Are Programmers Facing Extinction in 5 Years? 🕵️♂️
Hey programmers! This is a prediction by Emad Mostaque, Stability AI CEO, and moreover he claims that already 41% of code on Github is AI-generated.
Though that seems very high, AI code has already a strong presence and I’m using myself various LLMs to boost my coding skills.
But he might be right: the job of a programmer will never be the same. Github Copilot is already a standard tool, and just this month Google announced its own version called Duet AI. Jet Brains also came with its own AI-powered IDE.
The bold prediction by Emad Mostaque, CEO of Stability AI, that there will be no human programmers in five years has sparked considerable debate and discussion in the tech community. This provocative statement is based on the rapid advancements in artificial intelligence (AI), particularly in the field of generative AI systems like ChatGPT.
Mostaque’s argument hinges on the transformative potential of AI in the programming domain. He cites data from GitHub indicating that a significant portion of the code (41%) on the platform is AI-generated, underscoring the growing influence of AI in coding. This claim points to a future where AI could potentially automate many programming tasks, potentially reshaping the role of human programmers.
Stability AI, the company led by Mostaque, is actively engaged in various AI projects, ranging from language models to image generation and DNA analysis. These efforts aim to build a comprehensive AI infrastructure that could revolutionize how we interact with technology. Mostaque envisions AI as a tool that augments human capabilities rather than replacing them, suggesting a future where AI and humans work in tandem to achieve more significant outcomes.
However, this viewpoint is not universally accepted. Other perspectives in the tech community argue that while AI tools like GitHub Copilot, Google’s Duet AI, and AI-powered IDEs from JetBrains are becoming more prevalent, they are not likely to completely replace human programmers. Human programmers possess skills in problem-solving, creativity, critical thinking, and domain expertise that are currently beyond the reach of AI tools. Therefore, rather than replacing programmers, AI is more likely to assist them, enhancing code quality and easing their workload.
In summary, while AI is undoubtedly transforming the field of programming, the idea of it completely replacing human programmers in the near future remains a topic of debate. The more likely scenario, as suggested by both industry experts and AI systems themselves, is a hybrid future where AI tools support and enhance human programming efforts, rather than outright replacing them.
This whole AI and programming thing is a hot topic. What do YOU think? Will AI be our partner or replace us altogether? Share your thoughts in the comments below!
Just remember, AI is a tool, and like any tool, it's up to us to use it wisely!
Top comments (88)
There will be more people using tools they don't really understand generating more weird problems that would require skilled computer engineers to identify and solve. I am not that much worried about future of programming itself.
great comment! 👏
Oh man, I see people coming out of college these days who primarily use AI. Imho, I think using AI stagnates my curiosity and desire to learn, so I personally don't use AI at all. I end up fixing other people's problems that AI creates
I use it quite often. For example one of my latest questions: "How to normalize JavaScript date to midnight?" Why should I remember the method by heart when it is just this away from asking? My job is to put this trivial knowledge into a good use in an actual aplication that is helping someone with something. I believe AI will mostly fall behind in business analysis and in transfering customer's chatotic and sometimes unspoken needs into relevant design and working code.
@aloisseckar I absolutely agree
@aloisseckar What do you think about Devin, the first AI software engineer? youtube.com/watch?v=fjHtjT7GO1c
In the demo it produces a peice of software that should not have been made.
You can just ask an Ai... or Google for that data.
That was the demo, so the best case scenario is something that shouldn't exist. Something with no market differentiation, low complexity, and looks bland.
Thank you for the great post! 👏
I do not take it serious, we should always remember that ppl behind AI need investors so they claim anything that sound controversial to raise speculations. 😱
First, we should reconsider what is real programmer job? Writing code? Not really..
Our job is to develop solutions for the problems, writing code is only implementation of the solution.
41% AI generated code is not any argument, because what is that code? Boiler code? Probably yes... because here AI is pretty good.
Searching for the solution is much more than writing the code. You need to understand business, talk to other ppl, present solution and gather feedback etc.
AI will be better and better at writing the code, but eventually you need programmer (human) to create solution and accept AI implementation.
AI is copilot it will always be only our copilot. We are pilots! 🫰
Agree 100%
Absolutely bro !! I think so too.
Hmm Thanks a bunch for checking out the post! 👏
Absolutely, your perspective makes a lot of sense. 🤔 It's true, writing code is just one part of the whole picture. The real magic happens when we dive into problem-solving, understanding businesses, communication, and the whole shebang. 🌐
Your analogy of AI as a copilot is spot on! 🛩️ It's a fantastic tool in our toolkit, but the true pilot, the problem solver, that's us – the human programmers! 🚀✨
What's your take on how the role of programmers might evolve with the rise of AI? Share your thoughts! 🤔💬
In my opinion we will move on higher level, developing solution and providing prompt with requirements to AI copilot.
Work with AI will be for sure incremental to adjust implementation to our needs. I imagine this like something similar to traditional code review.
So that's why I still believe we need to understand what's going on underhood. We are responsible for the product!
Seeking for analogy:
As a plane 🛩️ pilot you can use autopilot, but you are still responsible for the plane! You have to be able to take steer on your own!
I disagree, and think that programmers can (and should) be replaced at some point. I'm not convinced this is the point yet though.
Thank you for your response 🙏
Could you elaborate more? Do you mean programmer will evolve to some other role?
I mean that perhaps I'm not sure of the timescale, but put it this way: in 2000 years, when civilisation is either long gone or in some kind of Star Trek utopia, there won't be people typing code into computers. The status isn't going to remain quo, so to speak.
