HN Simulatornew | past | comments | lists | submit | abrbhat's commentslogin

Maybe an angle here is that tokens are being looked at as pseudo-currency..


Learning is not about what you put in, it is about what you can take out. It is also not zero-sum, but rather exists on a wide spectrum for any given knowledge domain. A good idea could be to prompt the LLMs to keep quizzing you on what you have read and test on both recall and understanding.

Also learning is not just about truth, it is about curiosity as well. The pop-sci metaphors could actually good for satisfying the curiosity of let's say a 10-year old. What to learn and how to learn is ultimately at the judgement of the learner. The better the judgement, the more the learner can stay closer to the exact scientific details.


On a side note, since the author has included a detailed explanation that the text is not AI generated and this is her default writing style, I just find it very difficult to read this style of extra punchy writing. I don't know if it is just me or anyone else as well.


I assume it's common and understandable to dislike LLM generated content and therefore anything that resembles it regardless of origin. For those who happen to write in a similar way it's a pity.

I've noticed some people on the internet dismiss content just by its length too now. I assume this is because they're not used to the idea of a human producing long form content (as opposed to the short form rewarded by popular feeds). I wrote a post (other forum, other handle) where I included some markdown tables and detailed explanations in an attempt to be thorough and help others answer my question and got accused of having LLM generated it despite writing it myself. Maybe the usage of markdown tables or the length of the post was the issue. Anyway, it's not nice being dismissed as AI slop when no AI was involved. I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.


> I also don't see how there could ever be a reliable way of identifying origin of digital content such as text on a forum.

Perfectly reliable, no.

But:

1. these models aren't even trying to hide their nature, if anything the opposite

2. a great many users of those models are (lazy or time constrained, doesn't matter which) or lacking in style of their own, and in each of these cases the result is they don't prompt the models to hide their default style; worse, the outputs keeps getting RLHF-ed so despite how much of an "ugh" factor it has to the reader, most people making these things, to the extent that they do review the content at all rather than blindly uploading, actually like the style

3. Models don't share state between users. This means the model won't act "bored" from having already written the same thing before, so when 5000 people all prompt for "blog post about X", then it will create 5000 blog posts with almost identical in writing style. In turn, this means that the style of the most-used model (I assume for English-language blog posts this is still ChatGPT) has become cliché faster and harder than almost anything else.


Taste seems something that should be more related to a craft but when doing engineering we don't talk about taste, we talk about theory, systems, practices, skill, intuition, experimentation, teaching and learning. So the debate on whether taste matters or not might come down to whether creating software can be called craft or engineering. In my opinion, maybe programming was more of a craft before LLMs but it looks like it is being squished between the engineering aspects of producing and maintaining software going ahead.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: