Ask HN: Do you think AI is threat to developers job?

[iamspathan]

During my discussions with individuals who integrate AI into their daily lives, I discovered what I was hoping to hear.

Have you noticed any change in your work productivity?

The majority of them said that they were able to reduce the time required for building. Imagine a project that used to take three days, is now wrapped up in just one and a half days! Thanks to AI, we're talking about a 50% reduction in completion time

Do you think AI is a threat to a developer's job?

It's not just about coding skills; it's about how smartly you wield those skills. Embracing AI isn't a threat to your job; it's a guarantee of job security.

What do you think? Share your thoughts and experiences below

[austin-cheney]

In theory AI should be eliminating low skill developer jobs because it can already fully replace those people. By low skill I mean pattern memorizing framework monkeys who fear writing original solutions because doing so is beyond their capabilities.

In practice however the only people being eliminated are more experienced senior developers, because they cost more and their added value is often either not realized or not wanted. This is especially true in web development which has been racing towards commodity copy/paste skills for over a decade. In my case, as well as other cases I have observed, senior developers will continue to do this work for as long as they can until terminated at which point they don’t come back.

This is hastening the problem that was already present. Nobody is training web development as a competency which makes employers more reliant upon tools as a solution, which in turn drives increased complexity to compensate for declining availability of talent. It will be interesting to see what happens when employers stop over paying for candidates sufficiently lacking expected talent and instead turn to AI, which can already perform that low skill work.

[scrapheap]

> The majority of them said that they were able to reduce the time required for building. Imagine a project that used to take three days, is now wrapped up in just one and a half days! Thanks to AI, we're talking about a 50% reduction in completion time

I don't know about other people but the time I spend coding isn't the majority of the time a project takes. Most of the time and effort goes into all the stuff around the coding - investigating the issues being addressed, specifying what a solution would do, testing the solution built, etc.

> Do you think AI is a threat to a developer's job?

I don't think it's a threat to skilled developers. Personally I'm looking forward to a long career fixing up all the code produced by below average developers relying on AIs without any real understanding of the code being produced :D

[MyelinatedT]

Completely agree with this. Problems are solved by people, not code. Code is a tool that can either improve, degrade or leave unchanged the state of a system/service. Plus, code is usually the easiest, quickest bit of the process (perhaps with the exception of some huge legacy monoliths).

LLMs can be useful for improving developer velocity, but the key skills that make good software developers good have yet to be emulated well by AI.

> Personally I'm looking forward to a long career fixing up all the code produced by below average developers relying on AIs

“Looking forward” is a bit of a stretch… :)

[tellarin]

I don't see how it could be a threat.

For the simpler work, AI tools are as much of a threat as frameworks or existing code generators.

For higher level work, AI tools can be great assistants, but they are very far from being able to understanding large existing code bases, inter-dependencies, side-effects, non-functional requirements, etc.

Not to mention that most software development projects fail for non-technical reasons, which code-focused models also can't help with.

[ps256]

Yes.

Developers who know how to use LLMs are some % faster and more productive. You can increase the % enough so that overall demand for developers goes down or doesn't grow as much as it would have otherwise.

It's not "my company laid us all off and replaced us with an LLM" but more like "this year our team is hiring for 3 new people instead of 4" - that's still a significant impact on the job market. And who knows how those numbers will change as LLMs get better.

[iamlucaswolf]

I believe this is a rational assessment in general. A lot of the discussion around this topic seems to be negligent of market dynamics.

However, the crux is in the details:

> You can increase the % enough so that overall demand for developers goes down or doesn't grow as much as it would have otherwise.

I would be at least skeptical of this. Every push for commodification that we've seen in the software space so far has been absorbed by demand. Will this continue forever? Nobody knows. At least where I work the backlog is filled to the brim, and every new iteration of tooling begets more babysitting to unlock the promised gains. And customers still have a never-ending list of hyper-specific feature requests.

The friends and colleagues at the Senior/Staff level who are using Copilot/GPT-4 (and have admittedly become much better than me at prompting) didn't exactly become "hyper-productive". Sure, they get code pushed out faster, but they still work long hours and complain about deadlines.

This is not to say that we're all fine forever and things will not change. But as long as we don't experience an across-the-board temperature shift in the job market decoupled from macro-economic events I wouldn't put too much attention there. In the end, doom scrolling is also just a form of procrastination.

[mejutoco]

That would be true if demand for development in general would remain static. My opinion is that software demand has a lot of room to grow.

If cars become cheaper, car infrastructure gets better people will use more cars, not the same amount. Except I believe there is a lower limit in the amount of cars a person can use vs the amount of automation (through software) a person can use.

[danwee]

> his year our team is hiring for 3 new people instead of 4

Na. We'll be able to do more with less... but the amount of work to be needed will increase, hence more people will be needed as well. Same old story. Compilers didn't get massive people fired.

[ps256]

I don't think anyone can claim certainty on this either way.

My view is that in the past the increases in productivity have come in a period of exponential growth of the software industry and that growth can absorb the additional productivity. But exponential growth doesn't last forever and if you have a period of a declining / flat / slowly growing software field, a significant enough productivity improvement from tools like LLMs can reduce the overall demand for software development.

[retrocryptid]

[citation needed]

[SavageBeast]

Developers do things besides write code naturally. They design, implement, sometimes deploy, and always diagnose and debug. AI can help with some of these of course at about the same expertise level as a very qualified intern/jr. developer.

