What are your thoughts on this disparity in content filtering priorities between different regions?
Lastly, would you be interested in attending TechCrunch’s upcoming event in Boston, MA, where industry visionaries will share insights on building smarter and scaling faster in the tech sector?. For instance, images of Xi Jinping, Tiananmen Square, Tank Man, the Taiwanese flag, and symbols supporting Hong Kong liberation are blocked from being uploaded. Have you ever experienced content filtering like this on other platforms?
In China, models like Magi-1 must adhere to strict information controls, as mandated by a 2023 law that prohibits the generation of content that could disrupt national unity or social harmony né?. The company claims that Magi-1 can produce high-quality videos with more accurate physics representation compared to other similar open models né?. Do you think such restrictions reflect a broader trend in the tech industry, or are they unique to the Chinese market?
It’s worth noting that while Chinese models may limit political speech, they often lack filters for preventing the generation of non-consensual nude images, as reported by 404. Have you heard about the new video-generating AI model released by the China-based startup Sand AI? It has gained praise from prominent entrepreneurs like Kai-Fu Lee, the founding director of Microsoft Research Asia né?. However there are reports that Sand AI is filtering politically sensitive images on its hosted platform possibly to comply with Chinese regulations né?. But did you know that this model is so large with 24 billion parameters that it requires between four and eight Nvidia H100 GPUs to run effectively?
Interestingly Sand AI’s platform includes filters that prevent certain images from being used as prompts for video generation né?. This often leads Chinese startups to implement filtering mechanisms to comply with these regulations. Have you encountered any restrictions like this before?
Sand AI recently unveiled Magi-1, a cutting-edge model that uses a unique method to generate videos by predicting sequences of frames. How do you feel about the balance between content censorship and freedom of expression in AI models?
Moreover, other Chinese startups, like Hailuo AI and Shanghai-based MiniMax, also restrict the use of politically sensitive images on their platforms
