Do you have the idea that programming is now easy / can just be done by AI?

sock.rocker*

Curious about this after reading some opinions elsewhere and from listening to friends. It sort of feels like programming is being devalued and non-programmers consider it a sort of solved problem.

I am wondering what people's opinions here are.

T-Maxx

It's not just programming- it's revolutionising everything.

SuperBowserWorld

(a)

The tools get better which allow more complex systems and new systems and applications to be developed more quickly. These have tremendous amounts of complexity and tremendous amounts of failure points that require tremendous amounts of effort to keep running.

(b)

Go-to (a)

It's all programming. Maybe less people doing it or maybe more software will be required for new applications requiring more people. And no one doing it whenever AGI can do it itself.

506972617465

AI can help a clueless guy create a simple website, plug in the API and have fun with latest weather forecast. It by any means isn't ready to replace the actual programmers, aka people who know the context and can visualise and imagine the end result. Hell, even giving the correct prompt would require certain degree of knowledge and also - you'd have to know what to do with the answer you get. It can check the code, write the code, but without real brain behind "what is actually to be done" it doesn't mean much.

Long story short - a very helpful tool, good enough but I wouldn't go too crazy with "hurr durr I don't need to learn programming" - you do, because otherwise you won't be able to manage AI to get things done.

Source: my arse 25 years in IT

EDIT: God, it's been actually 31 years, the 90's still seem relatively "recent" and it's over quarter of a century ago… Scary

EDIT2: We got a recent hire who claims to be a guru on a number of things. I did call it BS, and I can feel the AI in 99% of the things they write and do (I even went to the extent of using AI to find the sources they used in their "paper" to claim it as "their work"). Yet, having been put in a situation that's new to them—something they don't know (because why would they?)—they proved to have zero knowledge and zero understanding. You need to know something to be able to even ask AI about it. If you lack this knowledge, you won't get far, and you'll get caught easily.

Example: You need to know that a local Linux firewall exists to ask AI about how to open ports and add rich rules. If you have no idea about, say, RHEL and its features, there's nothing AI can do for you. And then you get caught.

the_amazing_raisin

As a developer I haven't seen much in AI copilots that is massively useful. Yes being able to hand it some code and get it to write all your unit tests does save time, but I don't see why I need an LLM for that and not just some automation scripts

What I can see if project managers using AI to spit out some code based on customer requirements and use that as the code basis for the project

There's 2 ways this can go, one is the smart way where you use the AI tools to build a code skeleton with some basic implementation that your developers build on, test and deliver faster with high quality

The second is the dumb way where the project managers decide they don't need any developers, fire the requirements into an AI tool and just send back the output code with minimal testing

I do think the role of the developer is shifting from writing code to writing AI queries and filling in code. If it streamlines development without sacrificing quality then it's a good thing overall, however I feel like a lot of companies are favouring method 2 instead

I would say the role of the developer has been devalued somewhat. This is partly down to AI tools however I think the main driver is many of the large software companies running out of ideas and are now gutting development teams they don't need anymore

I do believe AI tools are having an effect and accelerating layoffs, however that's not because they can do the job just as good. It's more down to company leaders believing the hype that they can fire all their engineers and it'll all be fine

So the AI tools don't need to do the job as good as a human, they just need to be good enough to convince the CEO that the humans aren't needed anymore

L1011

Try debugging the spaghetti shite code that an "AI" spews out

False sense of productivity as you appear to jump forward then spend twenty times longer fixing the unfamiliar mess that appears

igCorcaigh

https://www.boards.ie/discussion/comment/123651687#Comment_123651687

The developer is still needed to sanity check the AI output of course, but things will improve as more sophisticated AI is built into the dev environment; we already have this.

_ptashek_

https://www.boards.ie/discussion/2058420009/do-you-have-the-idea-that-programming-is-now-easy-can-just-be-done-by-ai

It's just another tool to help increase ones efficient use of time, and has nowhere near the capabilities required to replace humans in the loop. I use it for tons of stuff, and see the nonsense it can spew out but also the quality it can help achieve (great aid for writing!).

