We are committed to Open Models @StabilityAI
— Emad (@EMostaque) February 27, 2024
We have given over 20 million A100 hours in grants to open AI research & projects across communities.
Stable Diffusion 3 will have an open release.
Same with video, language, code, 3D, audio etc.
Open collaboration is the way 🚀
MOAR FREE CODESHIT FRICK YEA
Jump in the discussion.
No email address required.
Help me understand as a normie. Does this mean their ai algorithms will not be filtered through globohomo and DEI? Or does it just mean that they'll make the algorithim public for everyone to see and use at their discretion?
Jump in the discussion.
No email address required.
Number 2. All their post 1.5 models are censored as heck, so the point they all fail with the prompt "Naked woman"
Jump in the discussion.
No email address required.
lame
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Both, kind of.
A lot of current AI restrictions are accomplished through pre-prompting: when you ask GPT or Gemini for something, before they give the AI your request they give it a lot of their own input, like "make sure to show people of all races" or "don't make scenes of graphic violence".
That approach won't work here, because if they do do it with an open model, it's trivial to make the same model with the pre-prompt edited down to remove all that stuff.
This doesn't mean it can't have its own biases from training or have some other censoring methods, but Google would really like to be able to have that for Gemini and they still had to resort to pre-prompt, so barring some major breakthrough only Stability made, SD3 should be much less cucked than its competitors.
Jump in the discussion.
No email address required.
More options
Context
More options
Context