Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

It's insulting to read AI-generated blog posts

blog.pabloecortez.com

Post reply on HN
Print view View original

431–440 of 565 posts

Re: It's insulting to read AI-generated blog posts

#431

Post by a_cool_username »

JohnFen wrote: post #326
ok_dad wrote: post #211

Earlier quoted context omitted.

> Code review is one of the places where experience is transferred. Almost nobody uses it for that today, unfortunately, and code reviews in both directions are probably where the vast majority of learning software development comes from. I learned nearly zilch in my first 5 years as a software dev at crappy startups, then I learned more about software development in 6 months when a new team actually took the time to…

I agree. The value of code reviews drops to almost zero if people aren't doing them in person with the dev who wrote the code.

I (and my team) work remote and don't quite agree with this. I work very hard to provide deep, thoughtful code review, especially to the more junior engineers. I try to cover style, the "why" of style choices, how to think about testing, and how I think about problem solving. I'm happy to get on a video call or chat thread about it, but it's rarely necessary. And I think that's worked out well. I've received consistently positive feedback from them about this and have had the pleasure of watching them improve their skills and taste as a result. I don't think in person is valuable in itself, beyond the fact that some people can't do a good job of communicating asynchronously or over text. Which is a skills issue for them, frankly.

Sometimes a PR either merits limited input or the situation doesn't merit a thorough and thoughtful review, and in those cases a simple "lgtm" is acceptable. But I don't think that diminishes the value of thoughtful non-in-person code review.

Re: It's insulting to read AI-generated blog posts

#432

Post by sings »

strbean wrote: post #311
alyxya wrote: post #54

I personally don’t think I care if a blog post is AI generated or not. The only thing that matters to me is the content. I use ChatGPT to learn about a variety of different things, so if someone came up with an interesting set of prompts and follow ups and shared a summary of the research ChatGPT did, it could be meaningful content to me. > No, don't use it to fix your grammar, or for translations, or for whatever el…

I just despise the trend of commenting "I asked ChatGPT about this and this is what it said:". It's like getting an unsolicited text with a "Let Me Google That For You" link. Yes, we can all ask ChatGPT about the thing. We don't need you to do it for us.

What is remarkable is the frequency with which I’ve heard so-called subject matter experts do this on podcasts. It seems to me a very effective way to communicate your lack of any such expertise.

Re: It's insulting to read AI-generated blog posts

#433

Post by ummonk »

collingreen wrote: post #294
ab_io wrote: post #206

Earlier quoted context omitted.

100%. My team started using graphite.dev, which provides AI generated PR descriptions that are so bloated with useless content that I've learned to just ignore them. The issue is they are doing a kind of reverse inference from the code changes to a human-readable description, which doesn't actually capture the intent behind the changes.

I tell my team that the diff already perfectly describes what changed. The commits and PR are to convey WHY and in what context and what we learned (or should look out for). Putting the "what" in the thing meant for the "why" is using the tools incorrectly.

Does the PR description not end up in the commit history after merge? A description of what changed is very useful when browsing through git logs.

Re: It's insulting to read AI-generated blog posts

#434

Post by ssivark »

stephendause wrote: post #375
ManuelKiessling wrote: post #312

Earlier quoted context omitted.

Why have the LLMs „learned“ to write PRs (and other stuff) this way? This style was definitely not mainstream on Github (or Reddit) pre-LLMs, was it? It’s strange how AI style is so easy to spot. If LLMs just follow the style that they encountered most frequently during training, wouldn’t that mean that their style would be especially hard to spot?

This is total speculation, but my guess is that human reviewers of AI-written text (whether code or natural language) are more likely to think that the text with emoji check marks, or dart-targets, or whatever, are correct. (My understanding is that many of these models are fine-tuned using humans who manually review their outputs.) In other words, LLMs were inadvertently trained to seem correct, and a little message…

I suspect that this happens to be desired by the segment most enamored with LLMs today, and the two are co-evolving. I’ve seen discussions about how LM arena benchmarks might be nudging models in this direction.

Re: It's insulting to read AI-generated blog posts

#435

Post by SAI_Peregrinus »

kyleee wrote: post #336
collingreen wrote: post #294

Earlier quoted context omitted.

I tell my team that the diff already perfectly describes what changed. The commits and PR are to convey WHY and in what context and what we learned (or should look out for). Putting the "what" in the thing meant for the "why" is using the tools incorrectly.

