Has anyone's job been threatened or lost due to AI, either directly or indirectly?

Miscellaneous and Personal Stuff I Must Share
21

My University admin had decided to go all-in on AI. When I raised concerns during a Dean’s meeting I was told (verbatim quote) “This is happening. Get on board or else”.

I’ve responded by eliminating every place in my classes that AI could possibly touch (chalkboard lectures, no provided notes, in-class pen and paper essay exams). I do not respond well to threats.

(I have no issues with computational AI in research; we use it all the time in my lab for coding assistance and protein prediction. I’m not down with LLMs)

22

A large number of technical writers at my company got let go. AI didn’t replace all the technical writers, but it allows 1 technical writer using AI to do the work of 2-3 technical writers who don’t use AI.

Also in the call center, its much easier to troubleshoot with AI, which reduces the skill level needed to do the job. So they are outsourcing some of the call center jobs so people with less education can do that.

23

I feel this is different than the internet. This is more like the industrial revolution. In the industrial revolution it wasn’t uncommon for a business to lay off 80% of its workforce when new machinery was brought in, and this happened across a wide range of industries. Eventually it all worked out, but only after decades and decades of fighting. The US government did everything it could to prevent workers from organizing, including sending the national guard after them.

I predict a lot of Gen Z, Gen Alpha, etc kids living with their parents and grandparents for long periods of time since they can’t find stable employment.

When the US revolution happened, about 90% of Americans worked as farmers. Then by 1900 it was about 1/3 farmers, 1/3 worked in factories. Now only about 2% of Americans work as farmers and maybe 10% work in factories. But with AI, by mid century the number of jobs a human can actually do better/cheaper than an AI or a robot will be massively shrinking. The number of jobs that require a human is going to keep shrinking as time passes. AI and robots won’t create new jobs for people, because there is going to be virtually nothing a homo sapien can do better/cheaper than an AI or a robot. Humans will be replaced from the economy the same way tractors and automobiles replaced horses, oxen and donkeys from the economy. There is nothing a draft animal can do better than an internal combustion engine.

My prediction is there will be a lot of unemployment, job insecurity and wealth consolidation in the next few decades. Maybe by the 2060s+ everything will be straightened out and we have UBI. But sadly, this being America, we will see a resurgence of plutocratic fascism long before we see UBI.

Its sad that the solution is so easy on paper. Either socialize or heavily tax the means of production (AI and robotics) as they displace large numbers of people in the labor market. Then use that wealth to support a social safety net. Much of the developed world will take this approach. In the US, the rich will use AI and their new wealth to keep the public misinformed and divided.

The next few decades are going to suck hard if you don’t have a large financial cushion or have family with a large financial cushion.

24

Back in 1998, a good friend of mine got some kind of purchasing agent certification and was hired by a restaurant supply wholesaler. As he described it, all he did was order consumables (not food, but stuff like napkins. toothpicks, take-out containers, etc.) for their distribution center. According to him, he sat at a computer and was presented with an item (e.g., toothpicks) and a history of how many were ordered last time, how many were sold, how many were sold last year at the same time, and how many were currently in stock. Then a recommendation was made to order X amount for a delivery in Y days. Almost invariably, he simply approved the order and made sure it got to the manufacturer.

AI? Not really. But it’s easy to see how a great number of purchasing, scheduling, and other mid-level management and administrative tasks can be reduced or eliminated by such systems even back then.

BTW, the pay was pretty good and he got bonuses based on performance. He did it for about two years and then took a better offer for another purchasing position.

25

Will someone explain to me how companies are profitable when vast swaths of the middle class have no income? Do plutocrats live in cloud cities already, completely insulated from the world below?

OpenAI Just Carpet-Bombed Mathematics (377 long standing problems solved)
26

I retired before AI became what it is now, but even then I could see the equation. The issue was not that I was a better writer than AI, but whether it could generate more revenue at a lower cost than I could.

Let’s say you write 100 grant applications in X amount of time, and you’re so good at it that you achieve a 50% success record. In the same amount of time, AI can write 1,000 applications, but achieve only a 7.5% success record. That means you got 50 grants, but AI got 75.

Sure, there will always be that application to the Ford Foundation or the Gates Foundation that absolutely demands a unique approach and multiple passes through the editing and rewrite process to polish. But that will be handled by the Director of Development (the only remaining human) or passed along to one of a handful of free-lancers with tons of experience in that particular niche, but who only manage to snag two or three assignments each per year.

