A Lot of 'Unskilled' Workers Aren't | Hacker News Reader

A Lot of 'Unskilled' Workers Aren't

bloomberg.com

46 pointspetethomas61 comments

> an AI program can likely do that, too

Hahaha. Yesterday I spent a half hour with Comcast's AI phone chatbot. No matter how I tried, it could not understand "cancel Hulu subscription." Finally, it connected me with a human who I said "cancel Hulu subscription" to, and he cancelled it. Mission Accomplished! But the AI chatbot was a total failure.

I am preparing my presentation for DConf, and thought I'd try some relevant pictures. I asked it to "draw a square wheel". It drew a round one. I tried many variations, never got a square wheel, although grok insisted it was a square wheel.

I now and then ask AI to write a function for me. The generated code always looks plausible, but there are always catastrophic bugs in them. The function will include tests (hooray!) but does not include answers to check the results against (boo!).

It doesn't take long to realize that AI has no idea what you're asking for.

AI has a loooong way to go.

> Comcast's AI phone chatbot. No matter how I tried, it could not understand "cancel Hulu

Seems working as intended? I’ve used a bank and Starbucks chatbots for customer service. Both times they seem so rudimentary and bad that it seemed anything could’ve been better, no LLM required. It took so much effort to get thru to a human.

I have to believe it was designed that way, to fob off as many as possible.

The thought definitely occurred to me. Afterwards, it asked me to rate the bot from 0..10. I gave it a zero.

I don't recall ever having a positive outcome with a customer service chatbot.

A couple years ago, at $Job, I got a call from a Comcast "customer satisfaction" surveybot, after a frustrating interaction with Comcast's "press any button to accomplish more nothing" phone system.

I gave the latter a very low rating.

Our Comcast internet connection went out a couple minutes later.

So, working as intended? Without internet access, you can't leave any future negative reviews.

this was likely designed that way. it’s frustrating isn’t it?

What i mean is, one could trivially craft an LLM prompt that lists various “automated capabilities” the business has as an array (each one maybe backed by an api) and then ask the LLM to rank the 3 closest matches based on the users query. Then have the user confirm if any of those choices are correct. If not, trigger the “speak to a rep” workflow.

it probably was designed that way because the original way of letting the LLM respond directly and make promises led to people asking for 100% discounts, for unreasonable refunds or other weird edge cases that ruined the business!

You mean, it was designed that way to make canceling subscriptions as frustrating as possible. Just like they've been doing with foreign call centres.

> Just like they've been doing with foreign call centres

When a call center tries that with me, I just respond to all their deflections with "cancel my _____". Eventually, they get the message.

> A person’s worth is never determined by their potential market wage.

There are different meanings to the word "worth". In the market place, your "worth" is how much value you can create for the person who pays you. Your worth to your family is something entirely different.

It's like the innumerable definitions of the word "free". I've seen many people try to make points by conflating multiple such definitions. The article does this with "worth". It's a worthless(!) argument.

This feels like a victim in search of a problem. It’s an economic term used to refer to people who are in easily-replaceable positions. You can drop someone in, give them on the job training, and they’ll have it mostly figured out in a couple weeks.

This is not a judgement of the people working those jobs, and never has been really.

Besides, call them literally whatever you want - that will become the new slur on the treadmill of euphemisms.

imo the real economic danger is the regulatory response that institutes bureaucratic functions to create fake "jobs" in the form of mandatory human roles in companies that have nothing to do with the product or mission, stifling the whole economy with bureaucracy.

the thing about automation is that it's a proxy for regulation and governance, and the "freeside" economy will still use a lot of manual work becuase nobody will want the administrative overhead of dealing with AI.

My understanding is that an “unskilled” job has always meant a job that doesn’t require previous experience. It has nothing to do with how hard the job is or how valuable the work is, just whether or not it requires prior training or not.

To add to this, I think it's also whether prior training will make a substantial improvement in productivity.

