Comment by swatcoder
4 days ago
The "open" in "open source" is traditionally about respecting a user's right to modify a library/application to suit their needs. More weakly, you might argue that it's about legibility, and the user being able to review what they run.
The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
Do weights provide either of those qualities?
> The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
> Do weights provide either of those qualities?
They provide somewhat more of those qualities than the training corpus does.
Not a lot, especially for "understanding", but more.
You don't need the previous training material to customize the weights.
I don’t need the source code to randomly change bytes in the compiled Linux kernel binary either.
Source code makes it easier to modify a binary, training data doesn't make it easier to modify a trained network.
Fine-tuning weights is easier than retraining the foundation model from scratch with a different corpus.