Give it 2-3 more years for the hype phase to settle down a little - then we can have a more serious conversation about the future. At the moment it's total lunacy, sprinkled with fairy dust.

Mrs OBumble

https://www.boards.ie/discussion/comment/123651687#Comment_123651687

Every developer who has been hired to do maintenance on an existing system says that about the previous developers work.

irishgeo

Until AI stops making **** up it will still need a human to check it.

eightieschewbaccy

https://www.boards.ie/discussion/comment/123652534#Comment_123652534

I work day to day with AI. Anything particularly difficult that I'm working on if I were to ask an AI, they more often than not hallucinate something that's entirely wrong. While I can identify that it's bullshit, an inexperienced developer could absolutely trust it. Being overly trusting of AIs is pretty naive and that seems to be becoming more and more common.

Giblet

I've been running Kiro on a semi complex spec, with very well written specs, code examples etc.

It can certainly generate a lot of code that seems correct, but it gets stuck on trivial things. If you cannot code, it will generate trivial examples for you and you might be amazed, but production ready Rust code, Microservices, Kafka decoupled handlers, data replication, high availability, docker files, Kubernetes specs etc. It will spin its wheels trying to hook up these things, losing context all the time and either make a mess so you have to do it by hand (throwing away a lot of what it did sometimes) or just hacking away until it works, which requires you to be a senior dev in the first place.

It's been running for 5 days also… which will cost a fortune when it moves out of preview.

So it can help, but it cannot take over just yet.

sock.rocker*

https://www.boards.ie/discussion/comment/123652104#Comment_123652104

I'm more talking about the perception of developers rather than the actual capabilities.

I find it incredibly useful, but probably only keep around 10% of the code it writes. It becomes less and less useful the closer you come to releasing a product.

To be honest, I feel like I got the last helicopter out of Saigon knowing how to program already. At least with my decade plus of coding, and by working with frameworks I already know, I can guide it and spot the nonsense.

sock.rocker*

https://www.boards.ie/discussion/comment/123652959#Comment_123652959

I tried Kiro briefly in the middle of a task and it failed completely, but it did look impressive. I think it would be better for me to try it with something completely new.

hfenton

Like the PC before it, AI is a tool, not a job destroyer. The work it affects will simply evolve into something else, not disappear. Systems admins didn't disappear when things moved en masse to the cloud providers premises, they just evolved into devops and cyber security etc. The technology that disrupts current practice in industry also opens new opportunities, new demand, new jobs products and activities that were not economically feasible or possible before. Software development possibilities are going to explode not recede. The electronic calculator, PC and spreadsheets were supposed to remove the need for accountants etc. The PC was supposed to do away with the need for most office staff etc. Instead, the work just evolved into something else. Calculators and spreadsheets do not do away with the need to learn math and accounting. CAD didn't do away with the need for architects and architectural technicians or their need to know about construction. AI won't do away with the need to learn programming and software development, of which programming is only one part.

silliussoddius

Doesn’t matter what AI can do if the companies selling it are burning cash like an Irish person in 2005.

littlefeet

(1) Someone needs to make a list of all the new developments that were going to lead to mass unemployment and follow what happened.

(2) I received a reply to an email that I knew had been generated by AI while trying to appear it wasn't, a total fail. I did a bit of digging on this, and the company has purchased a system that guarantees the receiver would not be able to know the email had been generated by AI, all I can say I hope they get their money back.

DublinWriter

It's not about how to code, it's about knowing the mechanics of platform you're developing under, i.e. ASP.Net, LAMP, etc.

Having said that, I don't think it's a viable long term career path in IT anymore. Just learn some Dilbert-speak, master some cool transitions in PowerPoint and call yourself an IT Architect.