i noticed that you can now continue conversations with multiple generations within a single conversation. I say “now” because it was my experience that once I interacted with a regenerated answer that became the committed direction. I couldn’t do anything with the other generated response. I don’t know if that was everyone’s experience so this may not be true for others and it could have been specific to using the chat interface.
Anyway well I do know that I can now continue conversations with the various generations I get. And it doesn’t cement the course of the conversation. At least not as far as preventing simultaneous ongoing interaction with both generations. But it definitely creates a wall the GPT doesn’t know about.
When I mentioned the dual-threaded conversation the bot was like “Yep. I know about the regenerated images and the associated conversation. I’m cool like dat.” Nice!
When I hopped over to the sub-conversation via regenerated image for the same prompt that side of the conversation was like “What are you talking about? I’m it. There is no other side to me.” Okay.
Did a quick little test went something like this:
me: pick a number between 0 and 9999999999
me: hey what number did you pick again?
sub-conv2: i didn’t pick a number.
me: right. my bad. pick a number between 0 and 9999999999
me: okay. what number did sub-conv2 pick?
sub-conv1: yeah man idk. looks like i don’t have access to that information.
TL;DR is it appears you can create daisy chains of sub-conversations within one core conversation and the subs don’t know about each other. they’ll know about whatever path got them to where they are but they won’t know the conversation contains other paths.
I obviously found this terribly fascinating. I kind of imagine a conversation like a synapse with dendrites that can never reconnect.
Also. Tbh. A little disappointing. Silly me thinking I could trust the bot to know what it could and could not access!