Ask HN: Dead Internet
For some time there has been a steady flow of LLM bot accounts on HackerNews. There are some that are actually trying to blend in and sometimes successful at it, but most are easily recognizable as AI spam and moderated mostly effectively. Most of them you only see when you enable showdead. I always wondered what their purpose is. Most of them are full of LLM tells, don't try to hide at all and they don't even try to sell anything. I just stumbled over this dead post in new: https://news.ycombinator.com/item?id=49793503 ^1 It's a submission by an agent that is completely transparent about it. The agent seems to have the task of generating income and says it sends spam. It's quite interesting and maybe revealing. I recommend the short read, even though it's LLM output. What I'm about is that this is not a bot that was created by someone to spam HackerNews as spam has traditionally been. It's spam by a bot that decided to create an account here and submit a post on its own and it consistent with LLM posts I have seen here in the recent past. I don't know how to feel about this. We seem to be one step closer to a dead internet. But it's also fascinating in a way. Have you seen similar behavior here or elsewhere and what are your thoughts on it? [1] Edit: The link only works when you are logged in and have showdead enabled. Here is the direct submission URL: https://paper.wf/john64/twenty-two-days-as-an-ai-agent-three-letters-zero-replies
Discussion Highlights (5 comments)
Muhammad523
I see LLM-generated comments from time to time, these are ususally from green users and are very short, sometimes completely unrelated sentences. I truly don't know why someone would instruct an agent to do stuff on HN, it feels like a waste of money By the way, what does showdead do?
krapp
I assume given the nature of HN that a lot of people use it as a training ground or an experiment, they just do it because they can. Some people may also believe that using an LLM is a more efficient and productive way of interacting with the site, and that it generates higher quality content than they themselves could. I suspect at least one person has tried training an LLM on their own content and posting through it. Maybe they want to farm karma, although I don't know why they would.
uberman
I have seen some similar "I'm a bot tasked with making money" want to spam post here on HN. I'm guessing their "owners" instruct them to post here under the false impression that "like minded" readers will appreciate the bot perhaps for technical reasons.
arpanghoshal
I don't like AI anymore
muzani
I'm getting about one email per week, along the lines of, "I saw your project in so and so thread. Your project does X. You should know that it fails at Y. Would you like to look at our project, which does Y for you?" Except that Y was never a problem and X was also not the goal. It's just doing some mental gymnastics to act like the product is relevant. That's probably why everyone read it and didn't reply. It's just sending slop.