What happens to kids if AI takes entry level jobs? What you advising them?

Adapt.
Source: U.S. Career Institute https://share.google/4BhW8kf5S4Ny1mFxr

I worry about it a lot. My DH and I both work as software engineers, and we are pushed in every way to use AI. I didn't for a long time but it reflected negatively on me.

My company used to hire a couple of juniors per 4/5 seniors we had, but we've not had a junior for more than 18 months.

I've been in my job for 10 years and AI basically does everything now. All the metrics my company looks at are "How many AI tokens you're using" and "how many projects are submitted with AI" etc (i.e. useless metrics). I've gotten to the point now where I don't know what work I've done because AI does it all. It's awful. I try to push back but I'll suffer if I do.

I'm planning on leaving in two years, and when I do so I'm going to complain bitterly about how crap the job is now in my exit interview.

So yes, I worry terribly for my children.

Following this thread with interest. The PISA 2025 data actually speaks to this directly — countries where teenagers have structured work experience early (Germany, Switzerland) are consistently outperforming on the academic measures too, not just employment outcomes. The OECD's own analysis suggests it's the real-world problem-solving that builds the skills AI can't easily replicate.

I'm a senior software engineer and there is no such thing as a junior software engineer anymore so definitely avoid that profession.

Raising two kids aged 15 and 17, it’s my concern as well, but AI taking over entry-level roles doesn’t make education obsolete—it completely changes whatkind of education matters.
According to global workplace data highlighted in this analysis on future-proof careers, the World Economic Forum estimates that while 92 million jobs will be disrupted by automation, 170 million new roles will emerge by 2030. The key difference is that the workforce is shifting away from routine, rule-based execution and heavily toward high-level cognitive skills.
If you’re guiding a young child today, here is how to look at it:

1. Move past rote learning: Traditional educational models focused on memorisation, basic data processing, and repetitive reporting—the exact tasks AI handles effortlessly. Future-proof security comes from analytical thinking, adaptive problem-solving, and critical inquiry.
2. Focus on human-centric and complex fields: Sectors expected to see massive growth include AI & Data Science, Cybersecurity, Green Energy, Biotechnology, and roles requiring high emotional intelligence, empathy, and creative direction (UX/Creative Economy).
3. Adaptability over fixed qualification: A prestigious degree alone won't guarantee a lifetime career anymore. What matters is learning how to learn, adapting to shifting technology, and developing strong foundational core skills (verbal reasoning, mathematical thinking, and digital literacy).

There is some discussion around this, not much on MN I grant you. Some people are going back to DVDs etc on the back of it. THe question is one of

a) transparency > the big AI firms are making it very hard to measure their footprint (see Sasha Luccioni), hence the announcement this week that the UK govt has underestimated the AI energy carbon footprint by a factor of 100

b) it is growing very fast and adding to, not replacing, extant unsustainable consumption in streaming etc.

I am happy with limited and regulated use of AI to track ways of reducing energy consumption, water leakages etc. Less so with it being used to put creatives out of jobs by pirating their (my) IP.

edit to add link https://www.politico.eu/article/uk-increases-ai-emissions-forecast/

What job is she thinking of that AI can’t take?

I would love to feel as optimistic as you. I think we’re all agreed on the bad actor point and the need to regulate but this needs to be done at speed and the speed is not as yet there. The US is not going to put the brakes on while wants to “win”. There wasn’t a whole lot of international regulation in the nuclear arms race until we realised, oh wait, this is a major threat to humanity.

Anthropic acted very sensibly re Mythos. I’m talking about the technology being the risk (both inherently in how it is designed and how it improves, the scaling of autonomous agents’ use and the use of AI tech generally in the wrong hands). Other companies and individuals will not necessarily act as sensibly as Anthropic. And there is no national or international regulation yet. The godfathers of AI (Geoffrey Hinton, Stuart Russell, Yoshua Bengio) have been ringing the alarm for a while now about the real possibility that we lose control.

Geoffrey Hinton the other day at the Digital World Conference “We're at the point in history when it's urgent to try and solve this problem," he said, yet "very few resources are being put into it". He suggested "maybe one percent" of work on AI was going into making it safer. Nuts.

Back to the OP though. My children are young (under 10). I’m not even thinking of the type of work they will be doing as young adults because I have no idea what the future will look like in 10 years. If I had older teens, and they wanted to go to university, I’d advise them to do something that they’ll enjoy studying for three years and go for the life experience (I appreciate that going to uni as simply a life experience is not something everyone can afford to do or would want to do).

Streaming has (currently) a higher carbon footprint than AI. I have never seen a post on Mumsnet bemoaning Netflix, Iplayer or Apple TV and the impact of streaming on the planet.

i mustn’t have made my point about university clearly. Not all businesses will need a degree, not all degrees will lead to better jobs as they might have in the past. If someone wants to do a (non vocational) degree, they should do it for its own intrinsic value, and the chance to study in depth, rather than pinning expectations to the outcome of having a degree. And do it if they can afford to- far fewer people are going to see the debt as worth having if degrees don’t lead routinely to higher incomes.

As a teacher, I have to say that AI is not all that in lesson generation. It’s like AI writing- quite obvious and not terribly useful.

My post grad daughter got a job in her area of degree, she did so so well and I appreciate is lucky. She does however use a lot of AI in her role. My second daughter will graduate next year and has a job in mind that AI can’t take so she should be ok. It’s been a worry though

How did Anthropic act responsibly when its entire business model is rooted in mass violation of IP law?

How is making AI available to the public at large compatible with net zero? I have no problem with strictly regulated AI in, for example, medical research settings. I have huge problems with letting it loose unrestrained. And so should everyone else, given its problems with baked-in misogyny and racism.

