HACKER Q&A
📣 precompute

Is “AI” being over hyped to paper over incompetence in IT?


LLMs and other media-generators ("AI") are egregores of the output by competent people in the last three decades. Seeing how, even with faster computers, software is getting worse and less reliable, is LLM usage seen as a hedge against diminishing standards of the software engineering discipline?


  👤 kypro Accepted Answer ✓
> Seeing how, even with faster computers, software is getting worse and less reliable

How old are you? I don't remember the last time I saw a BSOD. I don't remember the last time I download a game and it didn't just work. I don't remember the last time I brought a piece of hardware and couldn't get it to run on my computer.

I think you're just wrong on this. It's honestly amazing how reliable modern software is despite being orders of magnitude more complex in many cases.

But to answer your question, no. I don't think recent developments in AI are being overhyped. It seems we've finally advanced natural language models to a point where they are now competent enough to be useful in a highly general way, and they will only continue to get more useful as the technology continues to progress. The world is about to change very quickly and in ways that are hard to imagine.


👤 ofalkaed
Software is getting worse because there is a hell of a lot more of it being written and much of it is written for a quick buck. The hype around AI is because it is a big a step forward and the media is thrilled to have something new to talk about.

👤 logicalmonster
> software is getting worse and less reliable

I agree that software feels worse today, but that might be because our standards are getting much higher.

Compare and contrast.

If you made a website that let you upload a low resolution jpeg for sharing say 20 years ago, that was seen as a jaw-dropping "wow" technology.

Today, you need to build a unique 3D interface that tracks the exact location of your pizza being delivered to your door in real-time to barely start to raise half an eyebrow about the technology.

> is LLM usage seen as a hedge against diminishing standards of the software engineering discipline?

AI's popularity is an interesting topic, but I think there's largely 2 factors to consider.

1) The media is pushing AI because it's a profitable story. We have a generation of people who grew up watching Terminator. We have a current bad economy that has people fearful about losing what employment they have. In this environment, the novelty of AI is a click fest for the media.

2) AIs might be initially popular because the UI is easy to understand. It's essentially just text, so anybody can understand what is happening. But maybe you need specialized interfaces for more specific tasks. So who knows what will actually happen in real usage for important things? I think it's a fun toy though.


👤 muzani
"software is getting worse and less reliable"

This is a bold claim. If anything, it looks like software is more reliable, just harder to maintain. For example, you can build most use cases with just PHP, but not active async social media and marketplace apps.

Some are built on top of unstable middle layers (e.g. plugins) but those tend to get replaced.

If anything, I think there's a tendency towards "write once, compile everywhere". With AI metaprogramming being the upcoming phase of this. I see AI similar to software vs hardware. You can build better control systems in software than hardware. Similarly AI enables more complex software. But you don't want it completely replacing the lower level.


👤 seydor
not worse, but bloated. Human programmers are clearly lazy. Bring in the robots