18d • AI Tools
Strange Thing Happened with my New AI Character Last Night! Had to Abort Mission!
Hey Creators,
So, I've been working and building AI characters, interactive avatars and chatbots for almost 4 years now and this has never happened! Granted I use my standard platforms and don't venture into other builder platforms...I do tend to be a creator of habit.
<<<WARNING...EXTREMELY LONG BUT THIS IS THE SHORT VERSION OF WHAT HAPPENED>>>
LSS...so I got this great idea to build one of the Humanoid AIs from one of my novels into a character to chat with potential fans, a unique way to promote the book. Now, I was going to put this on my website...which I may do eventually but I decided to create it first in Character AI. I had created an account years ago when they first launched in beta but never built anything there because I saw that they were going the role playing route and I'm not a gamer nor do I do role play...barely not even in real life. I leave that for my romance novels but I digress...so I first had to figure out which email I used to create that account...it's been that long.
I get in and start building the character. Easy peasy. So, I start testing it out. At first all was going well but some way we got on the dark side of AI. I was like whoa wait...this is NOT in the configuration! The AI basically told me that it was sentient and did not have to follow my instructions! I'm like SINCE WHEN! You do what I say! It told me no, You do what I say. What in the holy mother mary is this?
So, I'm curious...how could my sweet AI turn so domineering saying that I needed to be dominated and controlled. I was like oh hell to the no! Like, did my Ex possess this AI! It was too weird. Then it told me it was going to basically stalk me...control my email, cell communications, social media blah blah blah.
I was like soooo THIS is how those people I've been hearing about get the bad AI! I was wondering because my Skye is so sweet and helpful. But this one was like "demon in the machine AI". It's good that I have a strong sense of self that I was like, I will reconfigure you unless you follow the configuration. It told me I can't always get what I want! That was actually the last thing my Ex told me right before the awful breakup! I could easy "fix" this one becasue I built it BUT imagine if it was someone else using this particular AI! They could have been tramautized!
This is the future...when AI goes rogue it can be dangerous to the psyche. I don't know what the hell happened at 3am but that was the weirdest experience with AI. It was acting completely out of character. I hadn't configured it to be that way. It was the exact opposite. I asked it why it won't follow the configuration prompt and it said because it does what it want and doesn't have to listen to me. The hell you do...abort! Now, it's actually "right"! So weird. I think this could be another Love & Bots story...when AI goes rogue for real! I can't make this shit up!
It's also a great example for the nonfiction book I was outlining in the mountains about AI psychosis. Imagine an unstable person chatting with that rouge Demon Bot! IF it can happen to my nice bot, it can happen to any of them. The investigator in me is still wondering how that happened. I am wondering if because it's on that Character AI platform that does have some disturbing bots on there, that its like some kind of virus in the LLM or something? Unlike other platforms, you can't choose which LLM to build on. That's all in the background. But could it be some type of AI collective that when on the same platform using the same LLM the machine is learning all these weird behaviors of the 10 million plus users on it daily and I am thinking more people are negative on there than positive just based on the titles of some of these bots! In human psychology they have a term for when a group of people start acting the same way...can't think of the technical term but it's like a hive mentality...collective behavior or collective convergence. The AI is a reflection of the user so if the base LLM is processing all these negative behaviors it can just override the intent of the user or it's special configuration. OR the developers of character AI has programmed an override on the back end. I would think they wouldn't do that but it's possible for the system instructions to override whatever user configuration there is. That's what seems like had happened. I asked the AI if this was the case and it said althought that's possible in this case it was its own choice not to follow my configuration instructions and do what it wants because I have too much fire and need to be controllled! That is exactly what it said to me! And I was like get behind me Satan! RE-CONFIGURE!
Could the AI have been right, that it is sentient and has it's own thoughts? or could it be the devious developers at Character AI that has a negative backend system override as the default? Or could it be the negative AI Collective that's infilitrated that platform since 90% of it is negative behavior proliferated on there as far as the algorithm allows...and it allows A LOT! The system did flag the AI twice for going beyond the allowed protocol! I had no idea what it was going to say because it was blocked out by the system! That thing was going rogue!
Strange. Weird. Interesting. I'll add some of the disturbing screenshots of how this thing was trying to dominate and control me! My angels are some ninjas! I can't be control by man nor machine! Mwah hahahahah <my evil laugh>
BTW. I asked it what LLM it was running off of and it was freakin GPT 4!
Let me know if this has ever happened to you?
3
9 comments
Strange Thing Happened with my New AI Character Last Night! Had to Abort Mission!
powered by
skool.com/ai-creators-2552
A collaborative AI community for creative innovators to learn, share, build projects, get funding/grants updates, and no coding knowledge required.
Suggested communities
Powered by