Of course there are bright minds among working class people. A hundred years ago, before educational opportunities were spread over the population, there were uncountable of them. My only point is that on aggregate, the generation of boomer production workers did not have children that were more intelligent than themselves. (Two of those workers were my parents, by the way. So I'm included in the category “descendants of production workers” myself.)
I live in Sweden too, and one of the secrets of our success is that, wherever possible, you try to get the problems solved at the bottom, and not at the top. Usually management tells people what they want accomplished, but leaves it to them to figure out how to do it. The number one reason why disasters happen when US or worse British companies decide to purchase a Swedish one is the cultural clash between the business cultures. From a British perspective, Swedes aren't capable of being managed at all. If you try to manage them, and take away their autonomy in ways that are common in the UK, they will get terribly insulted, and then quit. I used to teach courses on the cultural difference back when fixing industrial problems was my full time job.
But I think that Tove's greatest experience is with the hospital/medical bureaucracy in Sweden. And it's very much not run along the same lines as things are usually done in Sweden, for reasons historical and having to do with the fact that it is the government and not the patients who are paying for the health care. Which you sort of need, unless you are willing to let the poor do without. But it means that a health care system can end up being accountable only to the people who count beans in the government accounting offices, or entirely captured by the healthcare workers, rather than accountable to the patients who often would prefer if things were done in another way.
Also: the problem with nurses who cannot read is mostly an effect of having a significant number of people who have moved to Sweden to get a nursing position from overseas, and who have only been in the country for a few months. They're very much not stupid people -- they are bright and ambitious and once they learn Swedish things go much better for them. This is a problem in how nursing education is done, and a problem in how you really need the new immigrants to start their training soon after they arrive. (Otherwise, you are stuck with the people who have no intention of doing the training but stick around forever, claiming to be learning Swedish and even going to the classes, but never reaching any acceptable level of proficiency. It's all a mess, and the tradeoffs all have severe defects, whatever you decide to do. )
This is a great point. Living in Sweden I probably take a certain degree of worker autonomy too much for granted. Taking away power from people on the floor and giving it to people higher up who know less about reality down there is a potentially disastrous idea. If nothing else my mother, who spent almost her entire working life in a factory has taught me that.
>>Also: the problem with nurses who cannot read is mostly an effect of having a significant number of people who have moved to Sweden to get a nursing position from overseas, and who have only been in the country for a few months.
There certainly are a lot of immigrants in the Swedish care sector. But the article in question was not about recent immigrants but about new students.
>>But I think that Tove's greatest experience is with the hospital/medical bureaucracy in Sweden.
I can't complain of decision-makers with moderate IQs in the health care sector, because doctors are very much selected on IQ-adjacent tests and (less, but still) IQ adjacent high school grades. I could complain about the health care system for a day or two, but not in that sense. When it comes to more and more not-so-bright people making important decisions, I'm much more worried about positions requiring engineering degrees, teaching degrees, and, especially, degrees in social work.
Ah, the "nurse-teaching-has-gone-to-hell" cohort around here are mostly complaining about the immigrants whose Swedish is lousy. And they are trying
to change the way they teach the undernurses as a result of this. (And with Sverige Demokraterna demanding a certain level of Swedish proficiency in the care takers that interact with the old, I think they are going to have to change their course of study.) I assumed you were reading the same sort of papers, but your complainers have the larger claim that Swedish childhood education in general has gone to hell?
>>but your complainers have the larger claim that Swedish childhood education in general has gone to hell?
No, actually they don't even say that. As I read it, they are close to explicitly complaining that their students are not studious enough. It is an isolated public service article and it was so unusual that I remembered it one and a half years after it was new. https://www.svt.se/nyheter/inrikes/universiteten-larmar-sjukskoterskestudenter-saknar-kompetens
I read something a few days ago which said that when the Anglosphere learned about the Swedish idea of work/life balance and thought that they should try it instead of aiming for all-work ... they fed these ideas back to the Swedish english-reading youth who got the idea that they were working too hard as well. But the people making the claim were members of the 'Swedish children have too long a summer vacation' crowd, which makes me discount anything they say (and want to throw my rotting vegetables at them, too :) ) But maybe something like that is going on? I'll ask my 18-23 year old relatives and see what they think.
In my experience, bureaucrats flee from accountability the way vampires flee from sunlight.
ah, I think this misunderstands how so many of them are taught to see their role: they are there to execute the process dutifully, not to achieve a specific result. For accountability you might only be talking about the higher-level folks, or the electeds who provide their marching orders.
I am not saying I 100% love this situation. But I am saying it is a better mental model for managing your expectations of bureaucrats.
I have long believed that a major factor in the 2008 Global Financial Crisis was the desire to employ more college graduates who couldn't otherwise get the prestigious-sounding job they wanted.