Interesting, that seems plausible. Before I noticed this error, I was getting an error "cannot process two messages at once" whenever I tried to start a chat. Perhaps the root cause is some infinite loop on the frontend spamming blank messages.

How can this happen in 2023? More importantly, how can a well-funded firm like OpenAI allow such lousy database bug to occur in their system? For all the hatred towards Google, I'm sure at least they wouldn't have this kind of problems once they open their (much more powerful) LLM.

Well, if someone wants to help fix it, they're hiring security engineers [0] for $200 - 370k per year, although you'd have to move to San Francisco.

[0] https://openai.com/careers/security-engineer-detection-respo...

A similar issue happened to Steam in 2015 https://www.forbes.com/sites/insertcoin/2015/12/25/steam-is-...

It almost certainly happens after the DB layer. I would bet 10:1 that this is a cache issue, although without knowing their stack I have no idea where the cache lives.

No entity is immune from bugs.

I wouldn't bet on that. Back in 2020 Google photos had a "technical glitch" that resulted in videos getting leaked to other users.

https://www.theverge.com/2020/2/4/21122044/google-photos-pri...

Maybe that's true (and hopefully it is) but when I was seeing this behavior earlier today it showed me a series of different conversations with related, but slightly different titles- it wasn't just random unrelated ideas, it really seemed like someone was workshopping an idea over a period of time.

If you asked the language model to generate a list of chat titles in a single query, that's what it would do. It's going to try to make them seem related.

I don't know whether they do it in a single query, though.

If you dig through some of the stuff, you will quickly realize that these are histories of real chats.

I definitely saw a chat history that wasn't one I started. It was a single question and response in Japanese (which I don't know). The title in the sidebar seemed to match the conversation.

Edit: I'm getting upvoted, so I want to clarify that I don't have proof, but this did happen, only for a single chat. I dumped all the text into Google Translate (don't judge me - you'd do the same!) and it was a boring scientific question. I was amazed that it happened at all and felt pretty gross about it. Like most other people, I also saw a lot more chats in the sidebar that I couldn't open.

Check your translate history: https://translate.google.com/history

Looks like I can still comment on this thread. For future readers, this phenomena was mentioned in OpenAI's postmortem: https://openai.com/blog/march-20-chatgpt-outage

That sounds like the perfect defence even if it were cache leakage!

EVL Co: "It's just the AI making stuff up. Any resemblance to a real conversation or contact information is purely coincidental."

User: "but I contacted the email and the person's name matched... and she said she had just used your product and that these inputs I listed for her sounded familiar".

EVL Co: "Look, you just don't understand statistics is all. Trust us. Coincidence. Happens all the time. No breach to see here... Press one to complete our survey regarding this automated customer service call."

It would be fascinating if OpenAI's random chat topics were Chinese Communist propaganda like in screenshots I saw.