I’m glad I retired. Best of luck to you.

27

But each company is trying to maximize its profits today (or this quarter), not trying to build a market for it to be profitable in the next decade. And it’s an arms race. Unless you own a company town and have a monopoly on the customers/employees, your attempting to save jobs at the expense of profit margin doesn’t mean you have a bigger market, it just means some other company gets that scrap of profit you left on the table. The “invisible hand” is not a source for good, except in very tightly constrained situations that don’t usually exist in the real world.

28

Agreed. But it gets worse. That’s the revenue side. The AI generated 50% more revenue.

Now let’s look to cost:
For 100 grants, we’re probably talking 1000 Weasel-hours. Let’s assume her salary amounts to $50/hr. For non-union jobs, a common rule of thumb is that the total cost of an employee (the “loaded labor rate”) is about 33% more than headline salary / wage rate. So we spent 1000 * 50 * 1.33 = $66,500 to generate that revenue.

The AI’s effort for its 1000 proposals probably cost about $500 ($0.50 each). And will be less in the future.

So 50% more revenue for ~1.5% of the cost.

Humans can’t compete with a machine like that any more than you can pay a living wage to a hand shovel-equipped ditch digger competing with a person operating a large modern backhoe. One 30-second backhoe scoop = 1 hour’s hand labor in good soil or one day’s hand labor in bad.

And guys who are good with a backhoe can average more like 15 seconds per scoop in ideal conditions.

29

I learned everything I need to know about free market capitalism from playing Monopoly. All the money flows to one person, then the game ends. Heck of an economic system.

30

It’s time we updated that game with data centers instead of hotels.

31

You both raise good points, but the missing factor here is that every grant applicant is using AI, so funders are being inundated with hundreds of identical applications, which means using AI to write a grant doesn’t really give you a competitive edge; it’s becoming the status quo.

Which means people who write grants, and do it well, are still distinguishing themselves.

Second: An AI might be able to write 1,000 applications but there are nowhere near 1,000 funders to apply to. I’m lucky if I can find 60 a year. Furthermore, if everyone uses AI they are still competing for a finite number of resources. One of the recent rejections I got from OJJDP funded 3% of applicants. Submitting more applications doesn’t necessarily translate to winning more grants. They can’t fund everyone even if everyone has a perfect application.

And the narrative is only one part of writing a grant. It also requires organization, planning, choosing projects that are aligned strategically, talking to project directors and program staff, endless meetings, and in the case of the most recent federal grant I won, 25 pages of attachments. I actually used Claude a lot for that project, I didn’t have time not to. But I didn’t let it draft the narrative for me. I took its suggestions for cuts and let it turn some lengthy paragraphs into charts, etc. I also used it heavily in planning the development of the application and will continue to use it to plan the implementation.

I may be one of those niche cases where AI helps me be a little bit better at my job without coming anywhere close to replacing me.

For grants I can’t even say I write better grants than Claude. Sometimes I’m really impressed by Claude’s grasp of granstpeak. My writing style is a bit different, I don’t know if it’s better, but it’s obviously human. And we’re living in an era where obviously human is going to stand out.

32

AI is a tool that one needs to wield competently in order to do one’s job, as I have learned from e.g. professional programmers. There is no question of “competing” with anything; operating the machine is part of your duties in that case.

33

Why would they be identical?

34

Because they’ve all been written for the same purpose by the same LLM.

I’m sure there are some program variations but assuming they consulted an LLM for program design, it’s pretty much going to be the same stuff and the same language used to describe it, with some fill in the blanks for the specifics.

35

IMHO: I’m worried about this part, both for myself and people new to the field…

I am a professional programmer, or at least used to be :sweat_smile: Haven’t programmed anything in months… now my job is to chat with a robot all day.

On one hand, Claude is definitely a force multiplier. With it, I can build by myself in hours what would’ve previously taken a whole team several weeks to build.

On the other hand, that means we need far fewer person-hours to get the same outcomes.

For the time being, this seems to mostly be affecting juniors new to the field. In a very real way, they’ve been out-competed by Claude. $200 in AI spend would go much further than hiring and training and paying an inexperienced developer for a month.