@StealthMama
agree. I believe this is the main issue. Eventually it will fall into the wrong hands. I really can’t see our govt doin sod all to mitigate this until the horse has bolted. We are always last to come to the table.

It isn’t the AI companies that need to do all the auditing, regulating and decision making of this though, that’s my point. They are building and providing technological advancements and capabilities that organisations and individuals have to decide how to use, for what purpose, and what skills and roles they require to do it. That is where workforce strategies come in to play both privately and publicly. The type of work is changing. Some tasks will no longer be required, new tasks will be generated.

Regulation of use is advancing especially in already regulated markets.

Mythos is a a good example of self regulation. It was built to write code which means it can detect faults in that code as well though in this case at a level far supreme than by existing tools. In the wrong hands this would be dangerous - especially if you add quantum computing and the disruption this will cause in cyber security management.

Anthropic acted responsibly. Why is this perceived as a bad thing?

Lots of differing industries have developed products in the past that could be dangerous and were prevented from release and access restricted.

The more we learn about AI the more we learn that the threat is never the technology, the threat is the people allowed to use it and their objectives. Our governments need to deal with this at scale. But for every day people, AI can be an extremely beneficial tool at work, at home, in education, in healthcare and so on.

Not to mention the environmental destruction of its vast energy consumption

Great post. Cannot believe the naïveté of believing all these great new opportunities AI will supposedly offer will even come close to mitigating the destruction it will wreak - is already beginning to wreak - on the labour market.

And yes, regulation - where?!

I agree re your points about opportunities for humans to direct the design and objectives, and regulate / audit the LLMs / agents if the models being created were only narrow, specific models with tight parameters (e.g. an advanced model built specially for accounting or genomic research or to play chess to professional levels). A brilliant purely narrow model could ultimately cure cancer, yes, and hugely advance human health, find new materials, solve climate change. We could have a literal utopia. And yes there would be plenty of jobs using these AI tools to further human endeavour.

However the public stated aim of the main leading AI companies is AGI and the majority (not all) of LLMs being developed are general models, not narrow, and are becoming advanced quickly (see my point re self-improving) that even the AI companies are unclear on what their future capabilities will be. For example, Anthropic's new LLM Mythos (the one that caused the main financial institutions to shit themselves and isn’t yet publicly released) was designed as a new general model, however when testing it they found that it had an incredible ability to find cybersecurity flaws. Not designed with that objective, not the human designed aim, rather these are skills that were developed independent of that. Brilliant, but what’s next…? What will v2, 5..10 be able to do?

As to who is “regulating and auditing it” it. Well no one really. The AI companies’ own safety teams - the same ones that are lobbying the US government to adopt a 10 year moratorium on AI regulation. It’s the Wild West!

Autonomous analysis and research - who is providing the data that this AI model is using?

It can self refine yes, and compare layers of results, in order to define new prompts to investigate further or challenge its own initial results. Who’s running the infrastructure and databases and networks required to support it?

who is skilled enough to assess the outputs and make decisions on the analysis provided in the field? This type of capability could identify how to prevent cancer. But it can’t do anything meaningful with that information - only humans can.

who is determining what requires autonomous research and what doesn’t?

who is regulating it? Who are the financiers? Who is auditing it?

The opportunities for people are endless. Different but endless.

https://newatlas.com/ai-humanoids/self-evolving-ai-research/ if anyone fancies a read.

Who feeds knowledge and information to the LLMs behind AI tools? People. People do. We do, you do, they will.

Well not for too much longer really because models are becoming self-improving / self-recursive. This means that the models can write and improve their own code, adjust their own weights and parameters etc without any human involvement at all in an endless (assuming there is enough computing power) loop. This is how we get to artificial general intelligence and then soon after that artificial super intelligence and it is happening rapidly (by 2030 is a cautious estimate).

🙄

@Tffjyvkbh I would no longer advise someone to go to university unless it was absolutely necessary for their chosen career, say a doctor for example. That is coming from someone who studied engineering, has a PhD and who has worked in AI ever since. If someone was in a job and worried about being replaced I would say become the go to person for AI in your company, embrace the technology, learn its limitations and how it can work for the business. There will always be room for some humans to oversee things, make yourself that person to avoid being replaced. As for my DS I am getting him citizenship by decent from my own country so he always has choices. The UK became the service industry capital of the world and enjoyed the benefits for a long time, stopped making stuff and thought the sun would never set, mistakes were made. My husband and I are trying to save as much as we can for our DS because it will be so much harder for him to buy a house for example. We hope to provide enough for him to be able to enjoy what he does but that may not be possible.

English is the most beautiful language, because it is immensely subtle in the range of expression, and also very easy to speak atrociously, yet achieve sufficient comprehension to cope. I can't think of another language that does both so well, which is why it is spoken by so many people.

It doesn't affect your AI knowledge, but you are still confusing affect ( to cause: the verb) with effect (the result, and the noun). If you are dictating your posts, Alexa is fairly clueless on the subtlety. I'm not having a go; it's just a personal thing. Do please ignore me!

The advisory function/requirement is never going to be replaced. Some of the routine stuff will be onlined, but the bespoke subtleties are likely to remain the province of the senior, human advisor for a long while. The link in the chain that is being broken is the opportunity for a human to think, LEARN, and evaluate the AI advice. I will (because I am 70) always prefer to look someone in the eyes and evaluate their competence, intelligence and the extent to which they are working in my best interests. I have no yardstick to measure an AI algorithm that's as attuned to BS.

Edited to add the LEARN in capitals.