Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizens for the delivery of clues about the past. CTOnews.com, September 15 (Xinhua)-- Kuaishou announced today that it has begun internal testing of AI's textual graphics function-Kuaishou AI in the short video comment area. By entering a variety of creative text, users can generate a large number of style pictures with one click, making it more convenient to interact with each other in the comment area.

It is reported that this is another AIGC capability that Kuaishou explored and landed in short video scenarios after Kuaishou AI conversation. it is also the first time in the industry to apply AIGC capability under large-scale App core business scenario review area, which can greatly enhance the interactive experience.

As previously reported by CTOnews.com, Kuaishou has previously tested a number of functions based on large models: bringing new features such as intelligent question answering and text creation to users in search scenarios. It is said that if you open Kuaishou App and enter Kuaishou search for input questions, you will have the opportunity to get the information and answers provided by intelligent Q & A products.

In addition, Kuaishou's video editing products "Fast Shadow" and functional shooting products "Yitian camera" have been tested internally in late June, including "AI animation video", "AI copywriting recommendation", "AI instant universe", "AI sketches" and other AIGC creation features. These functions allow users to easily convert their photos or videos into different styles and effects of content, such as animation, watercolor, Chinese style, oil painting, etc., and automatically generate copywriting in line with the context of the content.

The large language model developed by Kuaishou is called "KwaiYii". The CTOnews.com query happy GitHub page found that the official introduction shows:

Happy Big Model (KwaiYii) is a series of large-scale language models (Large Language Model,LLM) independently developed by Kuaishou AI team from zero to one. At present, it contains models of various parameters and covers pre-training model (KwaiYii-Base) and dialogue model (KwaiYii-Chat). Here we introduce a series of 13B-scale model KwaiYii-13B, whose main features include:

The KwaiYii-13B-Base pre-training model has excellent general technical base ability, and the State-Of-The-Art effect under the same model size has been achieved on most authoritative Chinese / English Benchmark. For example, the KwaiYii-13B-Base pre-training model is currently at the leading level of the same model size in the Benchmark of MMLU, CMMLU, C-Eval, HumanEval and so on.

KwaiYii-13B-Chat dialogue model has excellent language understanding and generation ability, and supports a wide range of tasks, such as content creation, information consultation, mathematical logic, coding, multi-round dialogue and so on. The results of manual evaluation show that KwaiYii-13B-Chat surpasses the mainstream open source model, and is close to the same level of ChatGPT in content creation, information consultation and mathematical problem solving.

Related readings:

"Kuaishou search is internally testing the first intelligent Q & A product in the short video live broadcast industry."

"Kuaishou's video editing product" Kuaiying "internally tests a variety of AIGC creation functions, which can generate copywriting with one click."

"Kuaishou self-developed model" KwaiYii "was unveiled, and CMMLU ranked first in Chinese."

"based on the self-developed language model, the Kuaishou AI dialogue function has been opened for internal testing on Android."

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.