April 21, 2011 - China Telecom today announced that itthe wing (of a plane) AI Open Platform(https://www.teleai.com.cn) is now online and open to the public.
According to China Telecom's official introduction, individual users can use China Telecom's self-research through the platform for one-stop useStar Models Q&A assistant, mega table analysis, Star Wise Pen, multi-dialect recognition, and text-to-graph capabilities.
Officially, government and enterprise users can experience online the development platform capabilities such as big model development, AI center, data center, Star Ocean PaaS platform, Star Intelligent Body platform, and nearly 30 core products such as intelligent applications, big data, solutions, etc., which can satisfy the digital innovation needs of the governmental, financial, industrial and other scenarios.
Note: Star Big Model is a fully localized trillion-parameter big model independently developed and built by China Telecom, which achieves breakthroughs in four major areas: semantics, speech, vision and multimodality.
According to the introduction, Star Speech Big Model breaks through the industry problem of single-model multi-dialect mixing, and can support 50 kinds of dialect mixing recognition. It also supports supernatural speech generation technology, which can realize high simulation reproduction, 5-10 seconds of speech can accurately match the tone of the speaker, and the similarity is more than 90%.
With the big model of Star Vision, China Telecom is the first in the industry to propose the knowledge-based deployment and control of everything, empowering millions of road cameras nationwide, with an average daily call volume of more than 600 million times, which are widely used in transportation, government affairs, emergency response, urban governance and other professional fields.
As the first open-source central enterprise big model, Star big model realizes full-size coverage of billion-, ten-billion-, and hundred-billion-parameter models, with a cumulative download volume of more than 40,000 times.