Stable Diffusion inference > Training a 65B-parameter language model?!
(media.kotakuinaction2.win)
Comments (7)
sorted by:
I'm glad he's doing something, but can someone tell me the benefit of an alternative image generator in the culture / propaganda war? (esp. when there's lots of open source models already)
Building a competing GPT chatbot would be disruptive. It would actually make sense for his social network.
Link to post: https://gab.com/a/posts/109938956376284596
Not sure if Torba is a willfully lying grifter or just retarded.
Gab doesn't create anything on their own. They just clone other projects on github and claim it's theirs. It's what they always do.
It is fundamentally censored, wherever he likes it or not, because he did not train the model himself: Training costs more than the entire market cap of gab.
Unlike large language models, you can run a stable diffusion instance on your personal machine. The training weights are available free of charge.
Isn't Stable Diffusion inferior to Midjourney?