Henrik: I’m going to create word embeddings for Karl Ove Knausgaard's book Inadvertent. We are going to use the word embeddings as your memory of the book.
This was very cool. The idea of how we could extend enhancements to the model, so its able to create a sense of structure re how to use various methods to enhance its memory/ knowledge base/ questions, is likely to be extremely important going forward.
Yeah, it seems like a promising path, rather than trying to bake everything into the model. Though that might work too, who knows what emerges. But there are a lot of interesting experiments integrating LLMs into larger structures, extending their powers. Very fascinating.
It would be interesting to integrate these types of ideas into your Roam plugin, have it do semantic search of the graph for example. Or add other tools it can access, like calling the Wikipedia API, or whatever fun idea you can come up with.
This was very cool. The idea of how we could extend enhancements to the model, so its able to create a sense of structure re how to use various methods to enhance its memory/ knowledge base/ questions, is likely to be extremely important going forward.
Yeah, it seems like a promising path, rather than trying to bake everything into the model. Though that might work too, who knows what emerges. But there are a lot of interesting experiments integrating LLMs into larger structures, extending their powers. Very fascinating.
awesome work!
It would be interesting to integrate these types of ideas into your Roam plugin, have it do semantic search of the graph for example. Or add other tools it can access, like calling the Wikipedia API, or whatever fun idea you can come up with.
definitely a great direction to explore!
another great one Henrik! Keep ‘em coming!
Thanks, k3n! I've been a bit slow so far. But I'm going to do two a month from now on! That'll be fun.