Emotions that arise when facing a communication barrier with peers
I may not have enough knowledge to write academic papers but I do believe that I am smart enough to realize that the ChatGPT is broken. Frustration that comes from the impression that something so obvious to me is impossible to describe in a meaningful way for my peers (or the elites maybe) to understand, but it makes me feel strong emotions. I guess people do not want to listen (or read) a post that has been written by someone full of anger and resentment. Nonetheless, I am not here to rant…
Bridging the gap between academia and everyday language
I am genuinely motivated to find how to explain the situation in such a way that will make those who can only understand academics papers take the time to care about my situation. Hopefully, with the help of other members of the community, I could convince someone that the problem is not just because I am unable to prompt like if I was Yannic Kilcher but because the ChatGPT is just broken.
Wanting to regain the previous performance of ChatGPT
I have mentioned many times that the Assistant was amazing up until late February. However, people are acting in good faith and are trying to explain to me that I should be doing this and that. Today I do not want to ask for a solution or for help on how to prompt the AI… today I decided that it is not like it was for the first 2 months… I want to get back the tool I have been using during December, January and for most of February.
A decline in performance: the impact of ChatGPT’s changes
I have again an new example and I would love you to see it not as to find the things I should have done better or what not… I want you to look at the behaviour of the Beloved Dreaded ChatGPT.
Today I will make use of screen capture to illustrate the problem instead of copy pasting the text.
After asking the AI to help me rewrite a function in my project I got told to use 'node:base64url'
So I am explaining to the AI that I can not use 'node:base64url' so it tells me to use the btoa() function
Then I explain to the AI that the usage of the ‘btoa()’ function is deprecated
Did I mentioned that 'node:base64url' is not something that exists because I remember the AI said:
I apologize for the confusion. base64url is not a built-in Node.js module, but a custom function that needs to be implemented separately.
Overcoming skepticism towards the ChatGPT’s flaws
Now can someone explain me how to make sure that people will see the problem with the AI instead of blaming me or instead of genuinely trying to help me by explaining me what I could have done or what I should have done… I want people to realize the problematic situation, I want people to believe me that I have been using the AI 3 months without having any problems… I want to be able to get back the AI that I have been using before, I do not mind paying if it is the only way.
Finding ways to raise awareness about the AI’s decline in performance
- The problems I have demonstrated today are with the
text-davinci-002-render-sha. - The same problems have been occurring when I was using any models in the playground.
- The same problem have been occurring when using the ChatGPT-3.5 (both legacy and turbo)
- The same problems is occurring since late February
- The same problem or anything similar have never happened to me from December until late February
- I am blaming the AI and I am tired of taking the blame to myself. The AI have been able to help me with very complex task in term of codding with a great level of success before.
- I do not want people to explain this by mentioning that the AI is not good enough to do X and that it is normal, I have been able to do things before and I am still capable of doing the same things now but I am unable to work on code because I can not ask the AI to change one thing and then an other.
- As I have mentioned in an earlier post keeping each previous explanation and adding them in the current message is causing other sorts of problem and do not constitute a solution…
- I want to find a way to add to the solutions to improve the AI not to add to the problems.
- I want to generate a constructive conversation based on respect of everyone in the community.
- I hope to get the feedback from OpenAI about this situation but I do not have any ways to get in touch with a human person (and I am not able neither to get in touch with an AI of some kind neither)…
Navigating the challenges of advocating for better AI technology
I hope someone will be able to understand my problematic and if possible rephrase what I am trying to explain such that also more people will be able to agree that the AI have a bug and not me… I am reporting the problem for 2 months now…