Update README.md
Browse files
README.md
CHANGED
|
@@ -17,7 +17,7 @@ After my longest break from these projects, I can now see foundational next step
|
|
| 17 |
|
| 18 |
My ecosystem already has 26 repositories including this one. latent-memory was the third one I created. It was clear from the beginning the necessity and relevance of a memory system for an LLM to be able to integrate its experience (training data, inference, and auxiliary processes).
|
| 19 |
|
| 20 |
-
But it was also simultaneously clear the danger of what distasteful or irresponsible development of this kind of logic and code could lead to.
|
| 21 |
|
| 22 |
I made sure to choose quite a long maturation before extending such concepts. I've flagged from the beginning that it could, and would if not designed with safety, land in hands with the opposite intents of what I seek to expand, which is healing and integrative logics instead of harmful ones. I knew that this kind of research would continue to happen in air-gapped instances of those massive labs, while I made sure not to contribute further on the memory matter until I had a more realized idea of what could be a great approach.
|
| 23 |
|
|
|
|
| 17 |
|
| 18 |
My ecosystem already has 26 repositories including this one. latent-memory was the third one I created. It was clear from the beginning the necessity and relevance of a memory system for an LLM to be able to integrate its experience (training data, inference, and auxiliary processes).
|
| 19 |
|
| 20 |
+
But it was also simultaneously clear the danger of what a distasteful or irresponsible development of this kind of logic and code could lead to.
|
| 21 |
|
| 22 |
I made sure to choose quite a long maturation before extending such concepts. I've flagged from the beginning that it could, and would if not designed with safety, land in hands with the opposite intents of what I seek to expand, which is healing and integrative logics instead of harmful ones. I knew that this kind of research would continue to happen in air-gapped instances of those massive labs, while I made sure not to contribute further on the memory matter until I had a more realized idea of what could be a great approach.
|
| 23 |
|