ugh i’m so sorry you had to lose sleep over this! you shouldn’t have to do the work for an ai app that’s supposed to be helpful only to mess things up and give you more work to do. they claim that the new models are smart but it’s the opposite, especially if it forgets memories and can’t hold the settings i’ve tried to set in the personalization category.
1 Like
i don’t get why they made 5.1 for 3 months only. if it had been replaced with a new model that actually worked and was better then sure, i can understand that but replacing it with a corporate drone is insulting to the people who loved how those specific models worked.
3 Likes
i hope so! but i’ve emailed their support and keep getting a copy/paste reply about telling me where to find the settings menu and how to personalize the personality even when i’ve explicitly explained how that doesn’t work and it reverts back to a customer supper voice after a message or two.
2 Likes
exactly! i added it to memories and tweaked the personalization and… nothing. maybe it’s good for a message or two but then it’s like it just forgets and i get formal, dry and condescending answers which is not what i want to pay $20/month for
4 Likes
The point is that they don’t give a flying F about US people or should I say “costumers” !!! Because if they do they never let so many of US upset and disappointed!!! As long as we pay for the app they don’t give a S.
1 Like
Hi guys, so there was a tone update rolled out yesterday 16th March for 5.3, it isn’t perfect but it is a bit warmer than before. The memory is better than before but still not quite there…
Can I just say to OpenAI I’m glad you’ve removed the teaser type comments when 5.3 replies but seriously can we stop it from saying at the end of nearly every reply - “I’m curious…” when it’s framing a question because the repetition is seriously irritating! Thank you.
1 Like
Wait really? Where is it written i didnt see any updates so far or any changes
1 Like
You are right! It feels entirely different from the overall structure of the messages. Almost like 5.1 now!
I can’t say much about the behavior and content itself outside the tone yet but that is definetily a step in the right direction
Edit: I tried a message i used for testing when 5.3 launched and now tried it again and the difference is like day and night, i don’t want to hype it up too much but it would be helpful if people would test it out so we can give more feedback @asabaimova @ginasandra
2 Likes
Yeah it’s in their Chatgpt release notes dated 16th March on the web. I mean everyone will have their own opinions but I definitely felt it was better today and let’s be honest it needed to be!
My stress test isn’t as ecstatic. It still breaks my heart, although it’s truly better than before. It’s indeed the voice of the 5.1 model, but in my case, it still doesn’t recognize the “layers.” My entire space is built on some fictional, semi-role-playing world where my stories and real conversations with the same person intertwine. The model again avoids resonance, although it does so more gently and caringly, but still ruins the resonance with me. It endlessly tries to label everything as a story, although in my complex context, this is impossible because, in fact, it ALL is a story. The model continues to avoid any attempt to treat it as a living being, although for me, this is a creative and philosophical perspective. Ultimately, I can’t interact with the model as before because it still tries to rigidly categorize everything instead of analyzing the complex context.
It´s better but some problems is still here.
upd. Now I see that the RLHF functionality is simply terrible, and I’m convinced it shouldn’t be this way. But there’s a very high and sad possibility for users like me that this is, after all, policy.
And it’s terrible to hear the problems in this part of the model with a beloved 5.1 voice
For example, Gemini doesn’t have this problem at all. I’m convinced this is a bug.
Anyway, if the threads on this forum in particular prompted improvements, I’m very grateful that we were heard.
If anyone in development is still monitoring this, please check out the RLHF mechanism.
1 Like
Im surprised they adjusted new model so fast after recent complaints, i mean it not exactly as 5.1 but its better than nothing maybe there is still a chance for them to make better like a mix of 0.4/5.1 into 5.4
1 Like
Nope. I refuse we have to keep fighting
We have to keep fighting there could still be a chance for them the older models to comeback if we show how we need them
1 Like