Development Trend of Metauniverse Technology (1): Generative Artificial Intelligence Continues to Write a New Chapter of Metauniverse

Editor’s note:Future industries, including the meta-universe, are an important opportunity to seize a new round of scientific and technological revolution and industrial transformation, and an important starting point for leading development. Since 2021, Metauniverse has become a popular new outlet of Internet industry, and it is constantly evolving driven by innovative technologies such as content production, scene construction, virtual-real interface and authentication mechanism. China Electronic News published a series of reports on the development trend of metauniverse technology, interviewed authoritative experts, scholars and leading enterprises, and analyzed the marginal changes caused by key technologies to the development of metauniverse, so please pay attention.

The highly sought-after metauniverse seems to have entered a period of pain after the explosion of fire. Especially in recent days, the news of NFT "ebb tide", virtual real estate "dumb fire" and the dissolution of Dachang Yuanhu team has made many people doubt the subsequent development of Yuanhu. In sharp contrast, the wave of AIGC (Generative Artificial Intelligence) technology represented by the big model swept the world, and the topic continued to be hot. This has caused many people to question: Is the Metauniverse "out of breath"? Is AIGC and the Metauniverse "complementary" or "complementary"? Next, will there be any new stories? Many industry experts believe that the relationship between the two is not competitive. On the contrary, AIGC is continuing to write a new chapter for the meta-universe.

AIGC opens a new era of "national creation"

To build a meta-universe, a large number of digital contents including various virtual scenes, objects, characters and events need to be filled first. Previously, the lack of content was considered as one of the core challenges faced by the meta-universe. However, the popularity of AIGC has subverted this situation.

AIGC, or AI Generated Content, refers to the use of artificial intelligence technology to generate various forms of digital content, such as images, audio, video, text, 3D models and so on. It is considered as a new way of content creation after PGC and UGC. Like the recent popular "anime face", "AI painting", "digital person" and "chat robot", it is inseparable from the support of AIGC technology.

Baidu CEO Li Yanhong publicly stated that in the next 10 years, AIGC will subvert the existing content production model. It is possible to generate AI original content at a cost of one tenth and a production speed of one hundred times and one thousand times.

Unlike human creators, AIGC uses artificial intelligence models to learn the patterns and structures of existing data, and generates new content according to input conditions. This allows it to generate a large number of high-quality content quickly and stably without being limited by the number and ability of human creators, time, cost, skills, inspiration and other factors, thus greatly improving production efficiency.

At the same time, AIGC can also customize personalized content according to users’ needs and preferences, so as to better meet people’s demand for content differentiation and customization in the meta-universe. For example, in Horizon Worlds, a social meta-universe, users can automatically generate a unique virtual image through AIGC technology without any design or programming skills, just by inputting their desired features such as hair style, clothes and appearance.

In addition, AIGC can also use artificial intelligence model to realize intelligent dialogue and generation between content and users, enhance user participation and content interaction, and further optimize the immersive experience of the meta-universe. For example, GET3D, introduced by NVIDIA, allows users to generate a 3D scene by inputting words or voices, and can interact freely in it.

Professional AIGC tools can not only improve the productivity of creators, but also significantly reduce the technical threshold for turning ideas into reality. At the 2023 Game Developers Conference held not long ago, Epic, Roblox, etc. have launched related tools to create meta-space by using AIGC technology.

Recently, OpenAI, which is in full swing, is also promoting AIGC to generate 3D content. The AI-oriented construction tools of metauniverse make creation an attraction for users to enter the metauniverse.

8K 3D VR Yuan Universe Live Room

Forrester analyst Lu Guannan said: "AIGC will make a more significant contribution to the existing content production process, improve the efficiency of content production, lower the threshold of content creation, and even make the creative process civilian. The content creation method supported by AIGC will become one of the important tools to solve the problem of content shortage in the meta-universe. "

However, he also pointed out that it is still difficult to say that AIGC is the optimal solution to solve the problem of metauniverse content, because many enterprises are still exploring the content requirements of customers for metauniverse scenes, and it is still uncertain which content requirements can be met through AIGC related tools or processes.

Open the meta-cosmic technologyimaginary space

From the technical dimension, the outbreak of AIGC in this round mainly benefited from the progress of algorithm technology, including deep learning algorithms such as confrontation network, flow generation model and diffusion model. The functions of various algorithms cover the selection of data weights, the construction of data samples from noise, the translation of texts in different languages, and the matching of image and text feature values. These technologies are in the same strain as the underlying core technologies of the meta-universe.

"To realize such a huge and complex virtual world as the Metauniverse requires the support of various cutting-edge technologies, and it is a long-term and gradual process. In order to get a close-to-real experience in the meta-universe, it is not enough to have only human players, but also an infinite number of virtual scenes, intelligent NPCs, story engines, etc., which can only be achieved by relying on AIGC. " Guo Wilderness, deputy general manager and technical director of Ali Yuanjing, said. He believes that in fact, from the beginning of ChatGPT and large model technology, it has reached a qualitative change.

