Stability AI’s Stable Diffusion 3 Medium has faced criticism for prioritizing the happiness of US puritans over accuracy in image synthesis. The model’s attempt to filter out adult content has resulted in anatomically incorrect and visually disturbing images. This censorship-first approach has led to the model struggling to accurately depict human bodies, with users on Reddit sharing examples of its failures. Previous versions of the model also faced similar issues, prompting Stability AI to reverse course and include NSFW content in training data for better results. Users have reported receiving bizarre and unsettling images when using the model, such as a man with giant-sized backward hands in response to a prompt. The model’s limitations in understanding human anatomy have been attributed to its strict filtering of adult content. Despite attempts to improve the model with updates like SD 2.1 and SD XL, it still struggles to generate accurate human figures. Overall, the focus on censorship has hindered the model’s ability to produce realistic and coherent images, leading to disappointment among AI image fans.
Source link
Source link: https://www.fudzilla.com/news/ai/59167-stability-ai-messed-up-its-own-ai
AI’s own AI fails again, stability compromised #AIstability
Preeti
637 Views
0 Votes
GIPHY App Key not set. Please check settings