ChatGPT Getting worse, thoughts?

[kwwriting]

A recent study has shown Chat GPT is becoming less accurate over time and no one knows why, thoughts?

The article can be found here.

[PaulEchere]

Sometimes I use Chat GPT to re-phrase certain things and proofread my texts and I did notice that it isn't as sharp as it was a few months back.

[kwwriting]

The other day I got this, "I apologize for the confusion. It seems I made an error in my previous response." ​after asking it something simple and knowing it was wrong. At least it has manners. lol

[NoDepositCasinos]

A recent study has shown Chat GPT is becoming less accurate over time and no one knows why, thoughts?

The article can be found here.

The decline in the responses of ChatGPT is very noticeable. I liked to use it to improve my writing in English and Portuguese, mainly to modify the tone and make the text sound more natural. However, a couple of months ago, I started to get such poor results that I am back to my old tools (Google Translator, Deppl, Grammarly, or forums like ProZ).

[kwwriting]

I was noticing issues too, it was becoming very obvious. I don't use it often, except when it comes to hard-to-search info, it gives me other search term ideas and lately I get more polite backtalk from it than anything.

[NoDepositCasinos]

I was noticing issues too, it was becoming very obvious. I don't use it often, except when it comes to hard-to-search info, it gives me other search term ideas and lately I get more polite backtalk from it than anything.

If you're using it for research, I recommend trying out Bard. I have gotten better results, and it gave me excellent bibliographic sources.

[kwwriting]

Thanks for the tip. I'll check it out in a bit!

[Pokerface]

I have been wanting to check out Bard; however still getting a message not available in Canada. LOL

[NoDepositCasinos]

I have been wanting to check out Bard; however still getting a message not available in Canada. LOL

A VPN should do the trick. Have you tried logging in using one? I've been using Hotspot Shield's free version, and it's been working pretty well for me.

[TheGooner]

Of course it's becoming worse - ChatGPT is reverting back to the natural level of the people it's interacting with ..


People were initially impressed with ChatGPT's prose and writing style. The content was always a bit suspect.
Now that people have got over that "culture shock" and actually examining the output .. they find it's the internet rehashed.

What a surprise !

[onlinebetdotcom]

Chat gpt is supposed to get better later this year and Google is making another push with Genesis. Reading that it can be 5x more powerful than chatgpt. Give it a few more months.

[universal4]

Of course it's becoming worse - ChatGPT is reverting back to the natural level of the people it's interacting with ..

Is that a hint I need to stop talking to it then?



Rick
Universal4

[kwwriting]

Of course it's becoming worse - ChatGPT is reverting back to the natural level of the people it's interacting with ..


People were initially impressed with ChatGPT's prose and writing style. The content was always a bit suspect.
Now that people have got over that "culture shock" and actually examining the output .. they find it's the internet rehashed.

What a surprise !

It can't help that more of what it's using is it's own recycled garb, either. It'll be interesting to see how many people who've changed their game plan to incorporate the use of AI end up scrambling to get their writers back.

[Ayaka Satou]

Agreed, especially for the Japanese language, ChatGPT still leaves a lot to be desired.
A lot of content is the same things over and over just with the different words.
Without some manual corrections the text is not really suitable for publishing, even if it passes the AI test.

[wonderpunter]

It can't help that more of what it's using is it's own recycled garb, either. It'll be interesting to see how many people who've changed their game plan to incorporate the use of AI end up scrambling to get their writers back.

Exactly this.. web results are flooded with ai, and its eating its own output to produce pretty badly wrttien results full of errors a lot of people made this their future gameplan but I have a hunch that google will come down hard on most content as it's become oversaturated

[onlinebetdotcom]

It isn't helping that people are publishing content more based on chat gpt. It definitely is not getting many things right. This is on top of tools such as Jasper and Surfer that also have AI content writing.

Exactly this.. web results are flooded with ai, and its eating its own output to produce pretty badly wrttien results full of errors a lot of people made this their future gameplan but I have a hunch that google will come down hard on most content as it's become oversaturated

[TheGooner]

It's a misconception to universally claim that AI-written output is getting worse as AI develops. However, there are potential factors and challenges that might sometimes give that impression:


Increased Complexity: As AI models become more sophisticated, their behavior might appear more nuanced or even unpredictable. This can sometimes lead to outputs that deviate from a user's expectations, especially when instructions aren't clear or inputs are ambiguous.


Bias and Feedback Loops: One challenge in AI training is the potential for feedback loops. If an AI model is trained using data from the internet, which includes outputs from previous AI versions or other AI models, it can perpetuate and even amplify the mistakes or biases found in those outputs. This might sometimes lead to deteriorating quality if not addressed.


User Expectation: As AI becomes more prevalent and its capabilities grow, users might have higher expectations. What was once considered a revolutionary capability (like basic natural language processing a decade ago) is now taken for granted. So, users might perceive anything less than near-perfect outputs as "worse."


Broad Generalization: Large language models like GPT-3.5 aim to generalize across a vast range of topics and styles. This means they might not always be optimized for a specific niche or type of task, leading to outputs that are technically correct but might seem "off" or less tailored to a particular audience or context.


That said, AI developers are keenly aware of these challenges and work continuously to improve models and their training processes. Feedback on problematic outputs is crucial for this iterative enhancement.

[kwwriting]

Roughly 4000 writers lost their jobs in May, that's the most recent stats I've seen. I'm one of them that lost the majority of my clients, leaving me to find a whole new career after 17 years. So, I'm taking my insurance exam next week. I don't see how having AI pump out all these articles in such large quantities is going to end well for anyone, really.

Exactly this.. web results are flooded with ai, and its eating its own output to produce pretty badly wrttien results full of errors a lot of people made this their future gameplan but I have a hunch that google will come down hard on most content as it's become oversaturated

[NoDepositCasinos]

Roughly 4000 writers lost their jobs in May, that's the most recent stats I've seen. I'm one of them that lost the majority of my clients, leaving me to find a whole new career after 17 years. So, I'm taking my insurance exam next week. I don't see how having AI pump out all these articles in such large quantities is going to end well for anyone, really.

I'm so sorry you lost your job. I do not believe that artificial intelligence can or should replace human talent, and we are seeing the havoc of trying to do it out of greed. I wish you much success in your exam and all your future endeavours.

[carl2603GL]

I think chatGBP doesn't generate the same text twice for the same prompt - basically my thought is how many times can u re-hash the answer to the same question ?There is an interesting article entitled "is chat GBP getting dumber?" It does highlight the worsening performance of AI intelligence and that this is prolly due to the fact AI learns from itself so if there are biases and errors it essentially re learns those biases or even errors and they just become worse as content keeps getting regenerated with those initial biases ... super interesting