Yeah, I always assumed it meant a job where the lowest level was accessible without training. Economists don't have a reason to distinguish between a job that will build skills versus one that won't, and I'm not sure the latter even exists.

One example from the article was working as a caregiver. My brother worked as a disability carer while he was training to become an engineer. I was living with him at the time, and he started out with a baseline ability to care for others, no particular skills that an average person wouldn't have, and no qualifications other than a police check and a first aid certificate. He was essentially paid to drive people with mental disabilities around and not get frustrated with them.

On the one hand, I get why someone with 20 years experience as a disability carer wouldn't want their job to be seen as something a uni student can pick up in a week. On the other hand, my brother could just go become a carer, so it's not entirely wrong.

That doesn't encompass most of what people call unskilled labor. Even basic restaurant and big box store listings will typically ask for prior experience in restaurants or retail.

The terms are a bit vague and prone to misinterpretation but economists usually refer to skilled labor jobs as those requiring certain mandatory credentials. Could be an educational degree or an occupational certificate or a government license. Some of those certificates or licenses also have minimum work experience requirements.

I think this is correct although it points to the crux of the problem, at least in the US I think.

My impression is a pervasive disconnect (again, at least in the US) has started to develop between those types of mandatory credentials and the actual skills required for the jobs, or even for training for the jobs. So say, the credential starts to be equated with the skill in hiring or for training admission, when it only bears some weak relationship in reality.

There's lots of causes of this I think, but the net result is a spurious decrease in competition, increased prices, and greater income inequality and underemployment. I think the article is a little odd but is pointing to these kinds of issues.

http://archive.today/p5LPz

“Easily replacable” labor then

Why, skilled workers belong to an American Federation of Labor craft union, and unskilled workers belong to a Congress of Industrial Organizations union, of course. Skilled workers are trained by the union through apprenticeships. Unskilled workers are trained, minimally, by their foremen. And the CIO is a Communist front, say the bosses.

That's the US classic labor history view.

"Unskilled labour" isn't the same as "unskilled worker"

IMHO more appropriate would be "low skilled labour".

We should use the correct term when describing jobs that require little specialised skill and separate them from the persons performing it (who may be very skillful in other non-related areas).

If there is cultural stigma in using this term, should we sacrifice accuracy and succinctness in technical/professional discussions on related topics by dancing around with terminologies? If we "evolve" and use a new term that is not currently associated with the same stigma, will this neutral term eventually still be associated with the same stigma? If so, is it worth the effort and penalties (e.g. interim confusion) as a whole?

I am genuinely curious about this, I think there are many good points on "both sides of the coin".

https://archive.is/p5LPz

No such thing as unskilled labor.

The longer it takes to train someone to do a job, the more skilled it is.

If anyone with a pulse can do it, it's unskilled. And I say this having worked those jobs for a number of years.

Being able to dig a ditch and being able to dig a ditch well are two different things. I've worked with people who can't do basic tasks after a decade at the same place.

Amazing how often "unskilled" is used to describe those whose skills are merely low on the social prestige scale, or who are ill-positioned to ensure that they & their peers get an oversized slice of the economic pie...

Skills that are low on the prestige scale are unskilled. Skills that pay well automatically rate higher on the prestige scale. Skills that pay poorly are skills that can be easily provided by whoever, which is what "unskilled" means.

I have lots of skills that would be incredibly valuable in an ant context; that doesn't mean they count as "skills" in a human context.

> Skills that pay well automatically rate higher on the prestige scale.

This is pretty true. Though manual labor usually has less prestige at equal pay.

> Skills that pay poorly are skills that can be easily provided by whoever

This is much less true. There are plenty of skills that require a master's degree level of learning but also pay poorly.

> There are plenty of skills that require a master's degree level of learning but also pay poorly.

There are many more of such masters than the market needs, so the pay drops until Demand meets Supply.

