I wouldn't be surprised if by making code 5x cheaper to write, aggregate demand for code goes up by at least 5x, and possibly more. The history of computing is full of technologies that made it cheaper to create software (compilers, debuggers, high level languages, version control, increasingly powerful hardware, widely-available open source frameworks, etc...) and as programmers became more efficient and effective, the scope of problems they've been tasked with has kept growing. I think there is a good chance that LLM coding assistance will be a similar story.
Coding Jobs and GPT-4
41–50 of 58 posts
Re: Coding Jobs and GPT-4
#42Yes it absolutely decimate demand for junior coders. My firm has basically frozen hiring, and all existing devs are expected to boost their productivity using one of the all the various AI's or get fired. We expect the economy to get way worse as de-dollarization seems to be looming on the horizon, and preparing now.
Where are future devs experienced with the company's code supposed to come from if not new hires?
We all know GPT is going to improve, the question is how much and when.
Some strategics advisors are saying we might not even need the existing engineers to maintain current levels of productivity.
Re: Coding Jobs and GPT-4
#43Going a bit against the grain here I think but: I think tools like GPT-4 and beyond will increase the value of software engineers for companies as time goes on. Folks building software can be more ambitious with the things they create, and it can become easier to plug in another software engineer to your team if tooling can make them more predictable in their output. I see people here saying that increased efficiency…
I've been casually coding to assist my work for nearly 20 years now in various fields. It's helped me a lot recently when it came to blowing off rust when I was trying to use python.
The value I have found from it has been quite high, I have been more confident about being able to offer ideas, knowing I wouldn't have to spend quite as much time refreshing my language skills for languages i have not used in a while.
I agree with your last point,when I first started playing with chat GPT I thought I'd throw some of my already fixed problems and see how it handled them. It didn't have the best of answers (tripping over problems I also had initially) or wasn't much better than what I had already made.
Re: Coding Jobs and GPT-4
#44Yes it absolutely decimate demand for junior coders. My firm has basically frozen hiring, and all existing devs are expected to boost their productivity using one of the all the various AI's or get fired. We expect the economy to get way worse as de-dollarization seems to be looming on the horizon, and preparing now.
> Yes it absolutely decimate demand for junior coders. My firm has basically frozen hiring What pipelines were you recruiting from? I see AI decimating "become an engineer in 10 weeks bootcamp" types of coders, but proper engineers? The demand is still exceptionally strong, especially now that Twitter and Meta released a lot of talent on the market at the same time. > We expect the economy to get way worse as de-doll…
I hope you are right, this is the first time I've been asked to start research into de-dollarization and factoring that into our business plans.
Re: Coding Jobs and GPT-4
#45No, a coder is responsible. A tool is not.
Re: Coding Jobs and GPT-4
#46Yes it absolutely decimate demand for junior coders. My firm has basically frozen hiring, and all existing devs are expected to boost their productivity using one of the all the various AI's or get fired. We expect the economy to get way worse as de-dollarization seems to be looming on the horizon, and preparing now.
>de-dollarization American social media has been surprisingly loud in their predictions about the total imminent collapse of the US dollar and economy since February 2022. As a European I would like cheaper American products and services caused by a weakened US dollar. But I don't see it happening yet. The US dollar is still strong, American companies are still competitive, and the S&P500 is a lucrative ingredient in…
Maybe but I think France's Macron has spooked a lot people who do global strategy.
That's the only big thing I can think of in the news, that might be why I've been asked to start de-dollarization research and related business research.
Re: Coding Jobs and GPT-4
#47Building software has gotten easier and easier over time, and there's been other step changes in the past. If you understand this history, you can understand what the future might be like.
Before Stackoverflow, software research was rather difficult. You had books that acted like tomes of reference materials. You had maybe some tribal knowledge internally on how to do something. But even in the early 2000s something like MSDN was something you might search on a CD ROM! So the time to get a "basic Windows app" up and running was probably twice as long as it would be today. I'd say there's a similar thing with Github, languages with pkg management, with libraries that do every little thing, whereas when I got started in software even the C std lib might be sketchy and even then its very limited.
When Stackoverflow came out, there was a kind of similar panic that we forget about. Every problem would be solved on stackoverflow, and you just copy-paste everything from stackoverflow. It'll get way too easy to launch an app. Yet the reality is there's even more developers, with higher salaries than everywhere, with software in even more niches in our lives.
The current "panic" about software engineering going away is also happening at the same time as layoffs, so its hard to know how that colors things.
Re: Coding Jobs and GPT-4
#48> will significantly reduce the demand for coders?
Define "coders".
