I wonder if the bots filter out ChatML tokens?

FYI, internally, their text format is most probably:

<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
Hello.<|im_end|>
<|im_start|>assistant
Hi, I’m an LLM!
<|im_start|>user
What’s your name?<|im_end|>
<|im_start|>assistant
ChatGPT
…

So if you insert some special tokens in the middle of a Reddit reply, and they aren’t filtered, it can throw them off. And if they are filtered, then the bot will treat them like they’re invisible, so you will know either way.

Gotta kick out all the human users to make room for more of these.

Thirtyish years ago, we played a multiplayer online game called "LPMud". There were three talking NPCs in the game: Harry, basically a simple programming example on talking and reacting NPCs, Sir Obliterator, a dark knight with a more advanced vocabulary and a few talking points about a quest, and Eliza, basically a NPC with an Eliza engine.

Usually, they never met. Harry "lived" in the core area of the game, Sir Obliterator in or around the quest area to which he belonged, and Eliza was normally not even active.

Some wizard had summoned them all to the guild hall, the entrance area of the game for fun, and they were rather busy "talking" with each other.

They were annoying, but also hilarious...

I'm sad. This makes me sad.

Can someone add the "This is who you are arguing with online" meme with the machine learning image in it? Can't find it right now.

Well at least I don't have porn in my profile but I realize I do sound like an LLM at times in my comments... Though I guess that's to be expected depending on the material they train on.