2 ms·
I'd like to know if there's been any progress in the "memory" area of GPT. As far as I know, when you tell GPT to do something it runs the entire sequence of pr
by causality0 4y ago
I'd like to know if there's been any progress in the "memory" area of GPT. As far as I know, when you tell GPT to do something it runs the entire sequence of prompts over again from the beginning. This is obviously not practical for any applications that require ongoing collaboration, like using GPT integrated with your computer to execute commands and policies, or training GPT to be a better coding partner.
- grogenaut 4y agoWhy is it obviously not practical? Seems to be working in many places. Or is your issue that the memory is too small?
- causality0 4y agoIf it has to re-run the entire history every time you interact with it it will quickly slow to non-functionality in days.
- grogenaut 4y agothat's not what happens... basically it is using a ring buffer for the history. It only runs apparently (from another post today) 4k for chatgpt4 right now with 32kb tokens coming in beta. everyotype of memory tho maybe not as limited as this.ne I think has this This is why you see it start getting wierd after a bit, it loses the history and you've basically started a new different conversation. However