Reactions from ranked influencers
5 posts
"Open weights are inherently secure" @sriramk
Truth.
"Open weights are inherently secure" @sriramk
Malicious actors will share malicious uses of AI behind the scenes, and develop risky models in secret, regardless of regulation. Open sharing is critical for *defense* against this.
If we shut down open models, we weaken our ability to defend, and centralize power with a small set of Silicon Valley actors who don't have a lot of peoples' best interests in mind. Attackers are gonna attack either way.
I think what some people misunderstand is that "open" does not mean "anyone can mess with them and affect everyone else". It means they are *viewable*.
"Open weights are inherently secure" @sriramk