When Will AI Take Your Job?

I don't worry about AI automating away my job not because I don't think it's possible (though I do think it's much much further out than the hype would suggest), but because an AI automating away software creation is the economic equivalent of nuclear war: it would irrevocably alter everything about the way the world is run and is therefore impossible to adequately prepare for on an individual basis.

Most of us in software are automating other people's jobs: we learn and understand requirements and build things that make other people more productive or erase their jobs entirely. The rate of automation is right now limited in large part by the availability and cost of software engineers. If AGI can reduce that cost to ~0, then a massive percentage of the economy would be wiped out in a matter of months. What is any individual software developer supposed to do to prepare for that scenario?

It's not just issues of tech and economics but perhaps more importantly, of power. If workers have political representation (power) than these productivity increases can feed into wage increases and hour reductions. A big if. In USA, workers are pretty powerless. That means it is most likely to result in massive waves of unemployment and political instability.

When I first moved to the city in which I live there were NO homeless. That was in 1980. Since then the working class economy here crashed and the professionals took over. Now we step over bodies in the streets.

I think it is likely that the AI wave will be this, but worse.

BTW my last job was at an AI company as a high level tech contributor.

I think a lot of this thread (and HN in general) seems to think that their case is special and AI will not make any dent there. While this is true for many comments, it reminds of several studies that a majority of people think they are better than average in their field. We tend to view our tasks as special cases. People are trying to see how the current AI fits in their current process.

However this won't be the case. The AI might improve and improve. And companies might change processes to revolve around the AI. Just like how companies change their processes to cater to ERP systems, they might change them to cater to AI systems. AI won't have to talk to 10 teams and play politics for prioritisation.

If AI is solving everything and ultimately phasing workers out, then what exactly would you be educating your kids in? Human's limiting factor is the time that it takes them to learn something (18-27 years of education depending on the field). What job would be worth striving for? They'd be training for jobs that a future AI would be trying to phase out. This whole article just described how most of the the highest paid workers (doctors, lawyers, driving, logistics, film, music, radio, etc etc) would be automated. I don't know what you'd need a person for at that point. Feels like a race to the bottom.

This, just like many other takes, confirms that there will be a short term boom in applying AI to everything. If you are in your 30s or 40s and surf this wave correctly, this may very well be the last job you will ever do. So my plan is to learn how to plug GPT to excel, databases etc and provide AI for the smaller companies that can’t integrate AI like Salesforce and big companies like Azure Professional Services will not find them profitable enough to deal with.

Squarespace was supposed to take away the jobs of people who make websites, but it didn't. The same people who you used to hire to make a website for you are still doing that, but now they just use squarespace for it.

I saw a recent quote for a very basic, static website that was greater than $10k. This would be about a day of work to put together in squarespace.

AI is not going to take away jobs, it's just going to make the people already doing them more efficient.

It's really feeling like AI is the new crypto. Majority of the people hyping up AI are directly benefited from the raise in AI. The same way the people that were hyping up crypto either ran blockchain, dapp, or were VC invested into it.

I don't know the writer of this blog post, but I'm willing to bet he has posts about how crypto will replace the financial system.

There's an anti-productivity link round the front page that answers this well - the amount of useless work to be done is infinite, and growing productivity on useless work only lets you do more useless work.

The more useless work ai is able to do, the more useless work will be asked of it. The ai doesn't have opinions about what it should be doing, so without people to push it into doing the right thing, it's going to be answering a lot of emails

As a non senior data engineer, I think AI can take my job already, except that it requires my company to open up its code base for AI to consume and train seniors to be better prompt engineers so that they spend less time modifying code written by AI. There are a lot of edge cases but I think the tech is already here.

To escape from that fate I'm going to drill deeper into k8s, docker and Linux system programming. I know a better route is to launch a career that translates requirements from business stakeholders and develop an AI-friendly architecture so that our AI lords can conveniently code functions, scripts and programs without much interference. But throughout my career life I have tried my best to avoid that road (because I'm not interested in business), so I have to pick the second best way -- serve our AI lords by maintaining infrastructure for them, just like rhino birds serving rhinos I shall serve our AI lords faithfully in eternity, even after death when my coding skills can be consumed by them.