When collecting the latest progress of the Meta-universe, the reporter learned that ImageBind, the latest open-source multi-sensory AI model of Meta on May 16th, has been able to take vision as the core, combining text, sound, depth, heat (infrared radiation) and motion (inertial sensor), and finally achieve arbitrary understanding and conversion between the six modes.

Meta said that ImageBind can be applied to various modes and tasks with little training. If combined with other AI, it can also achieve cross-modal generation. For example, if you hear a dog barking, draw a dog, and give the corresponding depth map and text description. It can even perform operations between different modes, such as "bird image+sound of waves", and get "bird image at the seaside".

ImageBind shows the possibility of artificial intelligence generating multi-sensory perception in the future, which opens the door for designing and experiencing the immersive virtual world, which means that people are one step closer to the imagined meta-cosmic world.

"The breakthrough of AIGC technology has brought more help to the development of the metauniverse," said Fan Xiao, founder of Shanghai Weixiang Space-Time. "The metauniverse based on spatial computing and spatial AIGC can break the uniqueness and monotony of physical space and give it greater imagination and incremental value. We can polish the model closer to the real system and make the digital experience that once existed only on the Internet appear offline. "

Talking about the technical route, Lu Guannan analyzed that from the perspective of recent development, the quality of text-generated pictures has been improved based on the model of Diffusion, which has aroused the industry’s attention to AIGC; Then, because OpenAI publishes ChatGPT, it is proved that GPT model and RLHF can significantly improve the semantic understanding and text production ability in the process of conversation. Both Diffusion model and GPT model belong to large model or basic model, which is also one of the deep learning algorithms.

Compared with the previous generation model, the size of the current mainstream AIGC model has also increased significantly.

"For the metauniverse, new AI technologies can improve existing scenes and even create new ones. Like other technical fields, AIGC or generative artificial intelligence technology will have a significant impact on existing technical fields. " Lu Guannan said.

China Central Broadcasting and Television Station VR 3D Image Drawing Technology Exhibition Area

AIGC empowermentThe story of the metauniverse continues.

In fact, the controversy about the metauniverse has always existed. Recently CEO Ed Zitron, CEO of technology and commercial public relations company, publicly stated that Metaverse has died out and been abandoned by the business community, and investors have already turned to the fiery artificial intelligence. Tim sweeney, CEO of Epic Games, a game production company, countered Ed Zitron’s statement with the number of "600 million monthly lives". Previously, Roblox, the first share of the Metauniverse, disclosed in the latest financial report for the first quarter of 2023 that its daily active users once again hit a new high, increasing by 22% year-on-year to 66 million.

According to the latest White Paper on Ecological Development of Metacosmic Industrial Chain (2023) released by CCID Research Institute, by 2025, China will form a metacosmic industrial scale of nearly one trillion yuan. In terms of new technology infrastructure, generative artificial intelligence (AIGC) represented by ChatGPT has become an important trend of AI development at present and in the future.

From the perspective of the global market, the report of market research company Research and Markets shows that the Meta-Universe market is expected to achieve a compound annual growth rate of 40% during the forecast period (2022-2030), and the market revenue will reach 700 billion US dollars by 2030. Analysis Group, an international consulting firm, said that from 2022, the Metauniverse will make great contributions to the world economy. The contribution to global GDP in the next decade to 2031 can reach 3 trillion US dollars.

Lebaredian, vice president of Omniverse in NVIDIA, is optimistic about the future of the meta-universe. He said that it is expected that in ten years or earlier, a mature model that can create high-quality 3D content only through human-computer communication will be born in the market. The progress of artificial intelligence technology will provide more and more rich details and corresponding operation options for building a customizable virtual world. By then, the AIGC with a market valuation of trillions will better support the Meta-Universe circuit, drive the collaborative innovation of upstream and downstream industrial chains, and continuously inject vitality into the development of the Meta-Universe.

However, at this stage, challenges still exist. Lu Guannan pointed out that, firstly, for the existing AIGC service, how to reduce the cost and improve the response speed is the key to further expand its influence; Second, the usage scenario of AIGC in the industry is still in the exploration stage, and enterprises still need to invest some resources to explore how to integrate existing products with AIGC; Thirdly, the development of generative artificial intelligence technology has begun to optimize around more subdivided tasks such as picture generation, text generation and code generation, and in the process, it has explored commercialization modes, but it is still difficult to predict the time point of large-scale commercialization.

He suggested that from an industry perspective, regulators need to further clarify and refine the power and responsibility of generative content services. The platform side or service provider needs to provide tools to identify or improve the generation results of AIGC. The provider also needs to guide the platform users according to its own usage specifications. In terms of corporate and personal use, it is also necessary to actively pay attention to compliance and avoid potential copyright risks.

According to the technology maturity curve, the emergence of each new technology will go through five stages: birth period, bubble period, trough period, recovery period and maturity period. The birth of metauniverse is not a flash in the pan, but the innovation and change brought by the development of a new generation of information technology, which is the beauty of people’s digital life.Good imagination and expectation. As the main force of future content generation, AIGC will continue to write a new chapter for the "breaking first and then standing" of the meta-universe.

Author Song Jing

Editor Liu Jing

Mei bian ma Li ya

Producer: Zhao Chen
Reporting/feedback