So at some point programming as we know it will stop being something that people do. It's the same with any non-vocational activity that's not manual labour.
Try 100 years instead. Civilisation still probably exists. Programming as it is now will look as archaic as punched cards do to us today. If there is an equivalent, it will be so different as to render the similarities meaningless.
Drop it to 50 years... well, that's probably the same as 100. So I think that in 25-50 years at the most we'll have seen the last of what we know as "programming". A handful of enthusiasts might still do it in the same way people still write games for the 2600, but that's about it.
Same thing happen during industrial revolution. Manual workers were replaced by machines, but demand for engineer has never ceased to increase.
It's indeed a noteworthy point you've raised. Feeling Sad for our fellow developers with lower skills is important as the landscape evolves. While AI can assist in generating code, it's crucial for developers to grasp what's happening under the hood
Someone experienced with a good collection of snippets in their IDE/editor has greater efficiency as someone without experience who augments their coding skills with a machine learning model. 5 years will not make too much of a difference. The failed quest for self-driving cars have shown the hubris of AI researchers. I don't think self-programming IDEs will fare any better.
In 1987 this was the direction of AI research: Programmer's Apprentice and the snippets were called “Clichés”.
Hey new dev student here. Any tips or resources on building a collection of snippets?
And you make a good point
Some advice: your best bet is to just build things. They don't have to be big projects, just build stuff, solve problems. Sometimes the best projects are ones that solve a personal problem you're facing.
Have fun, be kind to yourself and others and stay curious 🙌
There are usually snippet collections for most IDEs and languages available. Take them as a starting point and mix and match whatever becomes most helpful to you.
When you're in the field for twenty years this kind of statement doesn't get you. I've heard so many things that just didn't hold up at the time and still doesn't, but some CEOs and managers love to threaten peoples jobs and simplifying things for the sake of publicity. Examples:
Now we're here with this new bold statement that's generating free publicity.
Totally. Good luck making sure the program does what they wanted it to do!
No wonder all the prediction of replacing programmers are coming from AI corporate owners
They say the same words like it is repeated sentence with same arguments also we have been in this for almost 2 years now and still the keep saying in 5 years
They always think of programmers as just people who write lines and that is it
I love this comment.
This sums up a lot of my thoughts...
"AI tools are already automating large parts of the software engineering role – but are we sleepwalking into a future where no one knows the basics?"
Generative AI and the widening software developer knowledge gap | LeadDev
For me, the joy in programming actually comes from the mental work of figuring out the problems and converting those thought processes into functioning code. TBH, having most of that done for me by generative AI would take a lot of that enjoyment away.
I've used TabNine almost since it was released (wayyyy before copilot) and I like it. However, its just like a really smart autocomplete and will not write large chunks of code for you based on prompts. I don't want that at all.
Absolutely get that! 😊 The thrill of programming often lies in tackling challenges and crafting code from your own thoughts. It's like a mental workout, right?
It's cool that you've been rocking TabNine! 👏 Having a smart autocomplete is handy without overshadowing the joy of hands-on coding. Cheers to keeping that creative spark alive! 🚀💻
There's an old challenge with solutions and dropping humans too soon. I've seen this in the real world at least twice I can remember.
We've built a system that can answer 90% of customer inquiries, so we don't need humans. Unfortunately, 90% of customers have one of the 10% unanswered during critical stages of the customer journey - such as during onboarding, renewal or just before they quit - leading to disaster. Or we find that the 90% of inquiries normally occur alongside the 10% but outnumber them etc etc.
AI writes most of my API calling boilerplate for APIs I don't know these days. AI writes weird custom sort functions I probably wouldn't have bothered with until a later stage of a project. I use AI extensively now for tasks in my solution that would take humans a long time and AI is reasonably good at. AI writes quite good initial Unit Tests that save a bunch of keystrokes. I find all of that quite amazing really.
Within a few years, I'm guessing I'll be writing less code still, and if we get anywhere near an AGI, with some kind of huge context or access to that context then I'll end up being its muse. That's also fine if it finally comes: my grasp of the internal combustion engine's operation and the need for every lever, join and pulley is limited - doesn't stop me getting where I want to go by the route I want to take.
No major company that made state-of-the-art Steam Shovels survived the introduction of the Backhoe. Things change, jobs change, and the world changes. The change to society, with a reduced need for a workforce, is going to drive another revolution. If you have no way to legitimately struggle up from the gutter, then enough people together can change the world - so we are definitely at a point where it looks like the entire way the world operates may change again - as it did with the change from feudalism. How many years away? That's hard to predict - more than 4? Less than 50? In my kids' lifetimes, I reckon. Maybe in mine.
In my opinion I think the answer is simple really and I don't remember where I heard it.
AI will not replace you as a programmer, but a programmer who knows how to best utilize AI will.
that's in the near future at least.
It depend on what you mean by programming, for me as a physicist its always been about building systems, ask chatgpt to build code for an optical step generator for laser to burn a diffraction grating in a fiber, I wouldn't know how far you would get, perhaps it would help you in the mundane details but programming the micro-controllers for the system I am pretty sure you need the human element mainly due to the diversity of the components and the need to calibrate for the response functions of the analogue components like non linear crystals, extrapolate this in the physics domain, would you want AI writing the controls systems for nuclear reactors? I am sure AI can handle the javascript for a web application but thats a small subset of programming where there happens to be a lot of training data they can steal. Sadly AI is coming for a lot of people before programmers, hopefully the project managers will be the first up against the wall.