There are plenty of students who can ace these exams but then utterly fail at getting things done at companies because they lack problem solving experience, a toleration for ambiguity or any number of interpersonal skills needed to complete complex projects. Until AI can just deal with other AI and not messy humans, there will be job security.

I think my job is reasonably safe simply because "AI" is never going to be as effective at it as "me + AI". Someone has to be there to understand the problem in human terms in order to know what to tell the AI to do. I've recently started pushing myself to take a more Systems Thinking approach to solving problems, and to start using AI in my work, precisely because I see this as a change that's coming. It is inevitable. I might as well embrace it.

Also, and maybe this is selfish, I suspect fewer people are going to see tech as a good career option in the very long term (10+ years). The experience I have now puts me ahead right now, but if there are fewer devs entering the industry in a decade it'll keep me in work for a long time. I'm think (hope!) I'm going to be OK.

If I was a junior right now, or a student, I would see AI as a much greater threat.

Seems like landlord continues to be the only secure career path.

What I find interesting is what professions are concerned about AI. For the most part they appear to be better paying white collar need-a-college-degree type of jobs. Tech, copywriters, marketing, designers, lawyers, accountants, etc. All feeling the heat.

The person pumping gas? Cutting hair? Janitors? Nurses? They, for now, seem to be immune.

As a side note, anecdotally, productive gains have a ceiling. Sure Copilot and ChatGPT free me up to focus on the heavy lifting. But my brain can't run that relentlessly all day. It seems to need to catch its breath.

I will admit that I am personally on the fence. There are parts of my job that could definitely be augmented by LLM ( I know, because I have been trying to use them as a way to test whether it could work in my domain ). Some results were cool and shaved off a fair amount of time for me ( for example, it cut down some time on proposed approach to problem, where there was a solution already ), but I can't say it would, in its current state, replace me ( Python and SQL code quality gpt generated seemed to rely too much on how specific the prompt was and I typically had to refine it as I went through it ).

But we are merely starting out. I actually started looking into building excel plugin ( first time ever too so, who knows how that will pan out ) for my use case ( categorizing some unstructured data into structured data, which is what GPT seems to excel at ).

And the one thing that was initially protecting me is that some companies, rightfully, identified ingestion of prompt data as a massive privacy issue. Now OpenAI supposedly said data ingestion from prompts and provided files will be opt in only, which may mollify some of those companies.

OTOH, some things can't possibly be replaced by AI ( last week we just dealt with a vendor requirements translation and, uhh, good luck with all that GPT ).

> The main winners of these trends will be all of us consumers, with access to much better content. Also, the best creators, who will use these tools to generate lots of great content and rake in its benefits.

Maybe yes. But I wonder. Take that example with disappearing local football leagues: what is impact for local communities? When everything is "centralized" to very best providers... Is best always best?

Before AI can take my job, AI can be my tool. I can become more productive with AI, many times more productive than now. Then the question will not be "is AI better than credit_guy?" but "is AI alone better than credit_guy who uses AI?". If I learn how to take full advantage of AI, I have a fighting chance.

Tanks may have made cavalry obsolete (was is tanks? I don't know), but they didn't make generals obsolete. When was there ever a military engagement that was described in terms of commanders facing off? It's been the story since the beginning of war.

Likewise there are decision makers in civilian life. They decide all sorts of things, sometimes relying on people to implement their decisions, sometimes machines. Sometimes they use a machine to replace a lower level decision maker. But there's always a boss.

This is what will determine whether your job is automated away.

here is my day-day from this last Friday; 1. Made circuit boards for a Photo-diode circuit(1nA max out. welp!) with Altium previous week. 2. Got bare fabs from local fab-house. 3. Received them, did opens and shorts between power and ground. Also verified internal planes were intact(no dead copper). 4. Stuffed one board(made 5). 5. passed smoke test(huzzah!!). But op-amp output zilch. 6. Debugged the entire afternoon found out that the enable to the reference chip was "pulled-down" instead of "pulled-up". I fixed it the circuit started working and op-amp output high. In the intermeaning steps between 5 and 6, I tried searching for specific debug tasks using BingAI(I was an early adopter hihi). The results were sub-par. essentially summarizing data sheets. Went direct to the source(ChatGPT) with alarmingly same results. Hence I decided to have the data sheet on one screen and my schematic on the other and traced the connections and found out my error within say 15 minutes.

