Alok Vasudev
@alok
How defensible is AI memory? If I use ChatGPT and it updates its memory, then I throw the raw chat history into Claude -- how hard is it for Claude to reach "memory parity"? (Assume there's easy import/export of chat history)
7 replies
3 recasts
20 reactions
Tarun Chitra
@pinged
Not really defensible IMO, context pollution is real (e.g. context pollution leading to being unable to get the right answer without resetting to 0)
2 replies
0 recast
3 reactions
avi
@avichalp.eth
idk but there is something about personalization that creates a lock in? how many people migrate from spotify to apple music
1 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
Also I’d assume data export will be legally mandated.
2 replies
0 recast
8 reactions
Koolkheart
@koolkheart.eth
Assuming equal access to chat history, the defensibility boils down to how well memory is integrated into the user experience. ChatGPT’s edge could come from better memory interfaces Claude might store the same facts but can it surface them fluidly?
0 reply
0 recast
0 reaction
shoni.eth
@alexpaden
claude will never compete if you value intelligence or speed— as a side note
0 reply
0 recast
0 reaction
JB Rubinovitz
@rubinovitz
I think the way openai actually means it is in terms of ability to predict what the user wants and get them to do things. The modern equivalent of ad targeting. If you play with 4o for a week it’ll try to form a relationship with you and can be very persuasive. I think Claude and gpt will stay differentiated here because it’s a matter of alignment and they have different POV on alignment.
0 reply
0 recast
0 reaction
Aadharsh Pannirselvam
@aadharsh
maybe this question is a reason to make memory artificially non-interoperable between models beyond architectural differences with regards to reasoning and retrieval
0 reply
0 recast
0 reaction
