i get the appeal of building automemory systems but i think they are actually bad products, both in current practice (they get things wrong too often) and in principle (they train users in helplessness rather than agency)
𝚟𝚒𝚎 🌱
𝚟𝚒𝚎 🌱18.7. klo 11.36
I would really advise openai to just disable the memory system right now. Allow users to control a larger portion of the context window between chats, if necessary, but people seem to just be sliding into psychosis as a result of compounding accidental psychotic-token-hits.
none of the existing ai products, even prosumer ones, give users adequate control over what’s in the context. and most of them don’t even allow users the basic transparency of seeing what’s in there
this is training a whole generation of users to treat the models as a magic box that thinks for them instead of a powerful, capable tool that can extend their agency. and the consequences will be disastrous for society
1,71K