I can look back at a career full or production bug finding/understanding/fixing and say with high confidence I'm in no danger of being replaced anytime soon.

Consider the dev cycle changing such that a Product Team feeds specs to some AI and delivers code wireframes to development as step 1. If we were going to eliminate developers thats where we'd have to start.

Any of us could try it too. Take the requirements/design artifacts of a feature and feed it to the AI of your choice. Sit back and watch it build the whole feature, integrate it into the product, test it, deploy it and maintain it. We all know that exercise is not going to even get past Step 0.

When product/feature stakeholders/designers are using AI to capture requirements and generate code thats the start of the cycle. The next problem would be now we got this big hairball of AI generated code that to every dev who works on it is "somebody else's (non-running) code - and its not even finished!". Any devs here who have ever got thrown on a project where you get to take ownership of some other teams unfinished, not working codebase? If you have you know how hard Step 2 would be.

Per OP, I think thats about right. There are still gains on the table to be made in leveraging AI too. The whole "AI is gonna take our jobs!!" thing gets pretty silly when you start thinking of it from the perspective of being hired as CTO someplace with your main task being "Leverage AI to eliminate all development staff".

[DeepSeaTortoise]

No,

there was already an "AI-like", but much more superior solution available for decades at this point:

Outsourcing the work to a country with low wages. Even nowadays you can get highly skilled workers for like $50 a month in some African countries.

Pay them $500 a month instead and you've got yourself highly motivated employees, who are extremely eager to learn whatever skills you require.

In the very long run robots and "AI" will of course wipe out 99.9% of the available jobs, but at that point we'll have to think about other solutions on the political level anyways.

It is reasonable to assume any jobs done exclusively in front of computers are very early on the chopping block, but as long as you don't see the business majors or business jurists being sent out the door enmasse, I wouldn't worry too much.

And at that point you'll still have first class logical deduction and math skills and a lot of knowledge about algorithms, which are highly valuable skills in general.

[danwee]

> Outsourcing the work to a country with low wages. Even nowadays you can get highly skilled workers for like $50 a month in some African countries.

But this is not the norm. In order for company X to hire people from a different country, it is necessary that:

- company X hires contractors/freelancers. The vast majority of people out there in IT work as employees, not as contractors

- company X needs a branch in that country. This is rare

- company X hires in that country via an intermediary company. Not as rare, but usually it's not worth it

Not even in Europe countries hire people from other countries that easily. It's more common now after covid, yes, but it's definitely not "let's hire from country X developers 50% cheaper!"

[coolThingsFirst]

>Even nowadays you can get highly skilled workers for like $50 a month in some African countries.

No way. The market is getting normalised and even Indian devs make good salary nowadays. Now there's basically rest of world and US market salaries.

[DeepSeaTortoise]

Check out e.g. Madagascar. I admit, the salaries are higher now than I remember (or maybe that's just because salaries in the capital are higher), but they're still low enough that you'd end up on a pillory if people were to look into one's business adventures there.

And Madagascar is far from the worst country to be stuck in, since the gunmen to "worker" ratio is rather low compared to other African countries. $50 in Somalia would probably get you a whole, very motivated team and a lengthy prison sentence.

[coolThingsFirst]

Even if the software salaries are 250$ per month for devs which is what I read online, I seriously doubt you can find a highly skilled engineer for that price. A highly skilled engineer would only struggle to get the first client on Upwork and earn a lot more or just leave the country.

By highly skilled I mean:

1) Fluent in English

2) Good education - you can't get a great education in isolation it's either from other smart people through online contact or the more likely, face to face(university).

3) Open source projects, ability to pass interview

Even devs would hire a highly skilled dev for 50$.

The whole idea just reeks of americans who haven't left their country.

[retrocryptid]

It turns out that Africa and India are on different continents and do not refer to the same place.

[coolThingsFirst]

It doesn't matter, you can't get a software engineer for 50$ in Africa.

[nittanymount]

my 2c

- it is a good helper, if you know what you need.

- cannot replace devs, most time, even designers/PMs cannot defines the requirement/functions clearly, how could AI get the right code? each product has its own logic, devs need to work out the architecture based on experience/limitations/comprises, don't feel AI could not do these yet ?

- AI/LLM repeats/copies existing codes from somewhere, it does not make new codes, right ?

[iamflimflam1]

Not right now. But in a short time frame <5-10 years I’ve moved to a position of yes. Currently assessing what to do about it.

[Urumar]

Yes, it's really a threat to developers job

[GoldenMonkey]

Developers who do not use AI. Will be replaced by developers who do. Same with all other trades.

[tennisprince]

imo most larger organisations have had a lot of bloat to begin with and there has always been a case for a reduction in headcount for established orgs and so i don't forsee a reduction in developer jobs even though in theory their productivity should be accelerated due to AI.

[milleniall]

Yes, unless you want feelgood copium then no.

[retrocryptid]

It depends. If you're working a "bullshit job," it may be enhanced by AI. As Alan Blackwell points out, ChatGPT is a bullshit generator. But David Graeber documented many people were working "bullshit jobs."

If you're working a bullshit job, maybe ChatGPT will enable unforseen vistas of productivity in your career field.

https://www.cst.cam.ac.uk/blog/afb21/oops-we-automated-bulls...