Hacker News Reader

Comment by abtinf | Hacker News Reader

It doesn’t mention the number one most important reason: the output is absolute garbage.

I can spot AI writing very quickly now, after just a few sentences or paragraphs. It became a lot easier to spot after I tried to use it in my own writing.

Calling it “slop” is far too generous.

If you know what you want to say, you might think to yourself “I’ll have this write an outline or a first draft that I will then thoroughly edit.”

And every time, what you’ll find is that the LLM output is fundamentally unusable. Points a subtly missing. Points are subtly repeated. Points are miscategorized. Points don’t make sense at all. Points don’t flow in a logical order.

If you try to use an LLM and you don’t know what you want to say, then it’s hopeless. You absolutely will not see the defects. If anyone who knows the subjects reads it, they will instantly know you are a lying piece of shit.

View on HN

> I can spot AI writing very quickly now, after just a few sentences or paragraphs.

Not denying this is true — but like a lot of what we've seen with AI, lets see how you feel in two years time when the models have improved as much.

I think it was actually Brian Eno that said it (essentially): whatever you laugh about with regard to LLMs today, watch out, because next year that funny thing they did will no longer be present.

> like a lot of what we've seen with AI, lets see how you feel in two years time when the models have improved as much

People have been saying this for years now though

And I think every year they have improved? I can say I am certainly much more impressed now than I was a year ago.

I have been testing them out in my day to day workflow every so often and I remain just as unimpressed as I was when I first tried copilot a year and a half or so ago

I don't think they are any better at coding than they were. I think people are lowering their standards to what LLMs can accomplish, not that LLMs have risen to meet our standards

I don't think the AI companies are systematically working to make their models sound more human. They're working to make them better at specific tasks, but the writing styles are, if anything, even more strange as they advance.

Comparing base and instruction-tuned models, the base models are vaguely human in style, while instruction-tuned models systematically prefer certain types of grammar and style features. (For example, GPT-4o loves participial clauses and nominalizations.) https://arxiv.org/abs/2410.16107

When I've looked at more recent models like o3, there are other style shifts. The newer OpenAI models increasingly use bold, bulleted lists, and headings -- much more than, say, GPT-3.5 did.

So you get what you optimize for. OpenAI wants short, punchy, bulleted answers that sound authoritative, and that's what they get. But that's not how humans write, and so it'll remain easy to spot AI writing.

That's interesting. I had not heard that. I wonder if making them sound more human and making them better at specific tasks though are mutually exclusive. (Or if perhaps making them sound more human is in fact also a valid task.)

I like that Brian Eno quote. If I recall correctly, he was also referring to nostalgia. Like, once the technology improves, you begin to miss the old rough edges. I know that I love seeing old images of Google DeepDream, for example.* It's the same reason why young people miss Playstation 2 blocky graphics, or why photographers sometimes edit their images for unreal Kodachrome color. The things annoy us today are the very things that we'll miss the most.

* https://en.wikipedia.org/wiki/DeepDream

you can check this post: https://news.ycombinator.com/item?id=44598695

>the output is absolute garbage.

You can recognize garbage, not AI.

AI produced garbage have taken the place of human produced garbage.

Saying AI is useless shit because advertisement, spam and the slop industry have put it to work is the same as saying all Human writing is shit because you got a penis enlargement spam email written by a human and you could recognize it as spam email. Even an AI will recognize the flaw in your thinking.

Reply on news.ycombinator.com