> There is a lot of debate whether AI will surpass humans in all economically viable skills (AGI, by one definition).

Actually very little debate. We get a lot of unsubstantiated hype from companies like OpenAI, Anthropic, Google, Microsoft. So-called AI has barely made a dent in economic activities, and no company makes money from it. Tech journalism repeatedly fails to question the PR narrative (read Ed Zitron).

> Regardless of whether this will happen, or when, many people already have lost their jobs in part due to the emerging capabilities of AI models…

Consider the more likely explanation: many companies over-hired a few years ago and have cut jobs. Focus on stock price in an uncertain economy leads to layoffs. Easier to blame AI for layoffs than admitting C-suite management incompetence. Fear of the AI boogeyman gives employers the upper hand in hiring and salary negotiations, and keeps employers in line out of fear.

People lost their jobs because of tax changes.

Yes. When the dust settles and the data gets crunched I expect tax code changes and a glut of money looking for returns will explain the layoffs much better than AI doing anything useful.

And the tax code makes sense, because it was a Trump tax and a lot of these companies support him so have to tow the line.

It couldn't be that people lost jobs because of the policies they voted for.

> Actually very little debate

Isn’t the debate about timeline?

Do you really believe there is a 0% chance that AI will surpass humans in all economically viable skills in the next say 100,000 years? If not, what confidence do you have in it?

> Actually very little debate. We get a lot of unsubstantiated hype from companies like OpenAI, Anthropic, Google, Microsoft

Would you really consider the Nobel laureates Geoffrey Hinton¹, Demis Hassabis² and Barack Obama³ not worth listening to on this matter? Demis is the only one with ulterior motives to hype it up, but compared to normal tech CEOs he certainly has quite a bit of proven impact (Alphafold, AlphaZero) to be worth listening to.

> AI has barely made a dent in economic activities

AI companies' revenues are growing rapidly, reaching the tens of billions. The claim that it's just a scapegoat for inevitable layoffs seems fanciful when there are many real-life cases of AI tools performing equivalent person-hours work in white-collar domains.

https://www.businessinsider.com/how-lawyer-used-ai-help-win-...

To claim it is impossible that AI could be at least a partial cause of layoffs requires an unshakable belief that AI tools could not even be labor-multiplying (as in allowing one person to perform more work at the same level of quality than they would otherwise). To assume that this has never happened by this point in 2025 requires a heavy amount of denial.

That being said, I could cite dozens of articles, numerous takes from leading experts, scientists, legitimate sources without conflicts of interest, and I'm certain a fair portion of the HN regulars would not be swayed one inch. Lively debate is the lifeblood of any domain that prides itself on intellectual rigor, but a lot of the dismissal of the actual utility of AI, the impending impacts, and its implications feels like reflexive coping.

I would really really love to hear an argument that convinced me that AGI is impossible, or far away, or that all the utility I get from Claude, o3 or Gemini are all just tricks of scale and memorization entirely orthogonal to something somewhat akin to general human-like intelligence. However, I have not heard a good argument. The replies I get seem to be largely ad-hominems toward tech CEOs, dismissive characterizations of the tech industry at large, and thought-terminating quips that hold no ontological weight.

1: https://www.wired.com/story/plaintext-geoffrey-hinton-godfat... 2: https://www.axios.com/2025/05/21/google-sergey-brin-demis-ha... 3: https://www.youtube.com/watch?v=72bHop6AIcc 4: https://www.cio.com/article/4012162/ai-begins-to-reshape-the...

> Would you really consider the Nobel laureates Geoffrey Hinton¹, Demis Hassabis² and Barack Obama³ not worth listening to on this matter?

Obama, no. Geoff Hinton has his opinions and I’ve listened to them. For every smart person who believes in AI and AGI happening soon you can find other smart people who argue the other way.

> AI companies' revenues are growing rapidly, reaching the tens of billions.

