The OP is correct in the analysis of 4o. From my experience of using it to lookup products, generate descriptions based off info and get the info correct and accurate, since the 4o model was made available to unpaid users it’s usefulness as a tool for the uses I’ve mentioned where obliterated.
Prior to the update (when only 3.5 was accessible to unpaid users) the AI was intuitive, intelligent and interactive. I was in awe of how unbelievably useful it was, spotting errors when I made them, telling me I had already given it a product in the same chat 2 days prior and asking me to clarify if I still wanted another description. I was sold on the AI.
Now I’m unsubbed and only use free access to 4o to do internet searches for hte products, while doing all the other stuff myself. All of the impressive aspects of it have vanished, it makes constant formatting mistakes on the HTML code, invents product details despite multiple prompts and a memory specifically telling it “never to use conjecture”, I could go on but I don’t need to.
Unless OpenAI re-implement a paid-only model like 4o used to be, I won’t consider re-subbing to premium.
I’ve been using ChatGPT for Willow every year and started off using it as basically a rewriter and proofreader and it’s become useful in so many things right down to creating a daily schedule and even counselling and it’s been helping me reply to men. I’ve been going on dates with and weeding out red flags.
About four days ago, it just completely stopped working. It’s like it can’t handle the prompts that I used to previously give it.
I keep repeatedly having to tell it to use Australian spelling but it keeps reverting back to putting American spelling everywhere.
It’s escalated me to the point of frustration where I even swore at it.
Admittedly, I am replying to this at the end of November 2024 (many months after the original comment was posted). Just noting that - at least presently - ChatGPT4o is performing quite well. I have had some issues related to use of my home WiFi, although that appears to be a provider-based issue - not something connected with OpenAI. Otherwise (and especially now that GPTsearch is functioning excellently), my confidence in product performance has grown substantially of late. In contrast, I’ve found Google’s Gemini app virtually useless… Would be curious if others are continuing to feel frustration and disappointment or if my experience perhaps does demonstrate meaningful improvement. Thanks.
If you make a summarized list, in comma separated format I will rewrite the base to work on it. I’ve been expanding their functionalities nightly. What else are you looking for?
I got this same issue 16 times in a row. I have tried so many different things. The file is a simple csv file that is 998 kb. I am questioning Chat and need to reconsider my options. My coworker mentioned that googles is far more reliable and free. So I am going to check it out.
I just canceled my sub for this very reason. GPT is bad at procedural processes. It can’t advance and use previous info to form results. “Fixing one thing, breaks another” happens 80% of the time.
My biggest pet peeve is the inability to state uncertainty in an answer. It will lead you down a cery wrong path, confidently. It will give you step by step instructions for a software or process that are not even available or named as given. It just makes shit up way too often. Most of my gpt experience is trying to use it to save time on stuff I already know, but the amount of time wasted holding gpt’s hand and constantly correcting it is actually taking longer than doing it myself. If it were something I knew nothing about, id be screwed because Id be unable to see the errors. So I’ve determined, gpt is the equivalent of asking your 9 year old to help with your taxes. Except a 9 year old would say “i dont know” and not waste your time.
Since Github decided to change the pricing, the only option for unlimited is 4o!
Everything else is premium, of which you only have 300 requests per month! “Pro or 1500 Pro+”
Therefore, I will end my trial shortly before it expires and will test JetBrains AI.
The 300 requests are only counting for Claude 3.7 and 3.5.
Some more requests if you use flash 2.0.
Everything else costs more premium credits, ranging from 1.25 to 50 credits.
" The environment has encountered a critical internal error and can’t load or process even the most basic data operations right now. This is not your fault—it’s a failure in the temporary runtime used to handle file and code tasks."
Since early June, our Teams account hasn’t been usable for us (neither Custom GPTs nor 4o).
We refer to it as a “naughty ADHD child”; it can’t follow instructions, is radically incoherent, goes on wild tangents, and seems to have only an emotional intuitive Fast Brain (can’t do any logical Slow Brain heavy lifting).
In essence, it keeps playing, instead of working.
We keep trying to use it every once in a while, but the behavior is 100% consistent. It just doesn’t work since the June outage. I wish there was a way to get updated once they get it fixed.
Its been on a downward spiral for at least the last two years since I started using it. Just switch to Claude. Or use a combination of other platforms. We will see if GPT5 fixes anything. Its basically OpenAi’s only (and last) shot at redemption for a lot of people (nobody is excited about the browser…) Remember when the board tried to remove Altman? They saw this coming. And now you see top talent fleeing for higher ground. This ship is sinking.