Yes, that’s the hard thing about having a “what changed” section in the PR template. I agree with you, but generally put a very condensed summary of what changed to fulfill the PR template expectations. Not the worst compromise

My template:

1. What is this change supposed to do?

2. Why is this change needed?

3. How was it tested?

4. Is there anything else reviewers should know?

5. Link to issue:

There's no "What changed?" because that's the diff. Explain your intent, why you think it's a good idea, how you know you accomplished your intent, and any future work needed or other concerns noticed while making the change. PR descriptions suffer from the same problem as code comments by beginners: they often just describe the "what" when that's obvious from the code, when the "why" is what's needed. So try very hard to avoid doing that.

Re: It's insulting to read AI-generated blog posts

#436

Post by alt187 »

dingnuts wrote: post #396
oceanplexian wrote: post #350

Earlier quoted context omitted.

LLMs write things in a certain style because that's how the base models are fine tuned before being given to the public. It's not because they can't write PRs indistinguishable from humans, or can't write code without Emojis. It's because they don't want to freak out the general public so they have essentially poisoned the models to stave off regulation a little bit longer.

this is WILD speculation without a citation. it would be a fascinating comment if you had one! but without? sounds like bullshit to me...

This sounds like the most plausible explanation to me. Occam's razor, remember it!

Re: It's insulting to read AI-generated blog posts

#437

Post by SamPatt »

oceanplexian wrote: post #350
ManuelKiessling wrote: post #312

Earlier quoted context omitted.

Why have the LLMs „learned“ to write PRs (and other stuff) this way? This style was definitely not mainstream on Github (or Reddit) pre-LLMs, was it? It’s strange how AI style is so easy to spot. If LLMs just follow the style that they encountered most frequently during training, wouldn’t that mean that their style would be especially hard to spot?

LLMs write things in a certain style because that's how the base models are fine tuned before being given to the public. It's not because they can't write PRs indistinguishable from humans, or can't write code without Emojis. It's because they don't want to freak out the general public so they have essentially poisoned the models to stave off regulation a little bit longer.

I doubt this. I've done AI annotation work on the big models. Part of my job was comparing two model outputs and rating which is better, and using detailed criteria to explain why it's better. The HF part.

That's a lot of expensive work they're doing, and ignoring, if they're just later poisoning the models!

Re: It's insulting to read AI-generated blog posts

#438

Post by pasteldream »

_smfg wrote: post #399
pasteldream wrote: post #376

> people are far kinder than you may think Not everyone has this same experience of the world. People are harsh, and how much grace they give you has more to do with who you are than what you say. That aside, the worst problem with LLM-generated text isn’t that it’s less human, it’s that (by default) it’s full of filler, including excessive repetition and contrived analogies.

> Not everyone has this same experience of the world. People are harsh, and how much grace they give you has more to do with who you are than what you say. You okay friend?

Yes.

Re: It's insulting to read AI-generated blog posts

#439

Post by fho »

ManuelKiessling wrote: post #312
jihadjihad wrote: post #23

It's similarly insulting to read your AI-generated pull request. If I see another "dart-on-target" emoji... You're telling me I need to use 100% of my brain, reasoning power, and time to go over your code, but you didn't feel the need to hold yourself to the same standard?

Why have the LLMs „learned“ to write PRs (and other stuff) this way? This style was definitely not mainstream on Github (or Reddit) pre-LLMs, was it? It’s strange how AI style is so easy to spot. If LLMs just follow the style that they encountered most frequently during training, wouldn’t that mean that their style would be especially hard to spot?

Don't Github have emoji reactions? I would assume that those tie "PR" and "needs emojis" closely together.

Re: It's insulting to read AI-generated blog posts

#440

Post by palmotea »

stephendause wrote: post #375
ManuelKiessling wrote: post #312

Earlier quoted context omitted.

Why have the LLMs „learned“ to write PRs (and other stuff) this way? This style was definitely not mainstream on Github (or Reddit) pre-LLMs, was it? It’s strange how AI style is so easy to spot. If LLMs just follow the style that they encountered most frequently during training, wouldn’t that mean that their style would be especially hard to spot?

This is total speculation, but my guess is that human reviewers of AI-written text (whether code or natural language) are more likely to think that the text with emoji check marks, or dart-targets, or whatever, are correct. (My understanding is that many of these models are fine-tuned using humans who manually review their outputs.) In other words, LLMs were inadvertently trained to seem correct, and a little message…

My guess is they were trained on other text from other contexts (e.g. ones where people actually use emojis naturally) and it transferred into the PR context, somehow.

Or someone made a call that emoji-infested text is "friendlier" and tuned the model to be "friendlier."

Post reply on HN