Trading stock and Azure credits don’t equal revenue. OpenAI, the leader in the AI industry, loses billions every quarter. Microsoft and Google and Meta subsidize their work from other profitable activities. The profit isn’t there.

> The claim that it's just a scapegoat for inevitable layoffs seems fanciful when there are many real-life cases of AI tools performing equivalent person-hours work in white-collar domains. https://www.businessinsider.com/how-lawyer-used-ai-help-win-...

A few questionable anecdotes? Given the years since ChatGPT and the billions invested I’d expect more tectonic changes than “It wrote my term paper.” Companies have not replaced employees with AI doing the same job at any scale. You simply can’t find honest examples of that except for call centers that got automated and offshored decades ago.

> To claim it is impossible that AI could be at least a partial cause of layoffs requires an unshakable belief that AI tools could not even be labor-multiplying (as in allowing one person to perform more work at the same level of quality than they would otherwise). To assume that this has never happened by this point in 2025 requires a heavy amount of denial.

I might agree, but I didn’t make that claim. AI tools probably can add value as tools. If you find a real example of AI taking over a professional job at scale let us know.

> That being said, I could cite dozens of articles, numerous takes from leading experts, scientists, legitimate sources without conflicts of interest, and I'm certain a fair portion of the HN regulars would not be swayed one inch. Lively debate is the lifeblood of any domain that prides itself on intellectual rigor, but a lot of the dismissal of the actual utility of AI, the impending impacts, and its implications feels like reflexive coping.

We could have better discussions if the AI industry wasn’t led by chronic liars and frauds, people making ridiculous self-serving predictions not backed by anything that resembles science. AI gets literally shoved down our throats with no demonstrated or measurable benefit, poor accuracy, severe limitations, heavy costs that get subsidized by investors and the public. Forget about the energy and environmental impact. Which “side” acts in good faith in the so-called debate?

> I would really really love to hear an argument that convinced me that AGI is impossible, or far away, or that all the utility I get from Claude, o3 or Gemini are all just tricks of scale and memorization entirely orthogonal to something somewhat akin to general human-like intelligence. However, I have not heard a good argument.

I have. You just need to take the critics seriously. No one can even define intelligence or AGI, but they sure can sell it to FOMO CIOs.

> Obama, no. Geoff Hinton has his opinions and I’ve listened to them. For every smart person who believes in AI and AGI happening soon you can find other smart people who argue the other way.

So your argument is "smart people are wrong sometimes so I will say he is one of those people". Not an incorrect argument in principle, but it can be used to arbitrarily dismiss any opinion you disagree with. I think what's more important is the delta of average sentiment among experts. The general opinion you'd get talking to reasonably smart, well informed tech-adjacent people on the impending approach of AGI is very different now than a few years ago, and these are people who can tell the difference between marketing hype and legitimate impacts.

>Trading stock and Azure credits don’t equal revenue. OpenAI, the leader in the AI industry, loses billions every quarter. Microsoft and Google and Meta subsidize their work from other profitable activities. The profit isn’t there.

They're losing money because they're spending a lot on talent, datacenter buildup and training runs. This makes sense given the returns on scaling. This is like how people had said "Amazon never made a profit" for years, even though their money was just being put into growth. A lack of profitability during growth isn't a weird exception that proves that OpenAI and AI companies are scams, it's pretty much the standard for the tech industry due to the benefits of being a loss leader when money invested early is instrumental for relative market dominance.

ChatGPT has 500 million active users. Over half of its billions in revenue is from ChatGPT plus subscriptions (which is mostly individuals paying for it, not b2b swap schemes.)

> I’d expect more tectonic changes than “It wrote my term paper.”

My specific example was a bigger deal than "It wrote my term paper". Even so, I know you used that example as a shorthand for a tool of trivial utility, but a tool which can effectively write at the college is invariably a big deal. However it seems that every advance in AI shifts the goalposts.

