Hacker News Reader

Comment by paxys | Hacker News Reader

People who think that tools like these will end software engineering as a career are the same ones who buy into the "10x rockstar programmer" myth and think that the job description entails sitting in a room and writing code for 12 hours a day and nothing else.

A software engineer's job will be the same as it always was – to translate unclear and always-changing requirements into something that works. Until ChatGPT, Deepmind or whoever else can learn to deal with my client or product manager, my job is safe.

In fact AI making programmers more productive is a good thing. It is only going to increase the problem space and we'll need yet more engineers to fill it, like all other productivity advancements that came before.

View on HN

Agreed, but the question is: how many other other programmers do you need on your team now? How much of their work will be taken over by our new AI magic boxes?

You have to see this from the perspective of non-tech companies, who see tech as a cost-center, not a profit or innovation center.

Does your regional grocery store that just needs some tools that can help it track inventory really care whether its code comes from a team of big brained humans or two programmers in a basement copy-pasting chatGPT answers?

You can say the same thing for any other advancement in software engineering. Did good IDE tooling reduce the number of programmers? Explosion in open source libraries? Cloud computing?

A single programmer with a laptop can do in minutes today what it took entire companies of hundreds of professionals and a large amount of funding a few decades ago. Yet the size of the industry hasn't shrunk in the same period – quite the opposite in fact. There is more need and demand for programmers today than ever before.

Agree with that as well - there's a good argument to be made that companies will find entirely new product lines and efficiencies once they free their developers from fixing low-level problems and bugs.

It's going to be an exciting few years to say the least.

How much software does your grocery store needs? Don't you think there will be a limit to that?

If cost vs revenue allows for more developers, yes, you can see the happy path. No wonder you observed this at the start of the software revolution. The app stores are already overcrowded.

> Until ChatGPT, Deepmind or whoever else can learn to deal with my client or product manager, my job is safe.

Given the rise of "NoCode", perhaps clients & managers are more willing to meet in the middle than one would otherwise assume.

Given how managers seem to enjoy meetings, I can easily imagine one sitting down with e.g. Dragon to speak with e.g. ChatGPT in order to clarify, disambiguate, and expand requirements list(s) before feeding them to e.g. Alpha Code.

Given that the amount of work to be done remains the same, new tools that increase the efficiency of workers will reduce the numbers of workers needed. Thus, even without machine learning, new tools make your position worth less and less; if nine people can do tomorrow what ten people can today, that is effectively a threat to your job. What keeps your job safe is a seemingly ever-increasing amount of work to be done where old tools can't be applied for increased efficiency (which is why most software engineers will have to keep their skill set constantly up-to-date). For how long will this be the case, though?

Every advancement in software engineering productivity throughout history has done the exact opposite. It increases the overall problem space for the profession and you need more and more programmers to meet the demand.

Yes, I 100% agree. I was sort of nitpicking about what kept your job safe since you seemed to imply that it was due to the shortcomings of Deepmind and ChatGPT. I argued that it is due to keeping your skill set updated, and, as you put it more elegantly than I did, the increased overall problem space.

I do, however, think there will be an inflection point in the coming decades, where the tools become more generalized and better at dealing with new problems. I might also add that the reason for this belief is simply that a lot of work is being put into making these type of generalized tools; but unlike Kurzweil, I don't quite believe it will lead to the Singularity. :)

Hey, if it can negotiate a Verizon bill then it stands a good chance of negotiating a list of features. Maybe you could pair it with a stable diffusion design bot that shows a selection of imagined designs that are based on the client's description of what they want. The client picks an option and the chatbot negotiates the price and timeline for it.

it's about deprofessionalizing it so cheaper people can do a "good-enough" job that you can't charge high prices for your service.

The loom and sewing machine didn't eliminate the seamstress, it just impoverished the profession.

It's like driverless trucks; what you'll actually see is mostly driverless outsourced and remotely monitored trucks where someone is making something like $1/day monitoring 5 trucks at once and switching it to remote control mode when needed.

We've proactively organized our economy to produce these kinds of outcomes. It's not the technology that's the problem, it's the unquestioned assumptions of how we've collectively presumed it will be used.

We just need to replace the product manager and client with AI too ;)

Well we can replace everyone with AI and not even need to exist at all. Until that happens, we'll still need good old fashioned managers and software.

Reply on news.ycombinator.com