Great product, great communication but: When you deep dive in their stuff, it doesn’t looks like they means it. It makes a great culture as well, can’t disagree with that and yet …
It started with Mythos. Awesome name indeed.
The communication was about that it was so good that it couldn’t be made available to the public. That may be true but the technical was obviously that it was because it had no proper safety net. And fable proved it months later.
Also, that was after the clash with the US government when Anthropic refused to remove the safety net. They did a good PR about how they held up to their own standard. And their next action is to make a model that exactly lack this safety and made it only available to selected entities ? Oof… they always have this double-standard.
They always emphasis safety research but provide no responsible disclosure or safe harbor about safety issue in their model, it’s only for cyber, not for AI.
I mostly use use Claude, probably will for a long time but always testing alternative and ready to jump at a moment’s notice because of the dichotomy between what they say and what they do. Yes, even a Chinese model if i must.
Top comments (0)