Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: What is your plan if AI ends coding as a profession in 2-3 years?
14 points by atleastoptimal 3 months ago | hide | past | favorite | 40 comments
Amazon Web Services CEO Matt Garman claims that in 2 years coding by humans won't really be a thing, and it will all be done by networks of AI's who are far smarter, cheaper, and more reliable than human coders.

I know a lot of people on this site insist that humans will never be replaced, but let's just say that this could happen. Are you doing anything right now with the assumption that may be a possibility? Do you fear, if you are a coder, losing your relevance in the industry?




A self-serving claim like all of the others about AI from people heavily invested in promoting it.

Currently “AI” can do maybe 2% of what a reasonably skilled programmer can do. I doubt the horizon sits just a few years away.

I’d like to give you a real answer, but I can’t take the claims and hype seriously. I have the same plan for the AI coding apocalypse as I have for a giant asteroid strike or the rapture: no plan. If it actually happens I will adapt, until then it doesn’t warrant worrying about. I have survived a few supposedly existential threats to my programming career over the last several decades, this one looks even less likely to come true.


I have to remind myself a lot of programmers these days have never seen a 'down' market, and get truly worried about the 'end of programmers'...

You have to be pretty long in the tooth now to remember .com crash, off-shoring, hell - even the 2008 stuff ended like 14 years ago.

Thats not even mentioning cobol, 4th GL langs, RAD, low-code, no-code and god know what else was supposed to put us all out of work.

This too shall pass... :-)


In the business since the late ‘70s. Two crashes, two AI winters, offshoring scares, 4GLs… seen it all, never lived up to the hype. jobs come back because software needs expand like gas.

Even if AI takes over most new development in two years we have 50+ years of legacy code to work on.

A decade before my career started COBOL got hyped as an end-user programming language. Then SQL would let managers bypass programmers. Then spreadsheets. Did Visicalc and Excel kill off the jobs? All of those things just created more demand for programmers.


Software expands like gas and most offshore software shops don’t give a flying f about quality, pad their hours and oversell programmers to multiple clients.


My plan is trivial, because I'm old and retired, so I just keep doing more of the same.

I'm here writing this because your premise is wrong. Greenfield programming is a lot easier with AI, if you're willing to ride herd on it, and correct the errors. The profession of programming has little to do with actually generating code, and far, far more to do with the overall structure, and coping with all of the geometrically growing number of corner cases.

Imagining the corner cases, and coming up with ways to cover them, a priori, is what programmers actually get paid for. If AI gets to that point, nothing except actual physical labor is safe.

Someone has to ride herd on the AI if it is to be useful. You can be that someone.


First of all AI might take most of the coding jobs in 2 years where those jobs are filled with people who write copy/paste framework type unoriginal code. AI is already writing better code in those areas. AI will not however write new original applications though.

That is kind of scary to think about because most of the web apps will be written by bots, and probably really shitty. Web apps have always been a race to the bottom but now it will be a sprint. The upside though, is that those of us capable of writing web apps without boilerplate bullshit might once again be employable. I guess it just depends on whether businesses are willing to pay for applications up front or would rather buy commodities and pay down the road on maintenance, fees, fines, lawsuits, and such.

In the mean time my back up plans are as follows:

1. Retain my current job writing transmission/data APIs for large enterprises.

2. Military. I can at any time switch my part time job to a full time job.

3. Analytics. AI might at some point take half of analytics but the second half will just become an echo chamber if taken by bots.


Focus on solutions. Coding is a tool for solving a problem, so the fix is to move up the food chain. Look for ways to use AI to solve a human problem.

Think in terms of spreadsheets. In the past, it took someone, a very specialized person, to do a full breakdown of a company's finances. Now, a relative math novice can do the same thing in a fraction of the time. Now they don't have to spend time solving 1+3. Now they can solve other more complex problems.

I'm one that believes that coding will change but it won't go away any time soon.


“I solve business problems, sometimes with code.”

https://www.kalzumeus.com/2011/10/28/dont-call-yourself-a-pr...


(2011)

It feels pretty outdated these days. All the roles are hyper specialized and engineers are a dime a dozen, and the whole profit vs cost center thing is pretty obscured by fashion trends instead (mobile, Next, crypto, AI) and big cyclical bubbles. I think we've long moved past the stage where technical ability alone is enough to get you hired. There's a glut of us now.


Don't miss the forest for the trees. Sure some of the buzzwords might be slightly outdated but the advice is timeless. A golden rule to remember is that people are hired to save money or add profit. If you have not or it is not the case anymore then your job is in danger.


Well, I get that part of the argument, but I'm not sure I fully believe it. It's not so much the semantics I'm concerned about, but the idea that you should try to get a role in the money-making parts of the business... is that really what drives hiring/firing decisions? Over the last decade or so, at least, it feels like most of those hiring/layoff decisions were driven by aspirational future valuations instead. When money was cheap, they hired everybody. When money got expensive, they laid everyone off. I don't know that the sales people survived any better than the devs, did they?

Seems like the macro staffing trends are more driven by market conditions than a reasoned accounting of staffing ROIs, whereas the micro decisions (who on the team to let go) is more a political and maybe seniority or performance thing.