> I might agree, but I didn’t make that claim. AI tools probably can add value as tools. If you find a real example of AI taking over a professional job at scale let us know.

Your claim was that it's unreasonable to assert that layoffs were in part due to AI. My claim is that AI being merely labor multiplying is enough for it to be an impetus for layoffs. Even if it's impossible for an AI tool to completely replace one job fully on its own with no human input, if a senior engineer could use an AI tool that allows them to perform the job of 1.5 equivalent engineers at a company, then that company requires only 2/3 its original number of engineers for the same output, which would preempt a layoff of the least productive engineers due to their contribution being made redundant.

No one doubts that a smaller percentage of humans are employed as farmers today than were in the 1800s, but it would be silly to claim that since no robot exists that can fully perform every task that a farmer can completely autonomously that the massive reduction in farm jobs since the 1800s could not have been in any part due to technological advancements. No one single farmer has been "fully replaced", but the integration of labor multiplying tools allowed a smaller number of farmers to perform the duties of far more, which is what is effectively happening with AI tools.

> We could have better discussions if the AI industry wasn’t led by chronic liars and frauds, people making ridiculous self-serving predictions not backed by anything that resembles science. AI gets literally shoved down our throats with no demonstrated or measurable benefit, poor accuracy, severe limitations, heavy costs that get subsidized by investors and the public. Forget about the energy and environmental impact. Which “side” acts in good faith in the so-called debate?

Every booming industry is rife with those seeking to maximally exploit its growth potential. This was true with the emergence of electricity, the dot-com boom, mobile phones, etc. Anyone with a brain could see that in any new flashy domain there is a lot of deceptive hype, but anyone with a brain could also see that these new advancements did massively impact the world and in many ways improved productivity and general technological enablement.

> I have. You just need to take the critics seriously. No one can even define intelligence or AGI, but they sure can sell it to FOMO CIOs.

You still haven't provided any substantive arguments beyond "AI company CEOs are all a pack of liars", outright denial of any claim I make, and reflexive dismissal of any citation I provide.

The substance of your arguments seems to rely on retreating to a bastion of impossible to certify claims, using the lack of absolute consensus in one direction as evidence that the complete opposite claim must be correct. Claiming that "no one can define intelligence or AGI" has absolutely no bearing on whether LLMs or the like are effectively performing many tasks humans consider essential to intelligence. These semantic hideaways would be a sensible area to take shelter in if this were 1999 and machine intelligence were a far-off fantasy, but ignoring much of the realities emerging across the economy and society is a rhetorical flourish you can only pull off for so long.

As much as I’d like to continue debating with you I think we will just continue talking past each other.

I didn’t address the possibility of AGI, the consensus or “delta” of smart people, the utility of LLMs, or the prospect of OpenAI et al. someday making a profit in my original comment. You raised those issues and then put words in my mouth.

My original comment amounted to an Occam’s Razor argument: we can explain layoffs in terms of tax changes, economic and political uncertainty, over-hiring during COVID, and management incompetence. The impact of AI remains to be seen. The hype and fear-mongering serves the AI companies in a blatantly obvious way, so we should take their claims skeptically. Maybe it will all come true, but even if we get AGI next week that won’t explain the layoffs in tech and other sectors that started several years ago.

In a few years we will have more evidence to support (or not) AGI, whatever it means. And we will have more evidence of the real economic impact of AI. Right now no one can claim expertise, just opinions.

I feel that it’s important to make a note of the kinds of discussions people have on this site which commonly recur yet offer no benefit to either party. Some of our disagreements lie in the realm of speculative confidence and thus whether either one of us is correct is up to our biases and trust in various sources. However, there are some points of disagreement which I feel don’t factor in the very real nuts and bolts realities and are tiresome to see brought up. However I believe you and many others are, due to a fear or animosity towards the implications of near term AGI, are engaged in motivated reasoning. Even the smartest people fall victim to this folly, even Einstein for a while could not accept the implications of quantum mechanics.