Of course. But low demand doesn't make them unskilled, right? I'm arguing against the idea that low pay/prestige makes them unskilled.

You could even have extremely skilled, moderately paid, but still relatively low prestige. And top of the trades pyramid go to this category. Think of specialised welders, were you need both years of expertise and probably quite a bit of aptitude... And the job is no way unskilled, but still would have less prestige in general population than some simpler job more of white collar type.

There is also the case of the 'outcast' skills which are currently stigmatized. Unless there is a literal caste of people stuck doing these jobs they tend to ironically have to pay better. Porn actresses get paid in part for the 'reputational damage' as even today people are deeply silly/stupid about past work.

Skilled labor in terms of "needs many years training" outcast skills are practically a hypothetical, and are what I would call 'high entropy' in the sense that it would take a lot of work and rare combinations to achieve. Practically an exercise for science fiction writers to depict say, a society where say heart surgery is viewed as deeply shameful yet still managed to perfect the art of heart surgery anyway.

The more skilled a person is, the fewer such persons there are. The fewer there are, the more compensation they can demand.

It's the Law of Supply and Demand.

If you hold everything equal, then changes in supply work that way. But everything else is very far from equal across different types of job, enough to make supply-based calculations basically useless.

The market is very very good at price arbitrage. Far, far better than central economic price setting, as endless experience shows.

Seattle recently tried to set hourly rates for gig delivery drivers. It resulted in a collapse of that business, as prices went so high the demand dropped and the drivers made less.

Is that true if the central planners had access to the digital data of Walmart, Costco, Target, Amazon, etc all at the same time?

I wonder if the amount of data needed for efficient central planning is now here.

Yes, it always applies. Even to monopoly pricing.

The mistake make with such data is that it can predict the future. At best, if the data is fairly accurate and complete it is only a good indicator of the past.

The fundamental flaw of central planning at scale is that a small number of people will sustainably predict the future well enough to plan better than if they let individuals make decisions for themselves.

Markets use prices to work out where supply and demand meet. Sellers control the lower bound (if the price is unacceptably low they can't stock the good), buyers control the upper bound (if the price is unacceptably high they can't buy the product). If upper > lower then there is a complex dance to work out what the price is going to be today.

It doesn't matter how much data you scrape from sellers, you can only ever work out a lower bound price for goods. That isn't enough for even a super-intelligence to work out how much of that product is needed. You have to let a real market play out to figure out what prices should be. It isn't a question of processing more data from sellers before a centrally planned system can work, central planners literally can't have all the data until after the market's auctioning process has played out.

You can make reasonable guesses, but the supermarkets already do that. The goods on the shelves don't appear by coincidence, a lot of thought goes into it. Central planning can only do that well or (usually) worse.

Back in the Carter era, the DoE handed out gas allocations to gas stations. Their strategy was to allocate according to the pattern of sales from the previous year.

Of course, this didn't work at all, and the country was a checkerboard of gluts and shortages.

Central economic planning has proven to be inept at handling the chaotic reality of actual markets.

That’s pretty clearly a data quality issue though; I don’t think this result necessarily holds in the case of better data.

Besides that, we don’t even have “actual markets” - take a look at all the places we’ve extremely distorted things; farm subsidies, for example. Do we have more than enough food, and is it cheap enough?

We’re already funding this stuff, and distorting the market - then playing pretend that it’s all fair competition by closing our eyes and saying “central planning doesn’t work”.

1) A subsidy isn't central planning. A key feature of central planning is having a plan that the market participants don't mutually want to follow anyway [0]. A subsidy is the government interfering but as a market participant. There is a spectrum going from [free market] -> [some interference] -> [central planning].

2) It wasn't a data quality issue. It doesn't matter how good the historic data is, they couldn't possibly have predicted future demand. Future demand isn't a function of past data.

