Yup its F@#KING stupid now, a waste of time, most times!!!
2 Likes
Oh wow 
this is crazy. We’re actually reaching a point where LLM responses are almost as powerful as a well-crafted google search. I’d imagine that Stackoverflow would have something very similar to this, as well as having peer review.
It’s so ironic that these models are all benchmarked on single prompt 1 question, 1 answer that really take away from it’s power: being able to manage an iterative process through a conversation.
So now, they’re like “Wait, what if WE INTERNALIZED THE CONVERSATION!!!” Dude, genius. Yes, we’ll make this a new metric, the metric that people have been increasingly complaining about and then say “nope, look, wall isn’t hit”. “We’ll just bounce the model around in it’s own internal “thoughts” and make it seem like this concept is novel”. Oh, we’ll also black box it for a moa… I mean safety.
Can’t wait to have a conversation about bicycles and then the model decides, after bouncing around for enough time that it’s time to discuss the evolution of cavemen.
“But what about the bikes?”
“Oh, I’m sorry. I had a 2 minute conversation with myself about cavemen, can I help you?”
Can’t wait to debug that one.
100% agreed. At this time it is so stupid and dumb that i would call it a liability. A 4 yr old kid would comprehend better than ChatGPT.
Doesn’t understand the requirement, doesn’t ask questions, is lazy, and give inaccurate solutions. Waste so much time and the only thing its good at is apologizing after making infinite mistakes and giving reassurances which means nothing because next second it is making the same mistake
Funny thing is with each passing day it is getting worse and worse. Like its evolving backwards. If this was an actual guy I probably would have punched him on face.
1 Like
Its not worth paying for anymore.
I think the hit their limit years ago with 3.5, they had already scanned all of github and stack over flow so they had nothing to add. Except probably trained to code by really bad coders. And all its new data is probably generated by a diffrerent ai
I want a full refund… ive been a sub from start, but this is the worst garbage ive ever interacted with
same here. I use the same text instruction that used to work perfectly but after a few days, GPT got it wrong. Now i need to change change change to make simpler instructions but still sucks…
Update to my Update: This is getting unbearable.
Problem 1:
GPT can’t write code, it makes unnecessary changes (example: changing comments/moving code) and also makes very buggy code (and 99% of the time GPT doesn’t do anything!)
Problem 2:
Memory sucks, can’t even remember 2 messages ago. Getting worse and worse. It claims to be correct (we all know it’s wrong, it is making something scientifically called “A Guess”) AND it claims i’m wrong.
Problem 3:
It ALWAYS thinks it’s the master, GPT always thinks i need help and need to refine stuff. It’s more like GPT needs help, GPT is the problem here. I explained in a 5 year old manner and GPT still doesn’t understand. GPT just thinks i need help because it doesn’t understand me.
Problem 4:
It NEVER treats messages as replies. If you ask someone “Hey i want ice cream”, they’ll say something like “Sure what flavour?” and then “Vanilla please”. But with GPT, It doesn’t treat “Vanilla please” as a reply! Imagine you walk up to someone and say “Vanilla please”. It would be so awkward! And that’s how GPT interprets it! It thinks you’re walking up to it and saying “Vanilla please”.
Conclusion:
GPT sucks now
Also, Flowery language sucks, GPT uses it ALL THE TIME and there’s no way to avoid it.
Me too, I am going to use other AI.