Long story short, in my professional career AI is not going to be of any immediate help(in its current avatar) for the next 3-4 years. Beyond that I probably would be long retired and gone off grid.

> get in the bandwagon of AI, start using the tools, explode your productivity

I'm reluctant to start using LLM-based coding assistants (e.g. Copilot) in my programming work because I don't want to taint my code with laundered code that was under a copyleft or other restrictive license. Is this a reasonable concern?

Oddly enough half of my job is learning to get answers (from the physical world), the other half finding out if those answers are true. So far no human has come close to threatening my job if they don't actually care about those things. So it will be when AI learns to care about the truth.

This got me thinking: as more of what can be produced with existing technology, is made trivial-to-produce through enhanced automation (e.g. AI), one area that will still require labor, and that people will still be willing to expend resources on to procure, is the creation of new technology.

It would be in our interest to remove legal impediments to raising capital for new ventures, so that more of the labor freed up by automation of existing industries can be allocated to the creation of new industries.

Such a world would see more kickstarters, equity crowdsales, crypto token sales and DAOs for collaborative creation, and more people employed in the ventures these capital raising projects fund.

One job that will always be AI proof: preschool teacher.

I think even if the utopian case happen and the machine wealth get shared among people and everyone lives happy in leisure while the machine does everything, even then it would severely curtain social mobility. There will not be a way to get better outcomes in life. An even larger number of folks won't have any upward mobility. A lot of human inventions in the past 2 centuries have owed to individual drive and ambition. Once that goes away, how do anything new happens.

In that first chart there is not a single test listed that people start at a job based on their performance on the test.

I think the mistake a lot of people make is assuming that their job is actually important. We got a little sneak preview of that during the lock downs in the last two years when people's daily routines got disrupted a lot and a lot of day to day work either stopped happening or was reduced by a lot. It had a lot less effect than you would expect if you assume that most people do something important/essential. Most of what people do in their jobs isn't all that important.

This is a result of automation that has gradually been introduced since the industrial revolution. People kept on working but the nature of what they do has become increasingly detached from things that people back in the day would have recognized as work or important.

We work less than ever and lot of "work" basically involves people having meetings, managing other people, or doing things in the services industry in a complex network of interdependent tasks and processes where you might question what it actually is that we are actually doing collectively or why.

Obviously, there is some value in doing these things because we're obviously doing them and getting something out of the process of doing that. At least subjectively. But what matters economically is that money changes hands and is used to create the demand for all these things. Without demand, there's no need to do a lot of the things that we do. And without that, the need to automate it also goes away.

So there's this paradox that automating what we do, removes the demand for the thing (or at least downgrades its value). And it helps create demand for more valuable things we can do.

Take bread for example. Rather crucial to a lot of people and it used to be the main component of a lot of people's diets. Bread now comes from factories and ever since we invented sliced bread it's been considered the best thing ever. Sliced bread (sliced by a machine) from a factory that produces bread at an industrial scale.

Except we pay extra for artisanal bread that is made the old fashioned way. So we now have people with college degrees opening bakeries and selling us really expensive baked goods that they made with their own hands. Could what they do be automated? Yes, that already happened. But the whole point of what they do and why it is valuable is that it isn't. The factory bread is low value. And the bread is just part of the experience. The whole bread experience involves interacting with the amazing people that bake the bread, that knead the dough, and visiting the beautiful shops in which they do that. Add premium packaging to the mix and it makes you feel really good about spending 10$ on a loaf of bread. Can AI automate that experience. No. It might bake some bread but it won't be worth 10$.

There's going to be a lot of that. People will look for activities that are valuable to others in some way. The automated stuff is inherently lower value, boring, and a commodity. And we need value creation to create demand for whatever else it is that we do for each other.

Under the current system, millions will lose their jobs and a handful of extremely rich people will get richer: capitalists. All of this will be completely lawful: is the law of the free market.

The only answer is to destroy capitalism and unite as workers. Only if we unite as proletarians, we can redistribute the inmense profits of automation and make sure to live a worthy life.