So we get a Tweet (X-eet?) announcing that OpenAI has expanded memory features which “remember” your previous chats, and the first thing I do is feed it this prompt, as inspired by Search Engine Journal‘s coverage:
“Describe me based on all of our chats.”
Damn, I sound good… except…
I ask it to remove a single word from the first response and wait for my next prompt. A single word, I know where it came from, but I’m not sure why it’s so prominently emphasized here. Maybe one chat remains undeleted from that work, or there are a greater number of deleted chats relevant to that word than I realized?
“Describe me based on all of our chats.”
Different answer. Damn, I sound good. What’s a little self-indulgent, vanity on Friday? I’m loving all this auto-praise.
“Describe me based on all of our chats.”
Different answer. And once more, which causes it to choke up with “Error in message stream” after producing about 20 words.
Oh no. Please dear companion, give me more. This can’t be the end!
Without clicking “Retry”, I use the same prompt again and this is what came out, another different answer.
Phew, I sound good again.
Weirdly, it knows (ahem, instinctively?) which chats not to acknowledge, and Thank God For That!
This could get real useful. It could also get really creepy, depending on what features evolve into across larger, shared organization accounts in the future.
But hot diggity damn, this feels so much better than “Googling yourself”!