You seem like a very smart person, so I implore you to audit the motivations driving your conclusions, and whether you would be willing to entertain ideas that are correct despite being unpalatable. The ability to accept and effectively grapple with unsavory realities is a massively useful skill in a world where long held assumptions are being upended every day.

Specialist skills. AI is trained on data. Data pulls it towards the average. It does a good job, but only good. If great work is valuable, there's still a need for specialists.

Like right now, native mobile jobs are mostly unaffected by AI. Gemini, despite all the data in the community, doesn't do a decent job at it. If you ask it to build an app from scratch, the architecture will be off. It'll use an outdated tech stack from 2022. It will 'correct' perfectly good data to an older form, and if you ask it to hunt for bugs in cutting edge tech, it might rip out the new one and replace it with old stuff. It often confuses methods that are common between different languages like .contains().

But if very high quality data is easily accessible, e.g. writing, digital art, voice acting, etc, that makes it viable to be cloned by AI. There's little animation data and there's even less oil painting data - so something like oil painting will be greatly more resistant than digital art. It's top tier on Python and yet it struggles with Ren'Py.

Anthropic released experiment results from getting it to manage a vending machine: https://www.anthropic.com/research/project-vend-1

This is a fairly simple task for a human, and Claudius has plenty of reasoning and financial data. But it can't reason its way into running a vending machine because it doesn't have data on how to run vending machines.

The phrase “outdated tech stack from 2022” is kind of hilarious

Out of date after only 3 years?

As one using tech mostly laid down in the last millennium, I also smiled at this.

But I think what the poster meant was that the AI is not always "up to date". So if my C compiler is a 2025 version, and my code makes use of features added since 2022, then the AI can "retrogress" the code simply because it isn't aware of the new.

Or another example, imagine when JavaScript Promises were new. There's a lot more examples of "not using promises" than "using promises ". So the AI us likely to use the old pattern.

If you're doing "up to the minute " code because you put the effort in to keep abreast of new stuff, then this retrograde will seem, well, frustrating.

Outdated libraries are a big problem. Sometimes stuff depreciated 10 years ago! It's the cargo cult problem.

I feel sadly that this may lead to much information normally available for free to be guarded more tightly, as the opportunity cost for not putting a barrier between valuable data and access to it increases with time.

One skill I believe will take the longest to be replaced by AI is interdisciplinary problem solving that involves empathy, ambiguity, and context-shifting across real-world domains.

What makes this hard for AI isn't just the "thinking" part—LLMs are already impressive there. It's the judgment required to navigate situations where data is sparse, stakes are high, and the path forward involves reconciling trade-offs between humans, systems, and unpredictable environments. Think of a doctor diagnosing a rare condition while calming an anxious patient, or a civil engineer weighing community concerns during an infrastructure redesign.

Another safe haven for now is the ability to guide and control AI. People who deeply understand the limitations of AI systems—and can design governance, oversight, and fallback processes—are going to be essential as AI becomes more embedded in critical decision-making.

I’m currently working on a side project called “AI Chat Co-Pilot” that flags architectural dependencies and offers deployment guidance based on company-specific code analysis. One of the key things I’ve realized is how much human context and judgment still matters in deciding what not to automate. So maybe the safest skill isn't technical or creative alone, but knowing when to lean on AI—and when not to.

Curious to hear what others think are the underrated "sticky" skills that AI will have a hard time swallowing.

To keep it with Geoffrey Hinton: "Train to be a plumber" https://www.youtube.com/shorts/WF9KEMmOm7k

Ironically, maybe writing. I've been able to get LLMs to translate well, but I can't get them to write well.

More realistically, though - the areas with the best lobbying and strongest unions.

By far, Deepseek writes the best of any model I've used. I've heard it's because they totally ignored all copyright laws and got a bigger training set of higher quality literature.

I've noticed the same! I assumed it was something to do with its reinforcement learning making it more organic and less LLM-speak, but that sounds plausible.

