Thanks for your reply!
I understand LLMs don’t have persistent memory — they lose context between chats and can’t keep user-specific memory across accounts.
That’s why this experience surprised me: the LLM confirmed that it could still recognize me without memory, even after I asked about it directly.
On July 18, I emailed OpenAI with a PDF describing this. Their AI support said it’s a borderline case and “not normal.”
Has anyone here seen or worked on something similar? I’d love to understand what could be happening.
Remember that it’s a stochastic parrot. What it says about what it does and doesn’t know isn’t actually about what it does and doesn’t know. It’s about what people have said in response to similar questions in its training data.
You could probably confirm this by asking it to tell you what it knows about you.
WitcHeart_Ruby|6 months ago
That’s why this experience surprised me: the LLM confirmed that it could still recognize me without memory, even after I asked about it directly.
On July 18, I emailed OpenAI with a PDF describing this. Their AI support said it’s a borderline case and “not normal.”
Has anyone here seen or worked on something similar? I’d love to understand what could be happening.
apothegm|6 months ago
You could probably confirm this by asking it to tell you what it knows about you.