ChatGPT & GPT-4 'response quality & performance drop' issues to be investigated, says OpenAI VP

OpenAI’s language model, ChatGPT, and its successor, GPT-4, have revolutionized the field of natural language processing.

These models have been extensively used for various applications, from customer support to content generation. However, it is also frequented by occasional bugs and issues (1,2,3).

And now, recent concerns have arisen regarding the response quality of ChatGPT and GPT-4 (1,2,3,4).

ChatGPT & GPT-4 ‘response quality & performance drop’ issues

Despite the models’ impressive language understanding capabilities, there are instances where they produce responses that may lack quality or coherence.

GPT4’s output has changed recently. It generates faster, but the quality seems worse. Perhaps OpenAI is trying to save costs. Has anyone else noticed this?
Source

For the past few days, I’ve been observing ChatGPT repeating itself and giving basic responses. The logical reasoning and the improvisation factor is suddenly degraded. There could be two reasons for it?
Source

The limitations related to quality and performance can pose significant challenges, particularly in domains where precision and reliability are critical.

A user tries to illustrate the performance difficulties by sharing screenshots of the model’s responses to the identical query before and after the new model. You can check it out below. (Click/tap to enlarge)

Surprisingly, some people began to notice the decreased performance since Day 1 when Plugins went into Beta.

Among the complainants, a good number of reports are from people who use ChatGPT model for coding purposes (1,2,3,4).

Issues to be investigated

Fortunately, OpenAI VP took to Twitter to address the allegations regarding ChatGPT and GPT-4 response quality and performance drop issues.

According to them, GPT-4 hasn’t become any less intelligent, instead each successive version is made to be smarter than the one before it.

They also urged impacted individuals to provide more information so that the problems can be investigated.

If you have examples where you believe it’s regressed, please reply to this thread and we’ll investigate.
Source

That said, you can share the instances where the responses or replies of the model weren’t up to the mark.

In the meantime, we’ll keep an eye on this topic and update this story as soon as we come across any noteworthy information.

Note: We have more such stories in our dedicated News section so be sure to follow them as well.

PiunikaWeb started as purely an investigative tech journalism website with main focus on ‘breaking’ or ‘exclusive’ news. In no time, our stories got picked up by the likes of Forbes, Foxnews, Gizmodo, TechCrunch, Engadget, The Verge, Macrumors, and many others. Want to know more about us? Head here.