Home > Technology peripherals > AI > Kuaishou internally tests 'AI Xiaokuai', which can use Kuaiyi's large model to achieve intelligent replies in the comment area

Kuaishou internally tests 'AI Xiaokuai', which can use Kuaiyi's large model to achieve intelligent replies in the comment area

WBOY
Release: 2023-10-26 20:21:05
forward
2517 people have browsed it

On October 26, Kuaishou officially tested the "AI Xiaokuai" account internally, providing users with a variety of AIGC services based on short video application scenarios such as interactive Q&A, image generation, strategy retrieval, and copywriting modification in the short video comment area. It is reported that “AI Xiaokuai” uses the large-scale language model “Ruyi” large model independently developed by Kuaishou from scratch as its base. Since this year, Kuaishou has made frequent moves in the AIGC field. This is also the first personalized application that Kuaishou has implemented in the comment area scene.

The short video comment area is naturally a public space for users to express, interact, and communicate. There are users’ demands for expression and questions under each category of content, and users’ questions are of various types, covering discussion and exchange of video content, professional opinions, etc. Doubts about information and expression of self-emotions, etc. In Kuaishou's community ecosystem of 376 million DAU, every expression and every question from a user is a sign of trust in the community, and behind it is the expectation of "being responded to."

In order to better meet users’ demands in the comment area and allow users to express their opinions, interact with responses, and interact with surprises, “AI Xiaokuai” came into being. It is understood that users can apply to participate in the internal test by searching "AI Xiaokuai" on Kuaishou. Qualified users can post comments from Aite @AI Xiaokuai in the short video comment area and receive responses from @AI Xiaokuai. "AI Xiaokuai" can not only provide effective answers and information references for users' questions, but also can effectively improve the frequency of interaction and information interaction experience between users based on the extension of short video information.

It is worth mentioning that "AI Xiaokuai" also has its own personality, that is, a Siyou AI account with attitude, ability, emotion, and motivation. Whether it's chatting about the details of binge-watching TV series or solving mathematical and physical problems, as long as users leave messages in the comment area, "AI Xiaokuai" will answer all questions. At the same time, AI Xiaokuai is also a friendly and lively chat partner. He can not only provide effective emotional support, but also "create memes" at any time based on information such as network hot spots and current affairs hot spots, striving to attract attention in the comment area

快手内测“AI小快” 可借助快意大模型在评论区实现智能回复

快手内测“AI小快” 可借助快意大模型在评论区实现智能回复

快手内测“AI小快” 可借助快意大模型在评论区实现智能回复

It is understood that behind "AI Xiaokuai" is the support of the powerful understanding, generation and interaction capabilities of the Ruiyi large model. The capabilities of the "AI Xiaokuai" released this time have been upgraded to the 66B version of the Ruiyi large model. This is also another technical upgrade after Kuaishou released the Ruyi 13B model in August. Larger models also bring more powerful capabilities

快手内测“AI小快” 可借助快意大模型在评论区实现智能回复

Since the second half of last year, Kuaishou has accelerated its layout of AIGC. In August, Kuaishou announced the start of internal testing of its self-developed large language model "KwaiYii", aiming to create a full-modal AIGC solution. This model is based on a self-developed large base model and provides users with multiple technical capabilities such as text generation, image generation, 3D material generation, audio generation, and video generation. Among them, KwaiYii-13B has performed at a leading level in multiple Benchmark tests, proving its excellent performance in natural language processing tasks

Based on the current technology accumulation in large models, Kuaishou has implemented applications in multiple scenarios such as search and comments. On July 8, Kuaishou started the internal test of its intelligent question and answer product; on August 8, it launched the internal test of "AI Conversation", and on August 18, it opened the internal test of the "Kuaishou AI Conversation" function in the Android version of the Kuaishou App; in September On the 15th, Kuaishou announced that it would start an internal test of the AI ​​text-based picture function for short video comments on the site - Kuaishou AI Play Commentary. By inputting various creative texts, users can generate a large number of styled pictures with one click, making it more convenient for interesting interactions in the comment area.

As the commercial implementation of AIGC ushered in a new wave of peaks, the integration of technology and industry and the richness of scenario applications are rapidly increasing. Short videos have a huge user base and strong social attributes, providing more application scenarios for the implementation of AIGC. In the future, Kuaishou will continue to delve into AI technology, promote the development of product forms and user experience, explore the second curve of the short video business, and promote it to become the infrastructure of the short video live broadcast platform

The above is the detailed content of Kuaishou internally tests 'AI Xiaokuai', which can use Kuaiyi's large model to achieve intelligent replies in the comment area. For more information, please follow other related articles on the PHP Chinese website!

source:sohu.com
Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn
Popular Tutorials
More>
Latest Downloads
More>
Web Effects
Website Source Code
Website Materials
Front End Template