When you learned about personalized GPTs that worked, when you showed your colleagues lots of things, incredible things.
And now, GPT 4o has an Qi of 5 and is incapable of respecting a single instruction, you too are ashamed of giving demonstrations?
When people expect to see “artificial intelligence” but in reality it is incapable of writing text without using a list of predefined words, don’t you feel like you’re being ridiculous for using this thing?
People were surprised at the quality, now you demonstrate and people laugh, GPT has become so ridiculous and useless.
I think the creators are ashamed because there are no more posts about these issues, they deleted over 100 posts to cover up their shame.
I understand some frustrations. Personally I have found that GPT-4o is socially retarded but is very capable with single turn conversations, which is mainly what it’s bench-marked on.
So for my own projects I use GPT-4o for single-use cases like converted unstructured text into a structured format.
For conversational purposes I use ChatGPT4 for random questions and Claude for coding projects. ChatGPT4o just ignores context and goes on it’s own path, patting itself on the back.
In your case can you show where the model is failing? Nobody is deleting posts that provide objective facts.
BUT
The good news is that other companies are releasing much more competitive models. You have lots of different models to try out for a very fair price & can mix and match to your desire.
1 Like
Hi, from my side it’s very simple.
I make a list of words: Explore, Fascinating, guide, discover the fascinating world.
he is incapable of writing a text without these words. Before GPT40 I never had this problem.
I tell him that he used a forbidden word and he replies: Yes, I used a forbidden word, excuse me.
He rewrites the same thing again with the same word.
I don’t even understand why they continue to offer and sell this feature after destroying it.
I think we need to urgently regulate AI services.
You can’t charge for a tool and then lower the capabilities later by using words like “laziness”.
It’s not honest.
1 Like
Totally agree. GPT4o does not listen well to instructions that go against its training data.
Maybe ask for it to re-write the text using synonyms? It’s not ideal but it may help
1 Like