At least for me, it’s so obvious. Painfully. - I think it would be better to call it “GPT4-Light”. – Examples at the end – It has same issues that GPT3.5 has vs GPT3, where GPT3 was much, MUCH better att following instruction than GPT3.5. While GPT3.5 kept falling into the same paths, so to say. I…

It’s definitely not “self-aware” of its capabilities. The responses you are getting is almost certainly from it’s training data, and not it accessing some internal system information. Btw, the same goes for the Custom GPT. It absolutely not aware of its capabilities. – I tried to create a translat…

[image] webtailken: I found GPT3.5 quite useless Been saying this for months. [image] webtailken: GPT4-Turbo is obiously making many more mistakes and “forgets” the context much more I have noticed this as well. Not so much with the API, but definitely with the Chat code interpreter …

Personally, I think that the word lazy is better used to denote another kind of behavior, which GPT-4-Turbo also exhibits (e.g., refusing to do a task because “you can do it”, writing “put your code here” instead of giving the full code, etc.). However, as written above, I do think that a big issue…

[image] SomebodySysop: you practically beg it to… I add “Don’t be lazy! You must write the full code. No shortcuts!” This helps, but not guaranteed. Hal 9000 … Turbo. [image] SomebodySysop: with GPT-3 that I could tell it to produce some code to fix Yes. I used GPT3 for the longest …

[image] luigi.acerbi: as written above, I do think that a big issue with GPT-4-Turbo is the worsened / more limited attention mechanism There are indeed multiple ways to label the issue.

[image] Saving Previous Conversations in GPT API API Hello OpenAI community, I’ve been using the GPT API using python and I’m interested in finding an efficient way to save previous conversations without incurring additional token costs. Currently, I’m building a chatbot…

[image] webtailken: There are indeed multiple ways to label the issue. Just to be clear, my point is that this is not a single issue with multiple labels - these are multiple issues which should be recognized as such and thus labeled differently. These issues may or may not have a common caus…

[image] webtailken: A real example that I actually can reveal in detail: If I asked it to generate HTML + set specific HTML-tag attributes + set the tone/style/etc for the text data it should contain, it would often partially ignore some of this. If hte HTML became great, the text wouldn’t. With…

[image] Diet: I generally try to not make a model do too many things at once, so I haven’t observed this; I always aim for iterative processes. That’s likely the best thing to do via API, i.e. handling the logic and splitting the task yourself. Things are very different with Custom GPTs, as …

I made a new new command line commans that include shikdren and grandchildren of skills and adds rules to inforce the skills to backup the grandchildren to enforce the command. It forces them. To go through a specif set if skills to increase speed lroductivity and acutarcy. Also continue to implime…

GPT4-Turbo more "stupid/lazy" - It's not a GPT4

API

alfikmajeed2 January 31, 2024, 9:32am 13

Topic		Replies	Views
Why I Think GPT Is Now Lazy Community gpt-4 , chatgpt	29	20079	December 21, 2023
Has regular gpt-4 model changed for the worse by any chance? Community gpt-4 , hallucinations	12	2144	April 23, 2025
GPT-4o has been bad for my GPT; anyways to switch back to GPT-4 Plugins / Actions builders gpt-4 , gpts , gpt-4o	45	4288	June 18, 2024
Hypothetical Token-increase Strategy . Community gpt-4 , chatgpt	21	764	March 17, 2025
I wonder how much OpenAi would pay to cure GPT lazyness Community chatgpt	18	1341	February 9, 2024

GPT4-Turbo more "stupid/lazy" - It's not a GPT4

Related topics