Tag: Natural language processing

Reports indicate that Xiaomi is expanding its own GPU computing cluster, with plans to invest in building its own large-scale natural language model to drive the development of its AI application services. Sources suggest that Xiaomi began this project several months ago, spearheaded by CEO Lei Jun. However, Xiaomi has not responded to these rumors. Prior to this, Xiaomi had already invested resources in AI application development, including the early launch of Xiao Ai (Xiaomi's AI assistant), the establishment of an AI team in 2016, and later investments in autonomous driving technology. In April 2023, Xiaomi established its AI Lab's large-scale natural language model team, recruiting numerous technical talents, including Luan Jian, former researcher at Toshiba China Research Institute, senior speech scientist at Microsoft China Engineering Institute, and chief language scientist for Microsoft Xiaoice, who serves as the head of Xiaomi's large-scale natural language model team. Xiaomi has already built a large-scale natural language processing (NLP) model, MiLM-13B, with 1.3 billion parameters on its mobile devices. In specific scenarios, its computational performance is close to that of the 6 billion-parameter MiLM-6B NLP model deployed in the cloud. This large-scale NLP model is also applied to the new version of Xiao Ai, enabling more natural conversational interaction. Over the past seven years, Xiaomi has undergone six expansions, increasing its technical staff to over 3000. Its AI technologies are applied across vision, acoustics, speech, NLP, knowledge graphs, machine learning, large-scale NLP models, and multimodal operations, and are used in mobile phones, smart IoT devices, robots, and vehicles.

Since announcing the introduction of Clova AI applications in 2020, LINE has continued to promote the development of AI technology applications in Taiwan. At the Techpulse 2023 Developer Conference held in Taiwan, LINE further explained the progress of its self-developed natural language processing platform, SmartText 2.0, and also revealed its future development towards automatically generated content technology. By combining natural language processing capabilities with elements such as voice, images, and videos, LINE aims to create more suitable interactive content. Since the most crucial aspect of AI lies in the semantic understanding of user interactions, especially since LINE services primarily rely on text-based interactions, understanding the user's intent through text has become a key focus for many LINE services. To more efficiently utilize natural language processing models, the LINE engineering team proposed the concept of "Natural Language Processing as a Service" (NLPaaS), standardizing the construction process of NLP applications and rapidly building different NLP models. This accelerates the understanding and processing efficiency of natural language content, leading to the development of its own SmartText natural language processing platform. The initial version of the SmartText natural language processing platform already handled functions such as document classification, multi-label classification, and topic detection. The newly announced SmartText 2.0 platform adds the ability to perform diverse natural language processing tasks, including article summarization, paraphrasing, question-and-answering, and customized advertising copy generation, thus addressing more complex automated text processing needs. According to the LINE team, SmartText 2.0 will continue to evolve towards faster content generation, creating more interactive content with users through automated generation.