Previously, I wrote about a noticeable decline in the performance of chatGPT Plus/GPT-4 Over the past 4-6 weeks. I have been wondering if that is just me or if a lot of other people are experiencing different issues with chatGPT too, specifically people in the US.
Table of Contents
Jan 29 – High error rates on ChatGPT
Over the past few hours, starting from about 3pm PST, chatGPT has been experiencing major issues. If you use chatGPT, the system may throw in the message “error in moderation” or the like.
Here is what OpenAI has shared on their official website https://status.openai.com/. There seems to be two separate issues:
but after that, there seem to be multiple errors, impacting the API feed, which is a big deal for developers.
Nov 16 content
I looked into this further today using Google Trends and this is what I found:
As you can see from the above picture, over the past 7 days (roughly between Nov 8 – Nov 15) in the US, a lot of people have experienced moderation-related errors or mistakes. The top rising related queries to “ChatGPT” are mostly about these issues. (In case you wonder, the long related query is: “chatgpt can make mistakes. consider checking important information”).
And they are going to Google to figure out what is going on. The “Error in moderation” issue was raised again on the OpenAI community forum or on Reddit. The issue seemed to be on late Nov 11/ Nov 12, based on search volume in the US.
According to Sam Altman, OpenAI has been experiencing a surge of usage after their Developer day last week, to the point that they needed to temporarily pause the chatGPT Plus subscription. I guess they are a “victim” of their own success.
OpenAI only recorded this generic message in their system uptime monitoring dashboard:
According to SimilarWeb, traffic to chatGPT web version has been relatively stable over the past 30 days (Oct 18 – Nov 14).
So my working hypothesis is that, since Dev Day, tons of GPTs have been built and shared by developers or just normal users. And in order to use GPTs, users need a Plus subscription. Hence, while the overall traffic level hasn’t increased substantially, the usage and demand for chatGPT plus has skyrocketed.
I personally have experienced a much lower cap for GPT 4 usage over the past few days. And when I hit the limit, I need to wait for around 2 hours+. This was not the case before dev day at all.
What have I done differently as a result?
- I subscribed to Claude Pro from Anthropic and have been having good experience with it for my writing needs.
- I subscribed to GitHub Copilot and have been having a good experience with it from a coding perspective as well.
- I have been testing the output for coding between ChatGPT 4 and Github Copilot and the responses have been quite different for the same prompt/context.
- The weird thing is that there has not been a clear winner between GitHub Copilot and ChatGPT 4 based on my limited experience so far.
Nov 29 update
It looks like “error in moderation chatgpt” is rising again in the US, over the past 7 days, from Nov 22 – Nov 29.
This is what I can find in Google Trends Today Nov 29 for the US.
Search interest for that term shot up massively on Nov 28, around 11am.
Looking through the OpenAI system performance report, this is what they reported for Nov 28 event:
That’s all from me. If you are using ChatGPT regularly, do you face the same issue recently? Share with us, using the comment box below.
Chandler