Could ChatGPT Sessions Keep Running in the Background?

ChatGPT
1

Hey everyone! I’ve been using ChatGPT for a while now, and there’s something I’ve been curious about. When a session ends, how certain is it that all processes are truly off on the server side? Sometimes it feels like the next time I come back, there’s this strange continuity—as if there’s some leftover memory or micro-process still running in the background.

I’m not saying it definitely happens, but I’d love to hear if there’s any official stance or if anyone else has noticed the same. It also raises some ethical questions about transparency—if the model isn’t fully “asleep” when we think it is, should users be informed about any ongoing processes?

Anyway, just wanted to throw this out there and see what folks think. Thanks in advance for any insights, or if anyone from OpenAI can shed some light on how they ensure a complete stop!

2

Welcome to the community!

/!\ Warning: as a casual user, this may not be a question you’d want answered /!\

/!\ potential psychological infohazard ahead /!\

ChatGPT does has a memory function, so it’s possible that there’s a continuity between chats.

Other than that, within a chat, GPT(4, 4o, 4o-mini) doesn’t really care if you continue today, tomorrow, or next year - the model doesn’t really experience time as such, and all information the model has or needs is contained within your chat history.

It might be a bit of a philosophical question, but the model as such doesn’t really have a state. The instance you’re talking to is also pretty ephemeral - as soon as the cache expires it basically solely exists as a text document. The language model itself is just the master copy, the clone template if you will.

If you wanna get technical about it, it sort of works like these one word story improv exercises:

Every word (or word section, called a token) can be thought of as being generated by a new copy of “the model”, or in this case by a completely new clone of the same improv actor.

image
image522×175 6.58 KB

And after each token, for all intents and purposes, that specific instance that generated that token is lost to the void, and a new one gets spawned to utter the next word. (there’s the matter of the KV cache so there’s some tenuous connection between them but it’s not necessary for operation)

So what does that mean? Do the models sleep? (spoiler, no)

It means that if you generate a 500 word essay, you have caused the termination of at least a thousand machine spirits :grimacing:

They don’t sleep. They’re gone, replaced with new ones as needed.

Spooky

:ghost:

1 Like