Ask HN: Thoughts on AI and the job market in 5 years? | Hacker News Reader

Ask HN: Thoughts on AI and the job market in 5 years?

23 pointsfunerr25 comments

I'd love to hear the HN-crowd thoughts on the job market. What will stay, change and disappear?

As a JavaScript developer I anticipate there will be no changes even with the advancement of AI. JavaScript developers will continue to be poorly prepared and entirely reliant upon large frameworks to do most of their job for them. Employers will continue to hire these people and continue to set low expectations.

AI will impose no changes, because demand for skilled labor will remain unchanged and investment in proper training will remain absent. The world has had the web for 30 years and fast JavaScript for about 15 years and still has not solved for this, so based upon historical trends this will remain unchanged and AI will not be significant enough. In order for the necessary disruption to occur employers must be willing to impose internal automation in excess of external output and simultaneously lowering hiring wages. Since neither of these conditions have ever occurred there will continue to be no disruption resulting in continually rising JavaScript developer wages.

I do anticipate one major change in hiring for JavaScript developers though, which is skills diversification. I anticipate JavaScript developers will become expected to have a greater diversity of skills outside of merely writing text to screen in a browser and may be folded into a sub-component of cloud engineering.

We're past the transistor equivalent of AI (GPT-4), and into the logic gate era (functions/plugins). So it's a slow period until people build an AI layer on top of the application layer.

But computers tackled paperwork and calculations. AI tackles mental labor and attention, similar to how steam engines tackled physical work. They cover the only subset that had to be manually done by humans.

I expect the industrial revolution all over again. Higher productivity means the few will exploit the many. Some consider it unethical, maybe it is. But you can boycott unethical coffee. You can't boycott unethical guns.

The best example today is the writer's strike. We see it in art too, where companies using AI are boycotted. These sectors are likely to take a huge hit. Friends in digital art end up just changing careers - printing tshirts, becoming flight attendants and so on.

There's still a demand for these jobs, and the ones who get through the winter will be worth more.

The same goes for other professions. Those in the bottom will go first. Hiring will freeze. Those at the top will have enhanced productivity and earn more. A logo designer can brainstorm a hundred ideas in 15 mins with AI and then draw it nicely.

I don't agree with the transformer=transistor analogy. Nothing is to say we don't find a new model 10x better than transformers. That would be like perfecting very tiny and cheap transistors, before even making integrated circuits!

I was very confused until I realised you weren't talking about electrical transformers.

The way I see it, GPT-3 is like someone with 70 IQ, while GPT-4 is like 120 IQ. It's not omnipotent, but having this block of human-level intelligence that you can just call through an API is world changing to say the least.

What if they create something of effectively 200 IQ? Or 5000 IQ, where we just abstract IQ like horsepower?

I would guess Amdahl's Law applies here. You get diminishing gains, to the point that a 10x better thing or 10 giga transistors on a mobile phone doesn't really make a difference.

But to use the transistor analogy, the mates working on transistor or quantum theory didn't envision we'd have this world where gig economy influences elections or strangers arguing with each other at the speed of light using only their thumbs.

So that's where Gustafson's Law kicks in. The 1x or 10x thing unlocks a new thing, which unlocks new levels of things all with their variations of speed, where it's a race to unlock the next next level thing.

The current level thing is AI agents. What happens when you put them in a community? Will that unlock new things? There's probably a certain speed and cost restriction, then you have AI nanobot cities giving you the power to regenerate limbs.

I think you misunderstand what IQ represents.

In ten years we will be complaining how giving bad input to LLMs gives us insane and horrible output.

End of the day we’re talking about people. Most people have no idea what they’re doing or want.

I predict a lot of meetings where people will be trying to figure out what the next big "thing" will be, since this whole AI/LLM deal was way overhyped...

There's a great future in plastics.

I'd love for someone to debate me on the below:

Like all other tech it will paradoxically make people more productive and somehow less efficient. For programmers, we're all going to write more code with AI but somehow get as much done as the programmers pre-internet.

In 5 years time the most noticeable difference, apart from even more startups based around the idea of AI is that more jobs will have "working with AI tools" as a bullet point, with varying levels of specificity, reasonableness (I expect some "minimum 7y experience working with GPT6") and relevance to the actual job

And HN will still be asking if all jobs will be eliminated in 5 years' time.

I don't know, it's not like jobs now are requesting that candidates know how to Google and stackoverflow - they're just assumed tools that help you do your job. LLMs will be in a similar category.

AI will increasingly replace human decisions wherever it's feasible. This includes recruiting and filtering candidates. Management will love the idea of replacing low-level labour, even at the expense of customer and employee dignity.

It will also make it much easier to spam convincing job applications, so you'll end up with bots talking to bots, and people on both ends being slightly worse off.

Are there any jobs where you are payed to setup an in-house AI system? All "AI" jobs I have seen are basically just API jobs more or less, where you're contracting out the actual AI to some other company that runs the models on your data.

Seems like there is bound to be far more jobs creating the apps, than actually building, training, configuring, the models.

Look for in house LLM ops and infra roles. “MLops” is the term I’m seeing when people ask me to recommend people in for locally hosted LLM infra standup. Know Azure if it’s enabling an org vs offering an LLM SaaS.

Tech support chat bots will be everywhere, and they will be ridiculously more effective and efficient than the current flesh-counterparts. No more "restart your computer 5 times". It will be great, except for the edge-cases that actually require a human thought. Those cases will suck.

The thing is if you have half a brain you end up talking to a support chat bot exactly because you're facing an edge case ...

I spent half a day on a HP call. The solution was install Mac drivers for my printer. These drivers were not on the site. I didn't even know Mac needed drivers installed for common printers - my previous printer ran perfectly fine without it.

Most of that time was just turn it off and on in different ways. Including a half hour long pause where the guy asked me to unplug my printer cables and tell him whether it worked when everything was unplugged.

AI would probably have fixed this in half an hour.

Or sent you on an endless loop of restarting. You would have been better off using the support forum

You talk to the agent ultimately for whatever they are gatekeeping: a refund for example, or your data, or a physical visit to fix the street side of your internet connection.

I like the analogy of retail self-checkout system.

It exists for a decade and the cashier job still there.

Don’t be afraid of the evolution of technology, humans take time to adapt to these evolutions.

Could apply to self-driving cars too, taxi drivers are still and will be still a thing in a decade.

One possibility is that many jobs in five years may involve directing a group of AI agents.

There will likely be "job sites" where most or all of the "candidates" are AIs. These AI employees will, for many jobs, compete directly with humans. Some maybe have realistic avatars in video calls and the ability to fully control desktop or web-based software.

The adoption will be gradual rather than immediately eliminating all of the human workers.

Sure, that's definitely going to happen in 5 years.

Just like self-driving cars were 5 years away in 2010.

No chance.

I really cannot wait for this exhausting AI hysteria to die down.

More and more startups pushing their chatbots as AI. Social media filled with these chatbots correcting and yelling at each other and building a more diverse and inclusive society.

Reply on news.ycombinator.com