Artificial Intelligence thread
- Thread starter 9dashline
- Start date
Sensetime deploys HyperPPL for multi-modal large model in cars.
8B parameter model using Nvidia Orin
i appreciate the discussion here on Deepseek by Steve Hsu and would like to read further into it myself
405B param Llama-3 LLM is out and I see no reason why it wouldn't be as good as other top of the scale closed model
Huawei AI is used to protect South African Railway from thief's
Figuring out use case and business model for new technology looks to be one of Huawei's strong suits. Honestly that might be one of the most valuable company traits this century. I'm predicting HW will be one of the first companies to make AI profitable.
Zhipu AI launches AI-generated video model Qingying: Generates 6 seconds of video in 30 seconds, increasing inference speed by 6 times.
Zhang Peng, CEO of Zhipu AI, announced at Zhipu Open Day that the AI video generation model Ying was launched on Zhipu Qingyan , which only takes 30 seconds to generate a 6-second video. From now on, all C-end users can experience the AI video generation capabilities of text and image through Ying.
According to reports, after entering a piece of text (Prompt), users can choose the style they want to generate, including cartoon 3D, black and white, oil painting, movie feeling, etc., and add the music that comes to finally generate a video clip full of AI imagination. In addition, Ying also brings more new ways to play, including emojis, advertising production, plot creation, short video creation, etc.
Zhang Peng, CEO of Zhipu AI , said: "With the continuous iteration of algorithms and data, I believe that Scaling Law will continue to play a powerful role." It is reported that the video generation model of the Ying base is CogVideoX , which can integrate the three dimensions of text, time and space, and refers to the algorithm design of Sora . At the same time, it is also a DiT architecture. By strengthening its complex instruction compliance ability, content coherence, and large-scale screen scheduling optimization, CogVideoX has increased its reasoning speed by 6 times compared with the previous generation ( CogVideo ).
1.6% is well within the rate of error. This is not a problem. 272 gpu failures is expected. Rest of the failures are software and networking failure which are still with margin of error.