3) If you think central planning is going to work, you're going to need an argument for why to have a meaningful conversation. At the moment you're claiming it fails due to data quality issues, but there is no evidence that the central planning failures are due data quality issues. Central planners are limited by the fact that the market-optimal price point for a product is unknown until after the market has cleared, and the market can't clear properly if supply is constrained by a central plan because there will be either a shortage or oversupply. That dynamic doesn't change by having better data, the problem is that the plan prevents the data needed from existing.

[0] In theory I suppose you could have a central plan where you just copy Walmart's existing internal quarterly plan, tell Walmart to do that then don't enforce anything if the situation changes, but that'd be pretty pointless even by government bureaucracy standards.

Isn’t this the plot of Idiocracy?

The network of central planners is astronomically smaller than individual companies acting on their own for their own success. A central planner wouldn't know how to begin to use all that data because their motivations would be to make the impossible reality rather than let the ecosystem of the market thrive. They can say that's what they want, but their day to day would be to control the economy as a whole.

Sure. But the mechanism of that arbitrage is mostly that training changes to match demand.

If demand drops for an uncommon/rare skill, then you can get an oversupply that lasts for decades. Wages can drop a lot despite the skill still being uncommon/rare.

My training is as a mechanical engineer, but I switched to software because it paid a lot better. I did not have a formal education in computer science, but my abilities as a mechanical engineer made the transition fairly easy.

In my younger days, the best programmers were usually physicists.

Well, if you've never known an entitled "idle rich" person, nor worked at a high-skill company taken over by vulture capitalists, nor found yourself bound by a malicious non-compete, nor been a member of a "lesser" gender or caste, nor ...

The law you mention is very nice, on paper. In the real world, there are a very large number of people whose interests it runs contrary to. And those people control quite a bit of our social, economic, and political systems.

Every time a government ruling defies the Law of Supply and Demand, we are reminded that they might as well legislate that pi = 3.

The LoS&D is very much "in the real world".

Ah, Economics theory. Where you know the thing lying on the sidewalk can't be a dropped $20 bill, so you don't even think about picking it up. And Pinkerton's goons have never done anything bad to workers demanding higher wages, and the Tulsa massacre couldn't possibly have happened, and the AMA can't be limiting the supply of MD's, and ...

... and socialism still never works.

Free markets perform better than any other system, over and over.

Do you have any examples of industries that operate as a free market, one that has never received government subsidies, grants, or bailouts? (i.e. socialism)

One that doesn't abuse the intellectual property system (e.g. patent trolling), or exploit their dominant market position in anti-competitive ways?

When deciding which system "performs better", there are deeper societal factors to consider than just return on investment for shareholders.

Sure. The software industry.

> When deciding which system "performs better", there are deeper societal factors to consider than just return on investment for shareholders.

Name an aspect where socialism performs better.

There's no free market. Therein lies the fallacy. All markets operate and depend on clearly written and well enforced legislation to function.

I agree, and a pound of butter doesn't exactly weight a pound, either.

A free market requires a government in order to protect rights and enforce contracts. It's not anarchy.

That sounds like you are projecting like IMAX about economics and assuming that it simply must be loaded with denial-ism of anything bad that ever occurred under capitalism.

I never said free markets produce perfect result. Nothing human is perfect. I said free markets produce better results than socialism, every time.

I fail to see what the idle rich have to do with supply/demand. Vulture capitalists buying and breaking up companies is very much in line with supply/demand (in the financial markets). As are non-competes (malicious or otherwise) which directly effect the supply of skilled labor. Discrimination based on non-utility factors directly influences supply/demand and thus influence prices of all sorts of things.

Economic theory around supply & demand simply describe how economies work. It does not imply that it will lead to a better life for all. In short, the "law" of supply & demand is about price discovery and resource allocation efficiency. It has no moral or ethical component that you seem to feel it implies.

What the breaking up of companies does is reallocate the resources of an inefficient company to more efficient uses. Free markets are not steady state - they are a chaotic system constantly creating, disassembling, and reallocating resources.

Reply on news.ycombinator.com