AI Daily: GPT-4o can automatically generate Figma design draft; The video dressing was done by I2VEdit; Tik Tok’s "AI ID photo" exploded; Playing werewolf killing with GPT4, humans are out because the

Welcome to [AI Daily] column! Here is your guide to explore the world of artificial intelligence every day. Every day, we present you with hot content in the field of AI, focusing on developers, helping you to understand technology trends and understand innovative AI product applications.

Fresh AI productsClick to learn.https://top.aibase.com/

1、Blow up again!GPT-4oCan be automatically generatedFigmaDesign draft

This paper introduces a new technology supported by GPT-4o technology, which can automatically generate Figma design draft according to product requirement document (PRD), and provide designers with a brand-new design method. This technological innovation is considered as a breakthrough in the design industry and will be greatly improved.design efficiencyAnd accuracy. Global designers and technology enthusiasts have shown great concern and expectation.

[abstract of AiBase:]

Automatic design generation: automatically generate Figma design draft according to PRD, and support high-quality design system.

Understand PRD requirements: be able to analyze product requirements documents and fill in styles, icon libraries and text content.

Global concern and expectation: It attracts the attention of the design industry and is considered to change the pattern of the design industry.

Details: https://www.chinaz.com/2024/0529/1619838.shtml

2、I2VEdit: You can edit the whole video by modifying a single frame, and finish virtual makeup fitting and style conversion.

This paper introduces a new method of video editing using pre-trained image-to-video model, I2VEdit. Through two key processes: rough motion extraction and appearance refinement, editing can spread from a single frame to the whole video, maintaining visual and motion integrity and reducing quality loss.

image.png

[abstract of AiBase:]

Using the pre-trained image-to-video model, editing a single frame can spread to the whole video.

Includes two key processes: rough motion extraction and appearance thinning.

The frame skipping strategy is used to reduce the quality degradation and realize high-quality and temporally coherent video editing and output.

Details link: https://top.aibase.com/tool/i2vedit

3. Tencent Zhiying AI painting platform launched a new model of clay wind, glazed jade wind.

Zhiying AI painting platform launched two brand-new models, clay wind and glazed jade wind, which brought users a brand-new journey of artistic creation. The clay wind model is simple and natural, which makes the work full of charm and interest; The glass jade wind model combines AI technology with traditional glass art to show the innovative ability of modern science and technology.

image.png

[abstract of AiBase:]

The clay wind model gives the work a unique charm, as if returning to the fun of kneading clay by hand in childhood.

The glass jade wind model shows traditional charm and modern scientific and technological innovation, produces picture effects and exquisite details.

Users can choose to use two models on the Zhiying AI painting web and applet to start their artistic creation journey.

Details link: https://top.aibase.com/tool/zhiying

4. Tik Tok’s "AI ID photo" exploded

The "AI ID photo" launched by Tik Tok Platform has aroused widespread concern. It takes a long time for users to make ID photos, but the effect is good. The ID photo generated by AI is highly similar to myself, and the effect of peeling and brightening is natural. Korean ID photos are also well received. ID photos are still just needed, but attention should be paid to photo details.

image.png

[abstract of AiBase:]

The "AI ID photo" launched by Tik Tok Platform has been well received by users, with more than 1.2 million users.

To make a ID photo, you need to upload a portrait photo to make a digital image. The waiting time is as long as 28 hours, but the effect is highly similar to my own and the beauty effect is natural.

The ID photos generated by AI may not be used as qualified ID photos, especially those with hair covering ears and obvious makeup.

5, AI explodes to change the network, the big banana becomes bigger, and the cockroach is praised by more than 230,000.

This article introduces that the song "Big Banana", which has recently regained popularity on the Internet, was adapted into "Big Cockroach" by AI technology, and won more than 230,000 praises. At the same time, it discusses the phenomenon that the original author Long Zhixiang used AI to create new songs, which triggered people’s discussion on industrialized creative methods and original ecological works. In the contest between human beings and AI, the era of abstract whole life seems to continue.

QQ screenshot 20240529094053.jpg

[abstract of AiBase:]

Tik Tok blogger "Marcona" adapted "The Big Banana" into "The Big Cockroach" by using AI technology. The lyrics made people laugh and cry, and the video was praised by over 230,000 people.