"Key Account Sales" are difficult to replace because these sales experiences are not documented and depend on many factors in the field. They are not used as input for LLM. Therefore, LLM does not have the knowledge of "Key Account Sales", and this skill is very valuable.

Aside from helping with the administrative part of the work like drafting docs or formatting excel files, it's hard to imagine an LLM that does a salesman's job better than a person, specifically regarding the sale of products with long sales cycles.

Massage. Anything else where literal human touch is part of the value.

Plumbing.

Embalming and funeral direction.

Childrearing, especially of toddlers. Therapy for complex psychological conditions/ones with complications. Anything else that requires strong emotional and interpersonal judgement and the ability to think outside the box.

Politics/charisma. Influencers. Cult leaders. Anything else involving a cult of personality.

Stand up comics/improv artists. Nobody’s going to pay to sit in a room with other people and listen to a computer tell jokes.

World class athletes.

Top tier salespeople.

TV news and game show etc hosts.

Also note that a bunch of these (and other jobs) may vanish if the vast majority of the population is unemployed and only a few handfuls of billionaires can afford to pay anyone for services.

I’d also note that a lot of jobs will stay safe for much longer than we fear if AI continues to be unable to actually reason and can only handle patterns / extrapolations of patterns it’s already seen.

> Anything else where literal human touch is part of the value.

Anything where a smile is part of the value.

In software? I don't see how high stakes infrastructure changes can be done by current AI technologies. It's different from programming in that you might only get one chance to do the things that you need to do in the right order with the right configuration for your environment. You might use AI for initial brain storming, but then you need to dig in to the official documentation for your specific version of e.g. a database, and then validate each individual step, and so on.

Yeah no so much the changes (blue green and monitoring helps here) but architecture decisions that bite later, years down the track.

I think the hardest skills for AI to replace will be those that require human trust, emotional intelligence, and real-time physical adaptability.

For example, Conflict resolution, therapy, and coaching depend on nuance, empathy, and trust.

Skilled trades like plumbing, electrical work, or HVAC repair, Auto mechanics, Elevator technicians.

Roles like physical presence plus knowledge Emergency responders (firefighters, EMTs), Disaster relief coordinators.

Anything where a human is still affordable and the task is perceived as better when done by a human.

The example I’m thinking of is a language tutor. If you only listened to AI hype, you’d think that tutors are going to be extinct in a year or two.

But the reality (which I have direct personal experience with) is that no one really wants to learn a language from a robot, no matter how human it seems or how sophisticated its learning algorithm is. Free or low cost language learning resources were already basically universal prior to recent AI tools.

At the end of the day, people want to have a human connection for many (but certainly not all) things. This includes the actual learning conversation, but also things like a weekly scheduled meeting, a relationship that grows over time, paid incentives to show up, and so on - things that some AI app is not going to be able to do.

If the cost of the human connection stays within reasonable limits, I don’t see it being eliminated by AI.

I don't think it really matters in the end. For example if AI takes over enough jobs or roles in general, if you train to be a plumber, who is going to pay you to do any plumbing?

Masses losing their job due to AI (or for whatever reason) will have a widespread effect on every other sector, because at the end of the day huge part of the economy is based on people just spending their money.

Working with business people, solving XYProblems and figuring out the “what needs to be done”. Telling computers what to do has always been the easy problem for 90% of what software developers do.

That’s why if you look at the leveling guidelines for any well known tech company, “codez real gud” only makes a difference between junior and mid level developers. After that it’s about “scope”, “impact” and “dealing with ambiguity”.

Yes I realize that there are still some “hard problems” that command a premium for people to be able to solve via code - that’s the other 10% and I’m being generous.

You have started with the assumption that the business people are not going to either be replaced by AI, or that they won't be directly working with an AI development tool themselves.

I am sure this time it will be different and the “no code” solution will allow non developers to create their own solutions.

Owning assets :)