@tante @noodle
Uhm, there are things I mostly agree, other not so much. For example, I do not think that data is that important. After all, if you have code, you have the recipe. And if you have the weights you can use lora with your data. You do not need the original data for that. I also disagree with your perspective on deep neural networks as glorified matrix multiplication.
https://mastodont.cat/@remenca/113561439882713414
No replies.
────
────