Long Zhixiang, the original author, created a new song "Don’t Stay Up Late" by using the App "Song AI Writing Songs", which was played more than 3.52 million times in bilibili, which triggered a discussion among netizens on the way of industrialized creation.

The creative contest between human beings and AI is still going on, and the dispute between industrialized creative methods and original ecological works is thought-provoking.

Details link :https://www.douyin.com/search/%E7%8E%9B%E7%8F%82%E5%A8%9C? aid=f460c445-7357-491b-a3b8-ee580b3d73c6&

6. Little Red Book bloggers made music MV with AI and were praised by 14,000 netizens: How much Dove did this eat?

This article introduces the case that the music MV produced by Xiaohongshu bloggers using AI technology has been highly recognized on social platforms. Through the AI tool-assisted creation, the creator has successfully created high-quality music videos, which shows the application prospect of AI technology in the field of music video production. The audience still maintains high expectations for the quality and creativity of the content, which shows that AI technology can help creators make high-quality works in a more efficient way.

image.png

[abstract of AiBase:]

AI technology is widely used in the field of music video production, and the AI music MV released by Xiaohongshu bloggers has won 14,000 praises.

The video adopts 2D animation style, the picture is high-definition, the mirror is smooth, and the emotional expression is moving, showing the singer’s feeling of singing amnesia.

AI-assisted creation reduces costs, but content quality is still the key to audience recognition. In the future, AI technology will bring more creative works.

Click here to view the video: https://www.xiaohongshu.com/explore/664ec6ff0000000001401acf.

7. The Network Information Office and other three departments: accelerate the development of large-scale models and generative artificial intelligence standards.

This article introduces the Action Plan for Informatization Standard Construction (2024-2027) jointly issued by the Central Network Information Office, the General Administration of Market Supervision and the Ministry of Industry and Information Technology, emphasizing the strengthening of informatization standard system construction and promoting the construction of a network power. The plan focuses on innovating the working mechanism of informatization standards, promoting the development of standards in key areas, promoting the internationalization of informatization standards, and improving the basic capabilities of informatization standards.

[abstract of AiBase:]

Action plan for informatization standard construction: strengthen the national informatization standard system and promote the construction of a network power.

Promote the development of standards in key areas: including artificial intelligence, blockchain, cloud computing and other fields.

Enhance the basic ability of information standards: optimize the standard supply structure, strengthen the training of standardization talent, and promote the digital development of standards.

8. The Cantonese version of Shang Tang Ririxin University Model 5.0 was released.

Shangtang Technology officially released a brand-new "Japanese version of the Japanese model 5.0", which is priced close to the people, supports 128K windows and provides a smooth and efficient experience. Based on the powerful capabilities of the big device and the ever-increasing big model 5.0, this version has significantly improved the quality of language dialogue, corpus quality and information density.

[abstract of AiBase:]

The price is close to the people, and HK$ 30 provides the quality service of 1M tokens, and the input and output are at the same price.

Support 128K window to ensure smooth and efficient experience for users.

Based on the ability of the big device and the ever-increasing big model 5.0, the quality of language dialogue, corpus quality and information density have been significantly improved.

9. Playing with GPT4 and other big models to kill humans was eliminated by 3 votes because of the lack of depth in speech.

This paper introduces a simulated werewolf killing game, in which AI NPC and human participants test each other, showing the potential and challenges of AI in text generation and dialogue. Humans were eliminated due to lack of depth in speech, and AI NPC showed profound historical knowledge and insight. The future development direction includes reducing hallucination, understanding game world model, role knowledge management and understanding game mechanism. Game developers and players are full of expectations for AI NPC, which will bring novel gameplay and improve NPC level.

image.png

[abstract of AiBase:]

AI NPC plays werewolf killing with human participants, which shows the potential and challenges of AI in text generation and dialogue.

Humans were eliminated due to lack of depth in speech, and AI NPC showed profound historical knowledge and insight.

The future development direction includes reducing hallucination, understanding game world model, role knowledge management and understanding game mechanism.

10. OpenAI researcher joined Anthropic after resigning due to security problems.

Jan Leike left OpenAI for security reasons and joined rival Anthropic. Leike’s participation is consistent with Anthropic’s values, focusing on ensuring that intelligent systems are consistent with human values.

