Group 1 - Tencent has recently welcomed Xu Tan, former Chief Research Manager at Microsoft Research Asia, to its Mix Yuan team, focusing on cutting-edge research in multimodal directions [2] - Xu Tan has a significant academic and industry background, with research on generative AI and content generation in audio, video, and speech, and his papers have been cited over 10,000 times [2] - Prior to joining Tencent, Xu Tan was with a domestic large model startup "Moonlight" where he was responsible for developing end-to-end speech models, indicating a shift in his career path [2] Group 2 - The exploration of multimodal research requires substantial computational power and funding, which is a heavy burden for startups [3] - Compared to emerging companies like DeepSeek, which primarily focuses on text and reasoning capabilities, large firms like Tencent and ByteDance have clear advantages in resources, ecosystem, and computational power for supporting multimodal research [3] - The Chinese large model landscape is transitioning from "wild growth to resource concentration," with early-stage startups losing their competitive edge as the focus shifts to data, computational power, and practical applications [3]
前亚研院谭旭离职月之暗面,加入腾讯混元,AI人才正加速回流大厂