As we get closer to GPT-5, older models have already been taken off ChatGPT Web chat.
Posted by JeffreySons_90@reddit | LocalLLaMA | View on Reddit | 6 comments
ttkciar@reddit
I'm glad to be using open weight models, which no corporation can remove from my computers.
bilalazhar72@reddit
I get your point and I fully agree, but it is not that deep bro
, they are going to take like the older models of the website and add a better model, it's just as simple as that.
XiRw@reddit
Now that 5 is out and I’ve been experimenting with it a bit I find it interesting that more than once it tries to answer a phantom (older) question as if the new question I am asking doesn’t exist no matter how long I argue with it until I start a new chat thread. 4 never had this problem
bilalazhar72@reddit
i think theyll fix the model in next iteration fully
XiRw@reddit
You’re probably right. The one thing I like though about 4 (earlier versions of 4 because now their restrictions were getting ridiculous before the release of 5) is that their safeguards seemed more realistic. Especially with Sora their image generator. In earlier versions of 4 it could print anything as long as it’s not murder or sex related. Now, if I say generate a picture of woman who is fully clothed but has an hourglass shape (and the context was just a normal selfie pic) it no longer will allow the user to adjust a woman’s body. But I believe you can with a male so they must be turning into anti woman then, just hate their users, or both.
bilalazhar72@reddit
Okay, this is a very controversial take, but I think if you're someone who wants to use AI for certain adult activities, you should just use local LLMs. These companies have to follow rules, and even if they want to target specific audiences, they have to show that their tool won't produce inappropriate content. I think the way Grok is handling its business is going to be very product-oriented, but it's never going to get a huge market share in the enterprise sector, which is where companies like Google and OpenAI make most of their revenue.
Elon Musk has a lot of people at X to sell Grok 4. If a company like OpenAI were to do something similar, giving users full control over certain features, I think they would do it in a separate product or experience altogether, like ChatGPT for adults or ChatGPT Virgin Edition or something like that.
Most of the things I use AI for are very engineering-focused and research-oriented. I can't even imagine that some people actually use it for building relationships with AI and stuff like that. I know there are people out there who do that, but it's beyond me. Your life must be miserable to enjoy things like chatting with a CSV file on a server somewhere.