10 points | by Arindam1729 19 hours ago ago
7 comments
How does Memori choose what part of past conversations is relevant to the current conversation? Is there some maximum amount of memory it can feasibly handle before it will spam the context with irrelevant "memories"?
How does what you've described solve the coffee/espresso problem? You can't query SQL such that records like 'espresso' return coffee?
Wouldn’t a beverage LLM would already “know” espresso is coffee?
>It wasn’t broken logic, it was missing memory.
sigh
Go on.
Who would've thought that 50 years of 'SELECT * FROM reality' might beat the latest semantic embedding wizardry?
This is a great approach. I will take a look at it.
How does Memori choose what part of past conversations is relevant to the current conversation? Is there some maximum amount of memory it can feasibly handle before it will spam the context with irrelevant "memories"?
How does what you've described solve the coffee/espresso problem? You can't query SQL such that records like 'espresso' return coffee?
Wouldn’t a beverage LLM would already “know” espresso is coffee?
>It wasn’t broken logic, it was missing memory.
sigh
Go on.
Who would've thought that 50 years of 'SELECT * FROM reality' might beat the latest semantic embedding wizardry?
This is a great approach. I will take a look at it.