But eventually it’s going to affect the rest of us too. Teams won’t need to be as big. Classical software products have no real moats anymore. The very idea of a computer programmer might become as quaint as having a secretary who types up your hand-written letters and emails them for you (which my dad had at one point).

Now anybody can make their own bespoke apps for their phone or computer with no prior experience, just a wishlist of demands. The comp sci majors are seeing decreased enrollments, and tech companies are hiring much fewer newcomers already (roughly 80% less? Seyed M. Hosseini and Guy Lichtinger: Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data - Stanford Digital Economy Lab ). That was a study done last year on even older data, and LLMs have gotten tremendously better since then, so I imagine the situation to be even worse for them now.

For now I still have a job, but I’m not really sure why. If my boss wanted to, he could easily replace me with a few hundred dollars’ worth of Claude spend. For that matter, our customers could easily recreate our product with a few prompts.

I guess I’m afraid we’re surviving on inertia, and once that runs out…

36

You point out something that has been lurking around my industry (aviation) for awhile, and not due to AI.

It’s one thing to put very powerful tools (meta-tools really) into long-skilled hands who know what their goals and meta-goals are. It’s a different thing to put the same tools in newbie unskilled hands. More than that, what does a newbie grow into when they’ve always only dealt with the meta-tools, not the tools?

How do we grow the next generation of folks like you (or me) who really know the biz and the work and … ?

In the aviation case the Big Idea is an airliner with a single pilot assisted by an intermittent remote human pilot on the ground who’s helping a dozen other flights in turn by data link. And of course eventually the remote pilot turns into an AI. As an old head captain I could see that working OK-ish.

That’s great while we have a supply of old head captains. But where do we grow new old-head captains when the first time a newbie pilot sits in the jet they’re all alone? It’s journeyman work but they’ll have had no apprenticeship sitting alongside a journeyman for a few years.

As you said, that plan will run on inertia for awhile. Then it won’t.


Of course that argument can be taken too far.

Having been in and out of pro software dev since the punch card days I remember old heads saying that if you weren’t fluent in assembler you didn’t understand what you were doing. You could maybe sorta do it, but you were fumbling lost in the forest unless you could have written it in assembler.

It happens my first real job was working in S/370 assembler. I moved on from that to generations of higher level languages and eventually topped out doing full-stack dev in SQL, C#, and html/javascript. Glad I did it, glad I’m retired.

The same argument is made today between the C & C++ people versus the Java / C# / whatever people. With 50 years’ of hindsight I can say that the “must have assembler” crowd had a point, but not nearly as big a point as they loudly insisted.

37

I currently work as a web developer-- my particular niche, since I have a graphic design background, is to take mockups of digital products that others have custom-designed for clients and ensure the final product is as pixel-perfect close as possible to the design. So I act as sort of a bridge between designers and back-end programmers-- I don’t do design myself much anymore, but I do some, and work in design media. I don’t consider myself a ‘programmer’ but I write some code, Javascript / node.js mostly.

Seeing the incredible advances AI has made on all sides of that equation is threatening, for sure. But for now at least, it’s actually helpful. When I use AI to drastically change or extend the background in a stock photo or get some code written in minutes that would normally take me much, much longer? Easy peasy. Makes me seem like a smarter and more efficient developer.

My son, who is currently in college taking computer science, took some time this past Summer to do some networking with tech people out in the Seattle area. He says that pretty much all employed programmers these days ‘vibe code’, meaning they prompt the AI on what to do. No more actual coding, no more human debugging. It’s made them so much more productive, a lot of them take on second jobs remote coding and make a lot of money doing so.

But this, I’m pretty damn sure, is a bubble of helpfulness, a small window of lucrative ‘good times’ for skilled workers such as us. At some point, the higher-ups are doubtless going to realize that we highly paid designers, developers and programmers who are currently taking advantage of AI are no longer necessary, get rid of us, and replace us all with relatively unskilled, low-paid ‘prompt jockeys’.

I was watching a recent Jon Stewart Daily Show where he interviewed Jacob Coxon, the Anthropic researcher who resigned, warning of the existential threat of AI. Despite his warning, he expressed optimism that, if we make proper use of AI, he still believes it could result in a utopian society: a future where AI has solved problems like world hunger and disease, humans don’t have to work, though they may do it ‘as a hobby’, and everybody will get a generous universal income, so they don’t have to work at all if they don’t want to.

