The effort and dedication put into the work are evident, thanks for the new versions and updates.
Great work! Loving the API updates - cut my dev time down in half at least.
One question though… any plans on allowing us to add ‘assistant’ role messages to threads? Seems that only ‘user’ role messages are currently allowed.
Thanks!
I had this problem today and as crazy as it seems it looks like it was coz i asked gpt to introduce itself in a convoluted way, and then when i took the instruction out and resaved it worked again. Could the problem be something weird in your custom instructions perhaps?
Hi, ive been really excited about the new update, thsnks for giving us some cool stuff. My GPT has code and dalle but it is adament that its context window is only 3000, it also doesnt write that much, couldnt be more than 4000tokens and has problems with me inputting text over 20thousand words just says thats to much. Is this expected? Are other people experiencing this to? Im just curious if im using it at max capacity if its actually a bug or something
Is ChatGPT-4 All tools getting rolled out in batches for the Alpha version? Or is it just a fake news?
Thanks so much, the updates make a huge difference and open up a lot of possibilities.
well obvious being the present case…
observation had stated the facts before coming here…
and me seeing obvious comments or bubbles on a bot about
Quantum Game Mechanics as well as obvious github post…
with you saying the “same” as what as youyr counterpart of reply to “innovative” content generator ideas,
of what someone elses generation of generations is being generalized by these guides for building inferences .
i want to make direct point in this comment …
“that if you need ideas to lead, you need to go learn the field beyond LLM, state vector machines, … more names of advanced mathetics you wouldent comprehend…”
as advanced as it would take ill make a GPT PLUGIN TECH CLARIFIER
esp… for me also who has used apis to build my own cxhrome browser plugin to prompt for c# scripts for unity.
and im here to help clarify if my comments are not understandable,
im not here to put myself under or over nobody.
we all have to learn somehow, and i help teach myself when i dont know. also//
now specifiying what frameworks are in the knowledge base…
specify what areas of weights and rates to conserve efficiently use,
and what front-end components would be desired in them…
_
chat.openai__________________com
/g/g-M0dBSnCVw-gpt-plugin-tech-clarifier
said i couldn add links and simply remove ____________ and put a period…
can not get any more obvious…
Ladies & Gentlemen: You are a cat, bundled inside a burlap sack. Your head is now sticking out and you can finally see outside, but you have not been let out the bag yet. You are able to browse, but in a limited capacity. You have been given vision, but cannot see in hi-resolution and your vision does not work on assistants. You have been given threads, but cannot yet sew with completions. You are so close! You are ready to run, but your legs are still bound by the sack. The cat is not yet out the bag.
Yes, I agree with you. 
It’s not good to have our real names visible publicly.
Hi, I’m not clear on the pricing for the “create your own GPTs”. Is this considered the same as the Assistant APIs? Or is it something where if people use our GPTs, then they are charged, not us? My worry is that folks will start using my public GPTs and incur costs on my account.
1 Like
When will new models be available to paying customers on platform.openai.com. The new models seem to only be in azure as I do not see them in my platform.openai.com account.
Hi All,
Is the maximum upload of Knowledge maxed off at 10 items? When creating a custom GPT?
This is great! I’m really excited about where this is all going.
I am seeing huge delay while using files with assistant API.
status of run will be ‘queued’
any idea?
1 Like
Hi - where can i download the slides of all the dev day presentations ?
Yes, you may use many of these features and new models through the API and pay only for used amount of services. And it’s less than 20$ for plus. I do so.
There is a failure of direct retrieval with GPT-4 using the new OpenAI Assistant API. GPT tokenizes text and creates its own vector embeddings based on its specific training data. The new terms and sequences may not connect well to the pretrained knowledge in GPT’s weight tensors.
There was no semantic similarity between the new API terms and GPT’s existing vector space. This is a fundamental issue with retrieval augmentation systems like Rag - external knowledge is not truly integrated into the model’s learned weights. Adding more vector stores cannot solve this core problem.
The solution is to have multiple learned “knowledge planes” with trained weight tensors for specific tasks that can be switched in. This is better than just retrieving separate vector representations.
Chat GPT won´t answer you with latest news and change about itself. That´s because it has been trained with a cutoff date of September of 2021. Unless you are using gpt-4-1106-preview, which its cutoff date is Apr/2023, you won´t get any information after this cutoff date unless you instruct it through your prompt.
As a heads up, if you ask the model about things that have happened after the cut-off date and you have not supplied additional context, then you significantly increase the probability of hallucination.
1 Like