I'm not a manager though, so maybe that's just how it works up higher and I don't see it. But from the peasant's eye view, at least, it doesn't seem like that's how those decisions are being made...


> I know a lot of people on this site insist that humans will never be replaced, but let's just say that this could happen. Are you doing anything right now with the assumption that may be a possibility? >

I don't. But even if it would, there's enough in the bank account to get through a (although maybe not each and every) re-skilling.

I see shortages of qualified workers left and right, but I think I would seek a trade where you could reasonably become self-employed without too much capital investment. Maybe carpenter/joiner.

The fields of medical assitance (nursing, physician assistance, operating theater assitance, or paramedic) would also be something I could see myself in, but there you are basically always an employee under bad conditions. Or maybe a trade in that direction like prosthetist/orthotist, but I don't know enough about the job market there.

Also teaching would be something I would consider, but I'm not sure if I could sustain the necessary studies financially (as this is rather regulated in my country, it would depend on how much of my bachelor's degree would be transferable)


If Ai doesn't work out there's going to be such a shortage of engineers in the next 5 years that we'll be asking $5,000 an hour.


Thanks for the chuckle :-)

My 'retirement plan' is to be the last living COBOL coder :-D $5K/hr should be about right :-D


This is certainly a thing. I worked along side people that used to be part of Sequent that made the Dynix PTX NumaQ OS and hardware. Sequent was purchased by IBM for a lot of money to destroy it. This left a lot of companies and government installations of Dynix unsupported. People could either pay IBM a massive ransom to continue support or hire one of the dozen or so former Sequent employees. I was part of a company that did the latter. I don't know exactly how much they were paid but they seemed super happy to assist at any hour for any issue regardless of how trivial it may be. One of them would assist us from his boat. I enjoyed working with these people. Super friendly, smart, it was a rare treat. So do you have a boat or remote retirement retreat with internet? :-)


Ohh - that sounds awesome! If I get to that point, I'll buy a boat ! :-)


> let's just say that this could happen.

This seems to be of great concern to you, but let me ask you this: How prepared for civil war are you? Or the zombie apocalypse? Do you have a bunker and a bug-out bag ready to go? If AI gets that good at programming, it'll be as good in other industries, and there's gonna be some upheaval if it gets to that point suddenly, in a bad way. Nevermind coding, what happens when all jobs can be done by robots? If you're worried about having a backup plan for coding if a hypothetical AI programmer job apocalypse comes, what's your backup plan for civilization if an all job apocalypse comes?

Which is really to say: Keep calm and carry on. Your anxiety isn't going to make the problem any better. Be prepared, certainly, but don't worry yourself into a curled up mess in a fetal position in the corner.


My current data engineering job is already less than 10% coding anyway. It's mostly talking and drawing diagrams.

I suppose if an AI will do all the work, I will probably be the one who uses the AI. Sure as hell the business people are not going to be the ones leveraging AI to build stuff.


My grandma worked as an accounting calculator (literally summed up piles of numeric trash) and when computers surfaced, they all were replaced. The same will happen with coders eventually.

But accounting is still a thing, it only lost that stupid “sum up a column” part. If you’re not only a coder, you’ll be fine. If all you can do is calling apis and recalling algorithms, you’re screwed.

AI will eventually try to replace everything in 20 years, including Matt Ceo. Make sure to stay an engineer cause at that point people who can shoot will realize people who own AI can not, but people who can shoot need engineers to run the thing.


If AI ends coding as a profession, it will have probably ended all professions, in which case there's no amount of planning that can save you from the economic catastrophe that we willfully walked ourselves into


I'm starting a second job as a solar installer next week. It pays about half as much as I currently make, but I'm looking forward to being in the real world again. Tech can be real lonely, especially remote, and after twenty years I think I'd like to try something different.

Was going to school for a bit but ran out of money, so it's back to the working class for me. As a 40 year old, I don't think I have a future in tech, or anywhere really lol.


You've less than 2 years to get out of Tech

https://www.indiatoday.in/technology/news/story/openai-japan...


Two years? In two years AIs will be smarter and more reliable than human coders? I don't buy that at all. (Cheaper, sure, I can see that. It's probably already true.)

To answer the question, though: I'm 62. If that happens, I just retire. That doesn't help everyone else, though...


You'll be useful for something. If nobody needs to code anymore, some enterprising person will come up with a financially lucrative way to utilize all the ex-developers. Will it pay as much? Maybe, maybe not. Will you survive. Yep!


Coding represents only a small part of a programmer's work. When tackling complex problems, AI still has limitations. Strengthening problem-solving skills remains essential for programmers.


What are the core limitations that AI has that humans don't? What would happen if the state of the art AI in 2-3 years has none of those limitations?


Then I'll figure it out then. No sense planning for things that probably won't happen.


Such hand wringing has happened for hundreds of years. Yes, technology has regularly caused short term shifts in people's employment. But the human world has one important element, humans. It exists for their benefit only. There are few people doing hand weaving now, but machine woven clothes are a net benefit. In my long career I shifted roles a number of times. Some people will need to do that.


