wary light
Hi all, so this is an idea I’ve had a few weeks ago that I never put to paper until now. When I saw Neuro struggling to remember her hype train record, I decided to start thinking about how one would solve this problem (even though I know next to nothing about AI).
Eventually, I came to a very silly conclusion: use SQL (or dictionaries, or databases, or whatever it’s called in the programming language you’re using.)
You could have a layout like the one shown in the image:
The way I envision it working would be, reading this from memory, Neuro would immediately now know that her relation with Hype Train's is that she got the world record for biggest hype train 3 times, and two of those times were in the last month. The last accessed date would make it work like human memory: if she hasn’t accessed a memory for a long time, it’ll be purged from the DB to keep it small (so hopefully it can be loaded in RAM for lower latency.)
I can also imagine this could be used for her to remember details about certain chatters.
Hell, I think SQL can even have images in the databases, but it’s been years since I messed around with sqlite and I never needed to use that feature.
Now, I’m going to be honest, like I said, I know nothing about AI so I’m not sure how Neuro would write to this, or how the auto-purging of memories would work, but from how quickly I’ve seen Neuro read, I think that this at least has a chance of being feasible.
floral venture