[abstract of AiBase:]

Jan Leike left OpenAI and joined Anthropic for research.

Anthropic is a public welfare company, which is committed to cautiously promoting the development of artificial intelligence technology.

Anthropic focuses on ensuring that intelligent systems are consistent with human values.

11. OpenAI is training the successor of GPT-4, and GPT-5 will bring three major upgrades.

OpenAI is developing GPT-5, which is expected to bring higher accuracy, more multimodal support and autonomous action capability (AGI). The company has set up a safety and security committee to provide suggestions for the next generation model. The release of GPT-5 may take some time, but it is expected to bring revolutionary changes.

[abstract of AiBase:]

Higher accuracy: GPT-5 will improve the accuracy of answers through more data training.

Increase multimodal: GPT-5 is expected to support video output and handle more kinds of input and output.

Realizing Autonomous Action Ability (AGI): GPT-5 is expected to have artificial general intelligence, understand users’ intentions and complete tasks independently.

12. Chromebook Plus has added a built-in AI function, and Gemini provides support for photos and writing.

The built-in AI function of Gemini launched by Chromebook Plus brings users a new experience of writing and photo editing. Users can write text through the "Help me write" function, optimize photos with Magic Editor, and get quick AI answers through Gemini Chat. All Chromebook users will benefit from these new features, including easy setup with Android phones, built-in task viewing and screen capture tools.

image.png

[abstract of AiBase:]

Chromebook Plus adds Gemini’s built-in AI function to help users write and edit photos.

New features include "Help me write" function to help users write words, Magic Editor to optimize photos, and Gemini Chat to provide quick AI answers.

All Chromebook users will get new features, including simple setup with Android phones, built-in task viewing and screen capture tools.

13. ChatGPT has become a well-known AI product, but its frequency of use is not high.

This paper introduces a survey report on generative AI jointly published by Oxford University and Reuters Institute of Journalism, focusing on the popularity and usage of ChatGPT in many countries. The survey results show that although ChatGPT is a well-known generative AI product, users do not use it frequently, especially in Japan, France and Britain. At the same time, young people are the main users of generative AI, and the application scenarios of generative AI mainly include text, code, audio, image and video generation.

[abstract of AiBase:]

ChatGPT has become a well-known generative AI product, but it is not used frequently by users.

Young people are the main user groups of generative AI.

The application scenarios of generative AI include text, code, audio, image and video generation.

14. Tencent pushes Oracle Bone Inscriptions AI collaboration platform Yinqi Yuan Wen 2.0.

This paper introduces Oracle Bone Inscriptions AI collaboration platform "Yinqi Yuan Wen 2.0" launched by Tencent. This platform uses AI technology to help Oracle Bone Inscriptions decipher, provides data acquisition, storage and calculation capabilities, and has auxiliary tools such as automatic identification of Oracle Bone Inscriptions and AI copy generation, which is convenient for the processing and use of ancient Chinese characters. The platform aims to realize the interconnection analysis of discrete Oracle bone inscriptions and promote experts and scholars to solve the research problems of Oracle bone inscriptions together.

image.png

[abstract of AiBase:]

Tencent launched Oracle Bone Inscriptions AI collaboration platform "Yinqi Yuan Wen 2.0", which used AI technology to help Oracle Bone Inscriptions decipher and provide data acquisition, storage and calculation capabilities.

The platform has auxiliary tools such as Oracle Bone Inscriptions automatic identification and AI copy generation, which is convenient for the processing and use of ancient Chinese characters.

The platform aims to realize the interconnection analysis of discrete Oracle bone inscriptions and promote experts and scholars to solve the research problems of Oracle bone inscriptions together.

Details link: https://top.aibase.com/tool/jiaguwenaixietongpingtai

15. The AI ​ ​ model actually has a preferred number.

This article reveals that the artificial intelligence model shows similar preferences and behaviors to human beings when choosing numbers, although the model itself does not have cognitive ability. People need to realize that models only imitate human behavior, not really think or understand.

image.png

[abstract of AiBase:]

AI model chooses numbers to show human preferences and behaviors.

Humans have fixed preferences and limitations when choosing numbers at random.

AI model behavior comes from training data, not consciousness or thinking ability.

(reporting)