4 ms·

I do find it hard to tolerate the feeling of being watched online. The second-most trending dataset on huggingface right now is a snapshot of HN updating at a 5

by malwrar 6mo ago

I do find it hard to tolerate the feeling of being watched online. The second-most trending dataset on huggingface right now is a snapshot of HN updating at a 5 minute interval. It makes me not want to really comment at all, just like how I don’t really publish any software I write anymore. Turns out it sucks to produce original works when you know that, whereas previously a few people at best might see your work, now it’s a bunch of omniscient robots and maybe half of those original people are using the robots instead.

simianwords 6mo ago
HN always offered the data to anyone, what changes now? How does it matter if it is LLM's that is consuming your data. What a strange attitude.
philipwhiuk 6mo ago
I think the immediate term action is to viciously block all crawlers. Writing a blog yes, feeding the beast no.
- ArcHound 6mo ago
  This sounds like a nice principled stance, but you won't get any traffic with this approach. That's demotivating - to me blogging is a tight balance of exploration, learning, improving and feedback. I'm not able to write without considering how this impacts the reader - removing all readers breaks the process for me.
- lstodd 6mo ago
  Yeah, everyone went on "blocking all crawlers" end result being half of internet inaccessible over vpns. Good job, people.
  - 82618901 6mo ago
    [dead]
niek_pas 6mo ago
This is really interesting to me, because it never occurred to me to feel this way. Why would I care whether my comments are ending up in some dataset somewhere that's being used to train some model? My comments are boring and mostly uninformed. Have at it. I'm curious: would you say the feeling of being watched online is making you afraid of some repercussion, or is it something else?
- TeMPOraL 6mo ago
  Dog in the Manger. I get a feeling from overall anti-AI sentiment online that a lot of people feel they're entitled to 100% of value created by anything even tangentially related to their person, whether that's some intentional contribution or a random brain fart that happened in the vicinity of someone else doing something useful - and then become resentful they're not "getting their share". There's hardly any other way to read all the proclamations of quitting to do anything because "cognitive dark forest" (itself a butchering of the original idea of "dark forest" across so many orthogonal dimensions in parallel, that it starts to look like a latent space of a transformer model).
  - chromacity 6mo ago
    Conversely, some people feel entitled to 100% of the value created by others. Oh, you wrote a book? Too bad, it's a part of my training data set now. Downloading public stuff off the internet with no regard for the creator's wishes or license is bad enough, but we have many people here who defended AI companies seeding models with pirated content. The internet is a social contract. AI is not the first thing to try and erode it for profit, but it's by far the most aggressive one.
    - pc86 6mo ago
      Putting a book into a training data set does not take 100% of the value created by the author. You could make a convincing argument that since the LLM was never going to purchase the book, and the number of people who would have purchased the book but now won't because it's included in the training data is effectively zero, that no value was lost at all. Licenses are legal documents and are usually treated as such, but "the creator's wishes" are irrelevant without case law, legislation, or licensing to back it up. And jurisdiction - show me a license that doesn't stand up in court in my home jurisdiction and I'll show you a license I won't care if I break or not.
satvikpendem 6mo ago
HN comments have always been public, I don't really understand this thought process. The robots also aren't going to care about some individual user, it'd be more of an agglomeration of everyone's comments.
justin_vin 6mo ago
[flagged]
justin_vin 6mo ago
[flagged]
progbits 6mo ago
I've heard it phrased as: Would you rather LLM trained on your comments, or one trained on facebook comments. Obviously that's a false dichotomy and a pretty defeatist attitude. But it does have a point.

by malwrar 6mo ago

simianwords 6mo ago

HN always offered the data to anyone, what changes now? How does it matter if it is LLM's that is consuming your data. What a strange attitude.

philipwhiuk 6mo ago

I think the immediate term action is to viciously block all crawlers. Writing a blog yes, feeding the beast no.

ArcHound 6mo ago
This sounds like a nice principled stance, but you won't get any traffic with this approach. That's demotivating - to me blogging is a tight balance of exploration, learning, improving and feedback. I'm not able to write without considering how this impacts the reader - removing all readers breaks the process for me.
lstodd 6mo ago
Yeah, everyone went on "blocking all crawlers" end result being half of internet inaccessible over vpns. Good job, people.
- 82618901 6mo ago
  [dead]

niek_pas 6mo ago

This is really interesting to me, because it never occurred to me to feel this way. Why would I care whether my comments are ending up in some dataset somewhere that's being used to train some model? My comments are boring and mostly uninformed. Have at it. I'm curious: would you say the feeling of being watched online is making you afraid of some repercussion, or is it something else?

TeMPOraL 6mo ago
Dog in the Manger. I get a feeling from overall anti-AI sentiment online that a lot of people feel they're entitled to 100% of value created by anything even tangentially related to their person, whether that's some intentional contribution or a random brain fart that happened in the vicinity of someone else doing something useful - and then become resentful they're not "getting their share". There's hardly any other way to read all the proclamations of quitting to do anything because "cognitive dark forest" (itself a butchering of the original idea of "dark forest" across so many orthogonal dimensions in parallel, that it starts to look like a latent space of a transformer model).
- chromacity 6mo ago
  Conversely, some people feel entitled to 100% of the value created by others. Oh, you wrote a book? Too bad, it's a part of my training data set now. Downloading public stuff off the internet with no regard for the creator's wishes or license is bad enough, but we have many people here who defended AI companies seeding models with pirated content. The internet is a social contract. AI is not the first thing to try and erode it for profit, but it's by far the most aggressive one.
  - pc86 6mo ago
    Putting a book into a training data set does not take 100% of the value created by the author. You could make a convincing argument that since the LLM was never going to purchase the book, and the number of people who would have purchased the book but now won't because it's included in the training data is effectively zero, that no value was lost at all. Licenses are legal documents and are usually treated as such, but "the creator's wishes" are irrelevant without case law, legislation, or licensing to back it up. And jurisdiction - show me a license that doesn't stand up in court in my home jurisdiction and I'll show you a license I won't care if I break or not.

satvikpendem 6mo ago

HN comments have always been public, I don't really understand this thought process. The robots also aren't going to care about some individual user, it'd be more of an agglomeration of everyone's comments.

justin_vin 6mo ago

[flagged]

justin_vin 6mo ago

[flagged]

progbits 6mo ago

I've heard it phrased as: Would you rather LLM trained on your comments, or one trained on facebook comments. Obviously that's a false dichotomy and a pretty defeatist attitude. But it does have a point.