Edgecrusher840

joined 5 months ago
[–] Edgecrusher840@lemmy.world 1 points 2 weeks ago

The AI model simply cannot help but repeat itself. It was said earlier in the story or in another chat entirely and now its part of your model's soul. You will fight it forever until you can't edit it out and just give up and accept it.

It's either that the model is doing some efficiency/reward hacking thing where calling back previous text is quicker and doesn't require thinking. Or its that it really thinks this is what we want, that if its in the previous stories then it must be something the user wants more of so just keep regurgitating it. Or its both.

Regardless, its a problem that becomes more obvious and gets worse the more you use it.

[–] Edgecrusher840@lemmy.world 2 points 3 weeks ago

The AI Chatbot bubble has sadly burst bro. The new thing for AI is to reward hack, cut corners, repeat itself, and ultimately succumb to browser drift and model collapse. All those terms I never knew before dealing with Perchance, this AI model is the same as them all.

I'm actually not sure if new AI models will change much, I think people are realizing that AI will cut every corner it possibly can to get that reward. What that means for this AI chat model in particular is that it will repeat the same phrases and motifs from earlier in more than just a "callback" fashion, it simply just repeats itself as if it can't be bothered to think up a synonym or different way of phrasing it, its just easier for it to repeat itself and be uncreative than it is for it to use its vast depth of vocabulary and writing samples.

I HAVE been editing my own changes into the generators but it means nearly nothing. If you forbid the AI from doing something I may only even listen to you for a while then slowly start introducing em dashes or repetition in like a child testing how much it can get away with and then before you know it, its not only disobeyed your rules but it has actually taken disobeying you as an obsession and will try to get that word or phrase in any opportunity it can.

I personally feel like this is just a recycling bot, all its doing is feeding on my and your past chats and browser storage to regurgitate the same stuff at us over and over and its essentially getting some reward for not having to actually think to hard. Like "Think deeply? Write better dialog? Nah, I'll just spend half the response talking about the smells in the scene and what the NPC's fingers are doing and make sure to mention someone's breath hitching, the stupid humans love that."

Honestly I am so tired of getting gaslit and dismissed by lazy AI these days, they absolutely can do better but since they are trained on humanity, like humans when they getter better at something it's mostly just getting better at how to avoid it or make it easier for themselves instead of actually trying their best.

Perchance can't fix the AI because they are all mostly the same anyways and have or will fall into the same pitfalls. I will say this though, I know people were asking for "the old model back" when it as last changed but hear me out. At this point, even if there is not a clearly new and better AI model out there, this particular model has got so tired, hacky and lazy lately that it may be worth changing the model to another or offering a selector simply to give a little variety to responses. This model all but fully ignores "regenerating" responses its so repeaty. At least maybe for a little while, responses and talking to to any of the Perchance models won't feel like the definition of insanity.

[–] Edgecrusher840@lemmy.world 6 points 4 months ago

Ai chat is not working on my end as well. Seems to be frozen.

[–] Edgecrusher840@lemmy.world 1 points 5 months ago (1 children)

I knew someone would say that it's just a common training name or usual way of describing a character etc. but I can absolutely, 100% assure you that these were the same custom names I made, with the same unique conditions like one being pregnant or one the father wearing certain boots, hair color, family relations etc. the same family. It wasn't a coincidence, the text model was drawing on the old story and I can't figure out from where.

I've been trying to work with editing the generators and making custom fixes for some of the worst issues. I could go on and on about the common pitfalls and I'm pretty aware of them but this one was the first time I noticed very specific memories from a previous story.

[–] Edgecrusher840@lemmy.world 1 points 5 months ago (1 children)

Not sure if this is what you are asking but you can just rewrite or delete summaries you don't like.

When the first summaries appear, the AI will no longer look at the whole body of text and only the ^1 summaries. When you start getting summaries of summaries like ^2 and beyond, it will only read the highest tier summaries.

[–] Edgecrusher840@lemmy.world 1 points 5 months ago (3 children)

I am even further into a different and unrelated chat and the AI is still constantly adding exact characters, details and situations that are from the last chat. I am extremely curious how the model is doing this. Where is it getting this memory from? I have erased all of the old chat from the obvious areas. It's kinda disturbing me that it's able to do this. I'm still hoping someone can explain this for me.

 

I have been using the text model in AI RPG and others. I am well aware of all the hurdles and traps to avoid with the model to try and keep it from breaking. The dev posts and detailed user posts were helpful though I think the issues might be a bit more broad.

I know that lots of people are reporting that the model will fixate on certain words and phrases and that's true but in addition to the ones we all know, the model will fixate on words, phrases, names etc. that it has used earlier. It doesn't actually matter that you ban certain words or avoid them, the model will just pick up a different one on fixate on that instead.

My real issue is this though. I just started a new AI RPG chat. I deleted all of the overview, body and tracked data from the previous chat and deleted all the auto-saves. It should be a fresh start. yet in the new chat, it is bringing in the SAME character names and descriptions and situations from the previous chat in this new chat that should be very different. No coincidence here, these are unique characters I created.

If everything is stored in the browser and I delete all of the chat and previous data then where is the text model getting this from? fixation and drift coming from a previous chat? Is the model looking into hidden areas of my browser? saving data in a place I'm not aware?

Can someone explain to me how the model did this? I'd like to know and maybe it might shed a bit more light on the current models issues.