Stewart’s rebuttal was that, as wonderful as the tech oligarchs like Musk make those kinds of utopian visions sound, the reality of what we’re doing with AI right now is helping military drones target the enemy more effectively, figure out the ideal times and situations for businesses to surge price, and replace skilled workers. In other words, all war and greed. This does not bode well for future use of even more powerful AI, even if it stays under human control and doesn’t take over and wipe us out itself.

And how would a universal income fly at all, he went on, especially in this country, where any payout without doing some sort of work for it is considered ‘Socialism’ or worse yet, ‘Communism :scream:’? But we’re to believe that some sort of AI-powered Socialist Utopia, where we’re all paid handomely for doing nothing, will happen one day?

And then, even if that utopic vision could or would happen, what’s going to happen to the human psyche when people have nothing to do that makes them feel useful? I mentioned earlier in this post that, for now, AI has made my job easier. But it’s also made it less satisfying-- having AI do what I used to do-- create an amazing image or write code that works well is easy, but feels very empty and professionally unfulfilling.

38

I have to assume there’s a tipping point. If half of the workforce became permanently unemployed, I think people would think a lot more seriously about universal basic income.

But yes, early results on how AI is being used do not bode well. I don’t predict a utopia. And as you noted, sitting around doing nothing doesn’t feel utopic to me at all.

That said, I expect AI to affect sectors unevenly. And you know there’s going to be some class warfare between people replaced by AI and those who still get to be employed.

The thing is I don’t know how much I buy the notion of mass job displacement in the first place. I can see it having a big impact on some sectors but not all of them.

That being said, I’d hesitate to recommend grant writing as a career path to a young social worker. I don’t know what’s happening in grant consulting right now but the field is probably taking a hit as well. Too bad for me because that was my long-term career goal. If it weren’t for the federal funding crisis I think nonprofits would have been the last sector to adopt AI tools. We’re just slow to change. But necessity being the mother of invention, people in the sector are desperate to find new ways to sustain their programming.

My agency lost $400,000 in operational support this year because of Victims of Crime Act (VOCA) cuts. How do you keep the lights on? By any means necessary.

39

Who would be dealing with Claude, then? Your boss? Force multiplication may mean fewer workers are necessary, so any individual’s job is at risk, but ultimately there has to be somebody around who knows what they are doing.

Ain’t no high-level languages in The Art of Computer Programming. Now, you might rightly say that the majority of computer programmers do not need to know theoretical computer science. But someone, somewhere needs to (perhaps that person, in this environment, now has a rare and valuable skill and is not trivially fungible with an AI)

40

What sectors do you think will not be displaced by AI? I’m not disputing the premise or arguing the point, I’m curious what you think, since you clearly have put a good amount of thought into this subject.

Skilled trades, for one, are probably safe, at least for the short term: plumbers, HVAC techs, electricians, maybe network engineers? Anything that requires a good amount of hands-on effort in addition to strong skills. Mechanics too-- it’ll probably be awhile before we trust a robot with fixing our cars and especially our planes.

I mentioned my son, going for a degree in computer science and networking with tech professionals in an effort to stay in front of the AI seismic shift. He said not even too long ago it seemed like a better idea to work toward something like project management rather than pure programming, since AI was still not good at seeing the whole picture, taking pieces of a project and combining them successfully. But that’s rapidly changing, with higher-level versions of AI being developed that automically prompt lower-level AI to do the programming ‘grunt’ work while managing the bigger picture. AI is working its way up the responsibility ‘food chain’. At some point, he said sardonically but I don’t think even half-jokingly, the only way to make money in the tech biz will be to start your own business and be the CEO. It’ll be nothing but CEOs making obscene amounts of money and low-paid ‘prompt jockeys’ telling the AI what to do.

Then there’s the alternate possibility of the AI bubble popping. Nobody wants data centers in their backyard. They consume enormous resources. AI companies, I hear, are subsidizing the public use of their AI agents and eating the huge costs themselves, to try to stay ahead in the AI race. But it’s not sustainable. Sooner or later the money will likely run out, as in the tech bubble of the late 90s, where tech companies that were basically vaporware would suck in billions of venture capitol money with not much to show for it, because capturing market share was thought to be the important thing-- the profits will come later! But of course they didn’t, and the ‘Web 1.0’ bubble popped. If that’s what’s in store for the AI industry, who can say what the aftermath of that will be?