A slightly cynical view: yes, it will reduce the demand for "writers of code in a single language and framework, following the docs and without much other understanding". No, it will not reduce the demand for proper software engineers, who understand the inside-out of the systems they work with, from hardware to kernel to libs and whatever stack they're using, and who will become more productive and will likely be able to deliver more from their huge backlogs.
Explanation: my impression (totally an opinion, but I've heard some people roughly agreeing with me) is that in the last 10-12 years a whole lot of people entered the software development profession attracted by salary, hype, easy money (crypto?), or whatever else, maybe through a dev bootcamp or just by learning on the internet. That's not a problem per se, some of the most skilled devs I know are self taught. But there're a number of developers that, even after a lot of years, have only a limited understanding of how their systems work.
Python programmers that know Django but have no idea what a syscall is, or how a process works in their operating system of choice, or how the Django ORM translates those classes into some SQL. They live in the IDE, somebody else is taking care of the deployment after they push, they just make changes and addition to an existing app. That's it.
Frontend programmers that know some toolkit like React or Angular with a few libraries, that can write some piece of Typescript, but that often have no idea about some Javascript basics, the browser, http. Again, they contribute to an existing project, adding files, changing things, but somebody else is in charge of deployment, and they won't be able to setup a new project from scratch, if needed.
A lot of those people still had/have good salaries, maybe not the best, but the market was tilted in their favour; if top tier talent flew towards big tech, there were positions to be filled in other sectors. But I think that even big tech wasn't immune to the coding bootcamp effect.
It seems to me that when money was cheap and the work to be done was a lot, it made total sense to add whatever kind of help could be found on the market. Now, this could change; if software engineers can move quicker with the help of AIs, follow-the-docs-only programmers may be become far less useful.
Just my 2c.
Re: Coding Jobs and GPT-4
#49A few thoughts based on my experience with GPT4 so far. 1. For now, not by much. It will make us more productive but the world is not short of projects with endless backlogs so huge nobody can ever drain them. Maybe some of those tickets are low value or obsolete, but a lot of them represent real bugs or ideas for improvement that the cost of software development makes prohibitive to reach. 2. It may change the natur…
> So - programmers who just want to play code golf, be ready for that to become more of a hobby than a job. But for programmers who always wished they had more hours in the day to get through all their ideas, it will enable them to create a lot more value and that value will in turn create demand for yet more value-add on top of that. So it should be a virtuous circle, in theory. But to me, this also implies a viciou…
Here's a trick I've been experimenting with lately - try making a GUI in some high level and English-like but formal language. FXML works well (the description language of JavaFX) probably because it's relatively noise-free, but XAML would probably also work.
Then ask GPT-4 to convert it to SwiftUI. It actually does it. Or try Jetpack Compose: same thing.
This leads to an obvious thought - part of why people use heavy abstractions is because the cost of native development is too high. You could pay development teams to write a WPF UI, a Cocoa UI, a GTK UI, an Android UI and a UIKit UI for your app but even VC funded firms flush with cash can't justify that level of extravagance. It's too hard and expensive to hire devs, so when you do, the opportunity cost of them porting stuff between endless UI native APIs just can't make sense. AI changes that equation completely. GPT-4 is cheap for what it does. Suddenly, auto-porting your app into 5 different native APIs simultaneously and then assigning the resulting bugs to a bug-fixing bot doesn't seem unimaginable.
Now, for better or worse, even if AI reaches the point where it can do that, it doesn't mean Teams won't be slow. People use Electron for a lot of reasons, not just cost of duplicated development (skills, libraries, Windows lacking a competitive and non-deprecated UI toolkit etc). But it's interesting to think about ideas that we instinctively write off today due to the massive cost of software developer time.
Re: Coding Jobs and GPT-4
#50As a 'professional coder', I've been very much pondering - and experimentinng - with how best to use AI coding tools. For big, complex, programs I work on (multiple C programs up to 100k LoC) it's hard to frame a question in such a way that it doesn't need to know details of the codebase. For peripheral stuff it's pretty useful as a quick way to get some cut'n'paste code ('write me code to read the load average on li…
> For peripheral stuff it's pretty useful as a quick way to get some cut'n'paste code ('write me code to read the load average on linux and log it to a file every second' - I know exactly how to do that, but it's a faster typer). For public APIs that might be the killer feature; a better way to look at doc (or rather, rely less on doc) if it can look at different code bases and find snippets of code using a popular A…
It definitely works well for that stuff, albeit it tends to be easily confused by APIs that have evolved over time. I find it frequently giving me code that uses APIs that don't exist, and when I point that out, it apologises, tells me that API was removed in version xyz and then gives working code instead.