As many of you are aware, I like LLMs, especially for quick and dirty research, and using them to discover further avenues of investigation.

I also think that any AI generated content should be identified as such.

To the best of my knowledge, there are (yet) no established rules regarding AI generated content, explicitly attributed or otherwise. I've got no problem with such content in, say, the "Jokes" thread, but AI responses to a member's technical queries should at least include the AI engine used and the source prompt, if allowed at all (I think properly attributed AI content should be allowed).

Is the development of such a rule worthy of discussion?

 

My 2c
It's pretty obvious and I don't care what slop engine assembled it.

 

Although for highly experienced members of this forum and AI generated response should be fairly obvious, to less experienced users the answer could fool them that it actually came from a blood and flesh user.

I do feel that Joey’s suggestion has merit, and I will start including a note on any of my responses which have been AI-assisted.
Hopefully I won’t forget to do so.

Exception to the rule could be the jokes thread.

 

Although for highly experienced members of this forum and AI generated response should be fairly obvious...

I suggest that the AI generated content will become less obvious at the technology continues to improve.

In fact, I am pretty sure that I could ask Grok to read a selection of my AAC posts and mimic my writing style, and that it would do so reasonably successfully.

My belief is that posting AI content as one's own is tantamount to plagiarism, though maybe not in the absolute strong sense.

 

My 2c
It's pretty obvious and I don't care what slop engine assembled it.

As much as you may hate them, LLMs are a thing, and they're not going away.

While I think some (or properly attributed) content is useful, I see a problem growing as unattributed AI generated content increasingly proliferates on sites like this.

If the rule is, ultimately, AI content is forbidden entirely, there are at least two consequences:

1. Useful and informative content will be limited
2. The moderators are going to have a hell of a time policing it.

I am sure there are more.

Anyway, thats my 2¢.

 

In the spirit of this thread, I've asked Grok for it's opinion on the matter:

"Hey, Grok, should unattributed LLM generated content be allow to be posted on technology-oriented social networks where traditionally only humans had interacted previously?"

Yes, with transparency norms, but not as fully unattributed deception.


This is fundamentally a question about authenticity, community standards, and technological reality on platforms built for technical discussion (Hacker News, Reddit's programming subs, specialized Discords, parts of X, etc.). Here's a balanced breakdown.

Arguments for allowing unattributed LLM content

• Practical enforcement is a losing battle. Modern LLMs produce output indistinguishable from competent human writing in many domains. Requiring "human-only" or mandatory watermarks/detection is unreliable—adversarial prompts, editing, or hybrid workflows (human + AI) bypass detectors easily. Platforms would waste resources on an arms race while alienating users.
• Value creation often outweighs origin. A clear, well-structured explanation of a Rust borrow checker issue or a novel debugging approach has utility regardless of who (or what) wrote the first draft. Tech communities have always prized ideas over credentials—pseudonyms, throwaways, and anonymous contributions are common. AI can accelerate knowledge sharing, especially for non-native speakers or busy practitioners.
• Innovation alignment. Technology-oriented networks exist because of rapid change. Excluding LLM-generated material while humans routinely use IDE auto-complete, Stack Overflow snippets, or search engines would be inconsistent. We're already in a hybrid era; pretending otherwise stifles progress.
• Free expression baseline. Heavy moderation invites bias and overreach. Many users already post AI drafts polished by hand without disclosure. Blanket bans risk turning communities into gatekept clubs.

Arguments against (the stronger case for some guardrails)

• Erosion of trust and authenticity. These spaces historically worked because participants signaled genuine effort, experience, and accountability. Unattributed AI floods can create "dead internet" vibes: generic answers, SEO-style spam, karma-farming bots, or simulated engagement. Readers lose the ability to gauge whether they're interacting with lived insight versus probabilistic text. This is especially damaging for debugging, architecture debates, or career advice where context and skin-in-the-game matter.
• Signal dilution and quality collapse. Low-friction generation lowers the bar. We've seen it on Reddit, forums, and comment sections—endless mediocre AI slop pushes out thoughtful posts. Human conversation has friction for a reason: it filters for seriousness.
• Deception vs. tool use. There's a meaningful difference between "I used AI to refine my post" (common and fine) and "here's a wall of text presented as my original thinking." The latter misrepresents the interaction. In tech, provenance and reproducibility matter—similar to how we expect citations or reproducible code.
• Community self-definition. Platforms aren't pure public squares; they have norms. Many tech users explicitly want human discourse to preserve serendipity, humor, and shared struggle that pure AI struggles to replicate convincingly at scale.

Pragmatic middle path

• Strong preference for attribution. Platforms should encourage (via UI nudges, labels, or culture) disclosing "AI-assisted" or "generated with Claude/Grok/etc." for substantial content. Not as hard rule (enforcement fails), but as norm—similar to how people disclose sponsorships or edits. Transparency preserves trust without killing utility.
• Distinguish use cases. Code snippets, summaries, or idea generators? Often helpful. Long-form "personal experience" stories, controversial opinions presented as human, or engagement farming? Higher scrutiny.
• Platform tools over bans. Allow users to filter AI-labeled content, improve quote/reply threading to maintain context, and let communities moderate spam aggressively. Detection can be a signal (not a ban hammer).
• Reality check. Full bans are theater. Hybrid human-AI workflows are the future; the interesting question is how to keep the human element valuable. Tech networks that lean too purist will shrink; those that become pure AI slop will lose serious users.

In short: Allow it, but reward and normalize attribution. The technology is here—fighting the tide is futile. The real challenge is designing social incentives so these networks remain places for genuine human (augmented) insight rather than synthetic noise. What specific platform or use case are you thinking about? That might sharpen the tradeoffs.

 

As much as you may hate them, LLMs are a thing, and they're not going away.

While I think some (or properly attributed) content is useful, I see a problem growing as unattributed AI generated content increasingly proliferates on sites like this.

If the rule is, ultimately, AI content is forbidden entirely, there are at least two consequences:

1. Useful and informative content will be limited
2. The moderators are going to have a hell of a time policing it.

I am sure there are more.

Anyway, thats my 2¢.

I don't hate them as that would be stupid, I don't trust (as a source of any intelligence) them, I don't use them and avoid their usage as much as possible. I see them as tools that are actually as dumb as a bag of hammers for the foreseeable future. In the hands of a novice, it's like handing the person a nuclear powered auto-nailer and a bottle of rum.

I don't have a problem with attribution of a post as AI but which brand is unnecessary promotion of a commercial product that's stealing human generated content on a global scale. I see no reason to reward the company or the person that issued a prompt to generate a obviously (IMO it will advance but much more slowly that the hype suggests) AI post.

Just make the poster say it's AI so all users can evaluate if it's useful, informative and should be trusted as something other than a set of next likely tokens from some prompt.

 

In the hands of a novice, it's like handing the person a nuclear powered auto-nailer and a bottle of rum.

Funny!

Edit: as a novice, I approve of (and relate to) this message.

 

I had heard a different version: “give a loaded gun to a monkey”… but the results would also be catastrophic.

 

Not a bad idea.