| ▲ | barrkel 5 days ago |
| [flagged] |
|
| ▲ | tomhow 4 days ago | parent | next [-] |
| Please don't do this here. If a comment seems unfit for HN, please flag it and email us at hn@ycombinator.com so we can have a look. |
| |
| ▲ | barrkel 4 days ago | parent [-] | | I understand where you're coming from, but there's also value in setting norms publicly, especially at times like these, at inflection points. The loud point and stare can have a bigger effect on behavior than the quiet disappearance of a message here and there. | | |
| ▲ | tomhow 4 days ago | parent [-] | | I know, and it's an ongoing and developing issue for us to manage and find ways of addressing, knowing that it's only going to get harder. Our position now (as always) is that the negative consequences of a false accusation outweigh the benefit of a valid accusation. We don't want to be a site that shames people or piles on. We can't really know what people are thinking when they post a comment, or why they might use a tool to generate or improve it. Downvoting and flagging are strong enough signals, and do enough to hide the comment if others agree that's what should happen, and that's what happened here. It's good to email us so we can see if there's a pattern of behaviour we need to know about so we can penalize or ban the account or take some other appropriate action once we can see the full picture. | | |
| ▲ | barrkel 3 days ago | parent [-] | | I understand. However, I interpret your statement as coming from your position: as a moderator, with a job to do as a curator of a community. How does the community learn what to flag? It's not about piling on. You can fix disruptive behaviour by picking off remarkably few individuals, but that behaviour is easy to classify. Everybody knows what a troll looks like. But you can't teach a community what is generally acceptable and unacceptable within the culture of the community without a conversation in public. Consider two forms of justice for a new rule: one day, it becomes illegal to smoke in restaurants. In the first form of justice, in the interest of avoiding pile-ons and vigilantism, secret police turn up at your home in the evening and confiscate your cigarettes, or fire you from your job, or simply disappear you. The right way to deal with smoking isn't to say anything in public; it's to inform your friendly secret police and they'll deal with it. They'll look for a pattern and they'll decide the punishment. The offender gets fixed. The decision isn't public, so it can't be criticized either. If a few visiting foreigners also get abducted because they didn't know what the norm was, well, tough. In the other form, justice must not only be done, but it must be seen to be done. The conversation happens in public. Everyone is empowered to teach everyone else about norms. A friendly word to the folks from out of town let's them know what's what. The community finds its own level of comfort for where the boundaries are. Not everyone is happy, but everybody can see what's going on. Now I'm not saying you're like secret police. You do however have a similar working model. You're not accountable to the community and you want to make decisions in private without any second-guessing. You mean well. But your tools are more suited to dealing with trolls than finding and socially enforcing community norms. I don't envy you. I would not like to be a moderator. Best of luck. | | |
| ▲ | tomhow 3 days ago | parent [-] | | > Now I'm not saying you're like secret police. You do however have a similar working model. You're not accountable to the community... We're nothing at all if not accountable to the community; our number one guiding principle in moderation decisions is to keep earning the trust of the HN community. Things like this can be blown up to seem as sinister as someone wants, including invoking imagery of "secret police" operating covertly. That's never how we've operated. Right from the time he took over HN, dang has gone to great lengths to explain moderation decisions and actions, and I follow his lead on that. There was a case a few weeks ago where I engaged publicly in conversation with the author and other community members about the perception that a post was LLM-generated, and other cases where I've commented in response to clearly-LLM-generated comments, and we'll continue doing that. All we're talking about right now is community members taking it upon themselves to publicly accuse a commenter of LLM-generated comments before we've had a chance to investigate. It falls into the same category as accusing users of "astroturfing, shilling, brigading, foreign agents, and the like" – which is explicitly against the guidelines. As I said in my previous comment, the harm caused by false allegations outweighs the benefit of valid allgations, and the established processes are perfectly adequate to deal with valid cases. |
|
|
|
|
|
| ▲ | ycombiredd 5 days ago | parent | prev [-] |
| I subjected some of my own writing to an "AI-detector" and was surprised to find what a high percentage of AI I must have in my DNA. Joking aside, I have found myself recently using phrases and styles of turning a phrase and stopped myself asking, "since when do you say that?" Since I've spent countless hours talking to LLMs, that's when. I even had a multi-hour conversation with ChatGPT about my idea for "perfectly flawed" lab-grown diamonds, several months ago and was excited to finally find a place to talk about what I learned, hence the comment you attribute to being from an LLM. I'm probably not an LLM, nor did I generate a response for this thread using an LLM, but you've just made me really self-conscious about how now humans, or at least myself, have to be wary that we don't start talking like the LLMs that are supposed to be talking like us. |
| |
| ▲ | barrkel 5 days ago | parent [-] | | If you want to hide the LLM tells better, you need more consistent combing. E.g. "hasn’t", "they’ve", "wouldn’t", "“natural” and “synthetic”" all use curly quotes, but other parts use straight quotes. Somebody tapping away at a keyboard will probably be consistent instead. Another big tell is the total overuse of antithesis. ChatGPT is cringy for this. "Not this, but that". I counted six instances of antithesis in your message. There were also three instances of hypophora, another beloved of essay generators. | | |
| ▲ | joshstrange 4 days ago | parent | next [-] | | The em dash use is also very telling. And before “I’ve been using em dash for a while”, no, they haven’t. This is the _only_ comment they’ve ever made that uses em dash. | |
| ▲ | throw310822 4 days ago | parent | prev | next [-] | | People, not sure you realised, but this thread is gold. Maybe even intentionally? You are debating over a message that imagines the consequences of generating progressively more natural-looking diamonds by adding them imperfections; you are claiming that the message doesn't look natural enough because it doesn't have the imperfections of a typed message (wrong quotes, wrong dashes). Fact is, with diamonds and intelligence we're moving to a Dickian scenario where progressively complicated and ineffective tests are needed to tell the artificial from the natural. And soon the Voight-Kampff test won't work anymore. | |
| ▲ | ycombiredd 4 days ago | parent | prev [-] | | Christ almighty people, even the Notes app on my iPhone (frustratingly, when you're typing something destined to be pasted in a console or, apparently a comment on HN) will "correct" quotes to the squirrelly quotes and my dashes to em dashes. I never want them and when I find them troublesome I will programmatically remove them before pasting, but in this case I didn't think they would bother anything. I didn't think anyone would want to read the 12 pages or so I had written on the topic so I pasted as-is. I'm not using an LLM to write for me. Summarize for brevity at times, sure, but my thoughts are my own. (Punctuation, perhaps not. I've written about my frustration with unwanted "smart" punctuation and the principle of least astonishment elsewhere. It is frustrating.) If I thought you cared enough about my thoughts on synthetic diamond-making (from a non-domain expert) I'd post the entire thing for your reading pleasure, but I gather you're more interested in trying to make me look like I'm using an LLM to write my replies. Here, from my phone, directly into the comment field, I am certain there are no "tells" that the content may have been typed elsewhere and summarized for brevity. Already this is more time talking about something off-topic than I'd hoped for. I was genuinely excited that there was something topical that came up on this obscure synthetic diamonds topic I had written and thought extensively about, and rushed to share what I had hoped would come across as a palatable sized summary of such. (You might be surprised to read me admit that I can be overly loquacious.) I'll shaddup already but this tendency for me to ramble on when something could have been said with far fewer words is exactly what I thought might be a good use case for summarization. Y'know, got a lot to say but trying to be respectful of other people's time and all. <|ENDTEXT> | | |
|
|
