Hacker News Reader

Comment by aetherspawn | Hacker News Reader

Nevermind this, it took me around 10 minutes to realise that Dosubot was an AI.

Well done. It seemed like a very natural conversation.

Your comment is a bit worrying, since precisely zero humans I know in real life write like that, especially on Github issues, and the first sentence was enough to trip my mental "AI detector".

Yeah, that really off-putting mixture of slavish subservience, patronizing familiarity, and appeasing corpo-speak should give it away immediately to anyone who has read their share of LLM output.

> off-putting mixture of slavish subservience, patronizing familiarity, and appeasing corpo-speak

This is why chatbots suck for customer support.

It doesn't help that customer support humans have been trained to behave like that before chatbots became common. LLMs take it to the next level, but there's been a few times where I found out the entity on the other end of the conversation was actually a human, but the responses were entirely unhuman and unhelpful.

Well, it is called dosubot, it has a "bot" flair, the message starts with a robot emoji, and its writing style is very recognizably that of a "helpful AI assistant". Plus it links to https://dosu.dev/ at the bottom.

I am not convinced this isn‘t some kind of state of the art trolling.

Your HN profile only exacerbates the issue.

It seems you're mentioning Hacker News (HN) and suggesting there's an issue related to a profile there. How can I help you with this? Are you looking for advice on managing your HN profile, or is there a specific problem you're facing that you'd like assistance with?

Dang personally wrote me an email telling me to stop claiming to be an LLM, or rather, asking if I was and telling me to stop if so.

Back then I saw it as some kind of private art project to write controversial comments that looked like LLM output. Discussions under my comments regularly devolved into fights over whether I am one. Some would call it trolling I guess. I got a kick out of it. To be honest, I would love to continue!

But just to be clear: All the drivel I ever commented on here was organically produced by my average, contrarian and unhinged brain.

I guess part of the thinking I wanted to provoke was about the fact that nowadays you just can not know anymore if you are talking to an LLM or not and how the perception of whether you are changes how you communicate. About how your communication changes, based on a single boolean of metainformation (isHuman/isAI/isLlm).

Bravo

Reply on news.ycombinator.com