I am wondering how much stuff (in terms of malicious behavior) one can hide in the (open) weights?
Certainly a lot less compared to closed weights :)
That is what I am not sure about...It is like, is having a public key closer to hacking the account than not having it? Theoretically, it is, but practically, it is not...
This is silly, people do inspect weights and prune/abliterate/whatever them
Also open weight models reveal their thinking, not to mention you can airgap the network around them.
Certainly a lot less compared to closed weights :)
That is what I am not sure about...It is like, is having a public key closer to hacking the account than not having it? Theoretically, it is, but practically, it is not...
This is silly, people do inspect weights and prune/abliterate/whatever them
Also open weight models reveal their thinking, not to mention you can airgap the network around them.