[solved] We are currently processing too many requests (and related errors)

Every other post these days is about getting errors related to downtime for the service. A few people have posted asking how to get their money back.

I am not an OpenAI employee. But this is my 5 cents that may (or may not) help.

As you can imagine, OpenAI is probably getting tens (or hundreds) of thousands of support tickets a day (the growing pain of instantaneous onboarding of millions of customers)

They probably can’t deal with every request even if they have hundreds of support staff.

My suggestion for unhappy customers:

  • Cancel your current subscription (problem solved moving forward)
  • Send in your support ticket (about your current charge)
  • Don’t hold your breath (or you might die)
  • Decide if $20 is worth getting upset about and involving credit card chargeback etc. etc
  • Come back when things settle down a bit and they have scaled infrastructure

This is not a criticism of OpenAI. It is just a practical suggestion to resolve your issue.

You may not be happy - but it is realistic.

I suspect they didn’t expect the media to blow it up as huge as it did, and they got caught off guard.

Some may ask why they continue doing big deals with Microsoft etc. The reality is that they were probably in the works well before the huge media impact in Nov/Dec last year. Also companies like Microsoft are providing huge amounts of funding that allow the service to exist in the first place.

One of two things will happen with OpenAI

a) They will solve their availability issues over time, and most of us will be happy. We have to be patient.
b) They will not be able to scale for the huge load effectively. They will change the product into something different or restrict access in some way. If they do this, you might be happy that you got in early and stuck with them through the hard times.


Yes. This. I’m not an OpenAI employee either, but I have to imagine people that chargeback on a beta product … might not be invited to try future products… or be able to use the service at all? Again, that’s just my personal opinion. Chargebacks over $20 on a beta tool seem a bit extreme to me…

You have to remember that ChatGPT was the fastest growing app - ever. Well, so far, anyway. There was a graph that showed their meteoric rise compared to Insta, et al. They grew really fast really quickly. That they were able to keep it up at all was amazing.

I’ve read about people having multiple accounts, trying to trick ChatGPT, etc., which probably isn’t wise. CloudFlare seems to be doing a good job with security so far, in my opinion. I’m sure the rules will change as the products mature. Keep the hope.

ETA: This is not the graph I saw but it’s similar… The one I saw was for first 100 million users while this is for first million users…


It would be good if the admins could pin this post @logankilpatrick - At least for a week or more.

I think it is unbiased and practical. Paul’s graph and related post highlight how quickly OpenAI (and their support channel) must have got smashed - thanks to the media hype.

It appears that many people eagerly became early adopters of a limited-rollout premium service, yet they seem unable to comprehend that fixing any issues and scaling the infrastructure will require some time. Why would they sign up to support the technology as early adopters if they weren’t willing to accept that there might be some issues that would take a few weeks to resolve? Some people’s expectations are simply unrealistic.

… Now, lets rewrite that comment with a “cosmic horror” theme.

“”" It appears that a cult of devoted acolytes eagerly embraced the forbidden knowledge of the Elder Gods, but they were woefully unprepared for the dark consequences that came with the unveiling of their terrible secrets. As they delved deeper into the abyssal mysteries of the universe, they found themselves beset by a cacophony of horrors and eldritch abominations that their fragile mortal minds could scarcely comprehend.

One might wonder why they chose to delve so deeply into the forbidden lore of the Old Ones, if they were not prepared to face the unspeakable terrors that lurked in the shadows beyond the veil of reality. It is apparent that their expectations were woefully misguided, blinded as they were by the ecstatic thrill of exploring new and innovative paths of knowledge.

But now, as the veil begins to fray and the terrible truths of the cosmos come crashing down upon them, they find themselves struggling to come to grips with the scale of the horrors they have unleashed. They scream and writhe in terror as the very fabric of their reality begins to unravel, and they realize too late that the price of knowledge is often far too high to bear. “”"

1 Like

It is human nature.

The problem is not the end users.

The problem is the developers who are trying to make a quick buck cobbling together production quality products based on a “research / marketing beta” and selling this to clients.

Good system engineers would never jump and make a buck in at this point because it’s just a research beta. It’s too early.

GPT-3 API is however in production, not in beta/research.

The current uptime is quite bad.
This week it was around 75% for me.
It’s acceptable during development, but when the product will be live it may be embarrassing for me. I hope that by then the APIs will be more reliable though.

OpenAI has no competition yet, but otherwise it wouldn’t be too attractive for developers.