I have a lot of software projects I could finally get written.


> but let's just say that this could happen

For Amazon Web Services.

> What is your plan

Move off of AWS.


Someone else wrote: your job isn't going to be taken by AI, it's going to be taken by someone using AI.


Deleting AI code


Society is 3 meals away from a revolution.

Automate everything and what you will get is a massive riot.


Here's why AI won't end the profession... (at least not completely).

Programming is defined as the art of creating/writing instructions for computers to perform. If you're aware, then you will realize that for AI to do anything, it must be instructed/prompted it to do so first. In other words, telling an AI to do something, is essentially programming, and that there is no world where AI will magically do something unless someone instructed it to do so.

That said, what AI has basically done, is made it possible to instruct computers using the syntax of our native language, instead of a more primitive syntax. This is immensely helpful, however... realize that the syntax of our native language is also problematic in ways, particularly in its preciseness (if you don't understand this, just look at the legal field, and why legal jargon is even a thing). And so, you'll eventually realize that it can sometimes be difficult to get a computer/AI to do exactly what you want it to do, no matter the syntax you decide to use.

All said, AI has made it easier than ever to program, but it doesn't remove having to do at least some sort of work of formulating instructions. That reality will always remain, and so long as there remains some demand for that work to be done, the profession will continue to exist.


Ai, in its current form, is very far from replacing humans. It's another tool and hasn't been even that good recently and not a cheap one either. I don't doubt that Ai will replace humans at some points in the future; I am just saying that it's very far from being the next 2-3 years.

The people invested in this AI hype have bought their own bullshit. They don't exactly feel aligned with developers, so their idea is to throw hundreds of billions at compute and hope it solves all their technical problems. This is not going to work and could potentially blow up in their face pretty badly.

If you are from the US, there are two existential threats to your industry/paycheck:

1. Offshoring. This time it might be real because the global south has improved considerably when it comes to language (English) and education. The people there are relatively skilled and might be able to execute. A few jurisdictions/countries are actually competent.

2. Big Tech blowing up in the US. Due to speculation, over-investment in AI, or mismanagement (ie: hiring lots of people and then firing a lot of people).

This happened before with cars, happened again with cars/evs (china) and I don't see why it can't happen again this time with tech.


https://github.com/dylanaraps perhaps?

I don't think AI will be able to do that, at least not what we are calling AI right now. I mean, for fucks sake, humans are still doing it and we can't even figure out how to use the exponential increase in compute to make computing faster, we still can't figure out how to build foundations such that a program will still run as is in 20 years. Something finished still to this day needs maintenance. I doubt an LLM is going to improve the state of software, so there's still a ton of work we need to do.


AI coding is a sort-of thing, but will also never become a 'thing'.

Maybe in the giant software houses, sure, generating code around constrained, prescribed requirements prepared by moneyed teams incentivized to feed the machine dense, specific, meticulously-prepared requirements is/will be a thing, but I have such a hard time envisioning a world where a guy who starts a regional brewery and a couple restaurants, who has ever-shifting day-to-day challenges with personnel + product segmentation, who has immediate, unexpected business requirements (not pure TECH requirements, mind you), who needs somebody they can trust to offer technical insight and guidance without wasting their time with how the sausage is made, I just struggle to imagine that that dude will fire his 'guy' and then just sit down and magically craft a perfect prompt or unleash an all-knowing AGI who can navigate the irrational nuances of humans who have money and are desperate to spend.

As a technologist, I'm literally boiling over with thoughts about AI / AGI, how to achieve it, etc. That said, as a human, I would be a fool to ignore the fact that PEOPLE can't even understand each other's intent or message with any accuracy. Get a guy from the Everglades and a guy from Mauritania together and you better believe there will be misunderstandings. Now imagine any non-SV non-elite-American and an LLM...

There's just no way that it'll do what you want unless you can ask it to do EXACTLY what you want, in the parlance of the predominant culture / colloquialisms that the model was trained on.

Who knows with AGI, but it doesn't exist, so I guess we'll cross that bridge when it appears. In the meantime, I promise you, all anybody wants is somebody they're comfortable with, who they can trust to execute even a shadow of what they themselves are on the hook for delivering to whomever they're beholden to.

AI can't make a client feel good about you. You aren't even "you" if you're an AI -- you're just a tool. Somebody who hires you to build a house doesn't believe in the hammer, they believe in the person wielding the hammer. They believe in the vision of what they will eventually receive. You have to have kind eyes and nuance and a soul to deliver that kind of value, and frankly, a computer program that can play on your emotions and manipulate you is not a computer program that you can trust -- it's just another human -- and courts exist because humans can't trust each other!

Who on Earth wants that as their (automated, unpredictable) hired help??


become a farmer


Have you seriously considered this, or is it a joke? I tried my hand at farming a little bit, and it's not easy at all. Lots of hard back breaking work, but also interesting opportunities for automation at scale. The small family farmers and ranchers I knew had incredible work ethics, seemed happy, but had basically no work life balance. It's a whole lifestyle...


Nuclear engineering, OnlyFans, maybe both

I mean either way I'll be spending a lot of time indoors risking my life with a bunch of guys around




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: