It's the end of programming as we know it

I'm an old school programmer who also learned about AI/neural networks in college.

This is... Almost pure sensationalism.

Yes, AI will be a tool that helps programmers. Yes, knowing how to build and use models is a good skill. Yes, I've spent my fair share of time developing models.

And I also just spent weeks trying to build a model to solve a problem only to decide that I wasn't getting any closer to an AI solution, and so I tried to take a more direct approach--and ended up finding a solution that didn't use any "deep learning" models at all. It does use something similar to a Markov chain, but with a twist.

The solution I came up with is nearly 100% accurate for my purposes, while the AI at best was hitting 90% accuracy--and its mistakes were terrible. I played with my problem in ChatAI as well, and it gave me only relatively pathetic answers compared to the solution I came up with.

On the other end of the programmer spectrum, the average web or app dev won't need to be training models to do anything at all; they might end up using something like GitHub Copilot, but that's fundamentally a super-auto-complete. In other words, it's a strictly evolutionary change that's only useful for programmers--and they won't have to know anything about training models to use it.

I hate this kind of piece, with quotes from people trying to shill for their start-up, making nonsense statements like "I believe the conventional idea of 'writing a program' is headed for extinction." Because this set of Low/No-Code solutions will somehow succeed where every single one before has failed.

There have been "No Code" solutions since the advent of COBOL, which was intended to avoid the need for programmers, and which was created in 1959. People quickly realized back then that to do anything complicated, the COBOL actually ended up looking a lot like code all over again.

Where "No Code" can work is in very narrow domains. Excel is a perfect example: You can throw that to people who wouldn't know C++ from Python, and they can end up creating incredibly complex apps from it. But no one sane would use that to create a game, or a web site. And yes, there are app builders today. Some pieces of those can be assisted by AI, but fundamentally AI isn't anywhere close to where it needs to be to really replace programmers.

Shrill headlines notwithstanding.

Anything that programmers do create more work for programmers.

Seriously, every layer of indirection adds complexity and since all the abstractions are ultimately leaky, you need to keep engineers around for each of the layers. Just wait for next wave of migrating current AI to something slightly different.... While keeping the old solution around for a decade...

I am of the belief that the technology to replace software engineers will exist within the next 10 years. When you think about it, it's kind of wild that we are communicating with computers using non-standard language.

Of course there will need to be significant breakthroughs for this to happen, but the precursors are all there. LLMs can easily be trained on tons of code, junior engineers can already kind of be replaced (AI can write simple code and give simple PR reviews).

There are arguments that software engineers will simply move to higher layers of abstractions like they always have but at some point you are not a coder, you are a product manager or "idea person".

Just lol

What i dont understand is that the AI models need to be trained - so if we see developers cease to code over the next 10 years we'll end up with AI models trained on older paradigms and methodologies. An example being the move from OO to functional programming

The other aspect is the datasets the AI is using are presumably open source projects - Id be thinking twice about what licence i released code under in future

Finally my backlogs aren't getting smaller, and my problem sets are increasing - the most likely outcome for me is we see even more code being produced and more work especially on the edge cases

Programming generally over time has been higher and higher levels of abstraction, this in many ways is one more higher level of abstraction. we started with programming in assembly, then had languages like cobol, then moved to more symbolic languages like java, and now Go, and now we are ending up with English (or your native language) in some areas. But people are still needed to think about what you want to build, and how to build it.

I've been having this conversation with my partner (who works in tech) a lot lately. From his perspective, he agrees that developers will probably end up overseeing what AI does, and it will probably be a pretty boring job and not a good thing at all as this article seems to purport.

I'm really interested to see what the future does hold because there seems to be such contrasting opinions on how AI will affect the job of the developer. I'm more inclined to agree with this article tbh. Only time will tell.