What's your take on Ai?

1

I’ve been staying up coding until sunrise. I even put in extra work to transition from front-end to full-stack web development. Finding a job has been really difficult, and while I’m still searching, I’ve been doing freelance work.

I’ve only been a full-stack developer for a short time, so I know I’m not as skilled or experienced as someone who has been working on the backend for years.

Three days ago, I was sitting in university cafeteria with some friends, and we were discussing an upcoming project. One of my friends suggested using AI—which, to be fair, I probably would have used too. But it had been a long time since I’d really thought about the whole “AI replacing developers” discussion, and it got me thinking.

I’ve been stressed about making enough money, and honestly, I genuinely love and enjoy coding, even when I’m stuck trying to solve a difficult problem. Of course, I use AI for research and learning instead of spending time searching through Google.

The feeling of investing so much time, energy, and passion into something, only to start wondering what AI might mean for its future, has been really frustrating. It feels like a threat to something I care deeply about and makes me question whether pursuing this career is worth it. I’m not talking about vibe coding or using AI to speed up development. I mean AI that can independently write and build software.

My friend mentioned Lovable, the AI web app builder, so I got curious and tried it. It generated a clean, standard portfolio layout—nothing groundbreaking—but it made me wonder: what if, in 20 years, AI replaces front-end developers? What about someone like me? I’m still a junior developer. I only recently started learning backend development, and I’m 21 years old. Should I be worried? Should I stop pursuing this path?

As a developer, what’s your perspective on where AI is today and where it might be in the coming years?:thinking:

do you think YouTube videos of this year, where developers claim AI will never replace developers, are actually accurate?

2

Welcome to the forum @rl_orion

Entire floors of typing pools were replaced by computers. The typists basically spent the whole day typing the same letters over and over again for different people. Same with people who used to make garments and carpets by hand.

Machines took over their jobs. Automation shifts the nature of human work.

Word processors, personal computers, and email transformed roles into administrative, data entry, and secretarial roles where workers manage digital data instead of paper.

Handloom weavers, spinners, and carpet makers could spend weeks creating textiles. The industrial revolution introduced spinning machines, power looms, and automated textile machines. Production speeds increased, costs decreased, and while many jobs were lost, new industries and roles emerged around machine operation, factory maintenance, and global logistics.

Automation removes certain tasks, but reallocates human labour toward higher value work, which requires problem solving and creativity.

A lot of the entry level work used to be boilerplate stuff. With AI the ground floor of what is entry level is going up. So more identifying and solving actual problems, rather than working on layout or components.

I think at the moment, AI is hype about slop. Some experienced and skilled users can make great things. But most people using AI are literally creating 10 bugs for each one they solve. These bugs don’t show up for weeks or months, by which time the person who created the PR or feature has already forgotten about it, since they didn’t write any of the code.

Others are using AI like a conductor, but don’t really have a good grasp of the nitty gritty details. It’s like a line of AI models communicating with the next one, with a person at one end of the line. This can create hallucination cascades. They have to “trust” the models are doing what other models told them to do. There is no way to audit the system. Only to see the output and match it against expectations. Which is basically how an LLM works anyway. This leads to structural flaws, security vulnerabilities, and architectural drift.

Happy coding

3

I avoid AI where possible, for multiple reasons:

  1. Environmental damage: it takes a lot of water for cooling, and AI datacentres have been known to produce noise pollution so severe that it has had deleterious effects on those living near the datacentres. Also, the heat off them affects the environment as well.

  2. Energy usage: because AI datacentres require a lot of power, it pushes up energy costs.

  3. Most LLMs are trained on stolen data and IP. In many cases, people were not asked before their work – writing, artwork, music, code, whatever else was scraped from the internet – was hoovered up by big AI companies. I’m not OK with that.

  4. Owing to the above, you cannot guarantee that the AI’s output is free from proprietary or stolen work. On a practical level, it means you cannot copyright the AI’s output. But it also endangers the FOSS (Free Open Source Software) community that relies on the likes of the GNU GPL and other copyleft licenses that allow, for instance, the reuse and building upon of code, for example to fork a piece of software and tweak it so it can run on a different platform. Freedom of use – including portability – is one of the many freedoms enshrined by FOSS. Because of these licenses, if a piece of FOSS software is found to have proprietary code contained within it, it could bring legal trouble upon FOSS developers.

  5. Increased hardware prices have also been a result of AI. Memory prices have surged, and so too has the cost of hard drives, and other components. This could make it harder for ordinary people to get access to computers or the internet in a world that widely relies upon the internet for common services.

  6. The surge in AI-made pull requests, and bug reports. These have made maintainers’ lives more difficult by increasing their workload. Oftentimes, the bug reports end up being duplicates and duplicates of duplicates. It also creates strain on the services used for hosting repositories.

  7. Tokens, which as I understand it, are basically AI companies asking people to pay for access to the AI and the information contained within. It subverts the very nature of the internet, which was designed to democratise learning and access to information.

  8. AIs are often wrong. They’re little more than statistical prediction machines that use probability to guess what next word they should use. They also ‘make up’ things that sound plausible but fall apart upon investigation, e.g., use in law, where AI has been seen fabricating entire cases or precedent when used for drafting legal arguments or documents.

  9. Skill atrophy. There is research that suggests that reliance on AI can lead to skills declining because said skills are not being used, instead just taking the AI at its word.

  10. … I grew up with Clippy, man. I learned about Eliza in school. Chatbots and various ‘AI’ projects have been around for a very long time. AI is little more than a buzzword.

So, yeah, that’s.. all of my thoughts. ^^;

4

Thanks @Teller !:blush: That’s a good point. Do you think learning the MERN stack is still a worthwhile investment if I’m thinking 20–30 years ahead? Or would you focus on broader software engineering skills instead?

5

that was inspirational, thanks @fiodora :grin:

6

both, learning a stack is experience that will benefit you in the long run, even if you later don’t use that stack, because it’s about learning about how the technogies fit together and communicate with each other

7

Twenty to thirty years is a long time.

Fortran is over 70 years old, but most programming languages share similar syntax and data structures.

In the future though, communication will become the most important skill. But it will rely on understanding data architecture, systems, and logic.

The choice of which set of languages to use will depend on the problem you are trying to solve. Each language has strengths, weaknesses, and constraints. If AI can write the code for you, you still need to decide which languages and frameworks to use.

Happy coding

8

And here I thought I was just being a luddite. =P

9

true, thank you for your guidance!

10

Lmao nah you’re not alone