As the breakthroughs in core technologies such as data, algorithms, and computing power, AIGC is promoting the paradigm transformation of content creation under the trend of symbiosis between virtual and real.
AIGC (Artificial intelligence-generated content) is a production method that automatically or assists in generating content through AI technology. With the continuous development and breakthrough of technology, artificial intelligence has an increasingly important impact and role in the fields of content creation such as painting, music, games, news, and art.
With the advent of the era of 5G large bandwidth networks, people are increasingly eager for more visually expressive digital content. The efficiency of generation of traditional digital content has become a bottleneck in the new era. As the next hot spot for exploration, "AI automated content generation" has stimulated a large number of industry demands and has also allowed us to see a new trigger point for artificial intelligence technology.
AI lowers the threshold for content production
As the saying goes, "Technology is productivity, which can improve people's quality of life and improve people's way of working." In the field of content production, we have seen the power of AI "subversion".
From the perspective of creators, the development of the content ecosystem can be roughly divided into four stages: Professional Generated Content (PGC), User Generated Content (UGC), AI-assisted Production Content, AI Generated Content (AIGC) .
PGC (Professional-generated content, professional production of content) mainly refers to the production of high-quality content by a professional team for commercial monetization; UGC (User-generated content, user production of content) confuses the boundaries between consumers and producers, and the creator is the user himself.
These are two stages of the current Internet content creation ecosystem, but its production potential is gradually being consumed.
To ensure quality, PGC often requires a lot of R&D costs, which is the main reason for the long-term losses of domestic long video websites. In contrast, although UGC lowers the production threshold and makes the community more prosperous, it is precisely because of the high creative freedom that it is difficult to guarantee quality.
In fact, the process of content creation is the process of creators processing, processing, structuring information, and selecting and using content carriers. A series of processes are based on the creators' acquired learning, which requires a lot of time and energy. With the explosion of concepts such as VR/AR and Metavers, future Internet applications are evolving into a rich media platform, with increasing demand for high-quality and diverse content.
When PGC and UGC are limited by production capacity and quality, and when the information processing capacity of the human brain reaches its limit, new production methods need to bring about content changes. From the perspective of development trends, AI+ content production will make up for the gap in content consumption and supply in the digital world.
AI completed the "Fuchun Mountain Residence Picture" and wrote a poem (in the red frame above)
Two months ago, Baidu used AIGC capabilities to instantly restore the fragment of "Fuchun Mountain Residence Picture" in just "1 second". The consistency between the style and the existing original works also greatly made experts very happy. Shocking; Tencent created the "Dreamwriter" news writing system, which can write in 22 prescribed scenarios, with an average publishing speed of 0.46 seconds; Himalaya can efficiently convert a large amount of text information in news, books and articles into audio through voice synthesis (TTS:Text-to-speech) technology, and these TTS contents have also gained a large amount of listening volume after they were launched.
AI technology not only helps improve production efficiency, but also helps further improve interactivity. For example, in the game "AI Dungeon", when the user enters text, the system will use the GPT-3 (Generated Pre-Training Transformer) natural language model to understand the script and generate the next few paragraphs of text, and can basically achieve the consistency of the front and back worldviews.
However, the development of AI technology corresponds to content production. At present, it is more about AI-assisted production. Creation has not broken out of the creative framework of PGC and UGC. For example, the creation of virtual people requires human encoding genes for them, setting up their personalities and backgrounds, and then interacting with the external environment. With the continuous iteration of factors such as data, algorithms, and computing power, AIGC will be the long-term direction. Behind the breakthrough of
AIGC in text, audio, and meta-universe construction
AIGC is a combination of artificial intelligence technology and a number of key technologies such as multimodal interaction technology, 3D digital human modeling, machine translation, speech recognition, natural language understanding and other capabilities.
From the perspective of technical capabilities, AIGC can be divided into three levels according to the difference in object-oriented and implementation functions.
Three cutting-edge capabilities of artificial intelligence generation content (AIGC)
At present, AIGC has made breakthroughs in text, audio, and metaverse construction:
AI creation tool function around text has achieved a major breakthrough. AI technology in text creation applications include recognition and translation, writing poetry/novel/news, etc. At present, text recognition has achieved high accuracy.
also made great progress in content creation, and production efficiency and interactivity have been further improved. For example, Tencent created a news writing system for "Dreamwriter" to write in 22 scenarios, with an average publishing speed of 0.46 seconds; in the text adventure game "AI Dungeon", when the user enters text, the system will use GPT-3 (Generated Pre-Training) Transformer) Natural language model to understand scripts and generate the next few paragraphs of text, and can basically achieve the consistency of the world view before and after; the continuous writing application of natural language processing model based on large-scale language models "Caiyun Xiaomeng" can already realize the AI creation of novel stories. Just give her a start of 1-1,000 words, and it can continue to write the subsequent story for you.
Audio-based AI creation interactivity has also been further improved. Currently, AI has been applied in the fields of music generation, speech synthesis, song production, etc., and its interactivity and real-timeness have been further enhanced. Tom Gruber has now created LifeScore, an adaptive music platform that can arrange music dynamically in real time. After users enter a series of music "raw materials" into LifeScore, AI masters will change, improve and mix in real time to bring music performances.
Compared with text and audio, AI image/video/3D model creation is relatively more difficult. Lip2Wav AI speech synthesis technology realizes the lip shape transformation of dynamic video. In 2020, the team from the University of Hyderabad in India and the University of Bath in the UK launched Lip2Wav's AI voice synthesis program. Creators only need to provide target voice content and character videos. The program can directly lip-convert the dynamic video, output video results that match the target voice content, and achieve extremely high similarity between individuals, rather than a universally applicable general model. The Omniverse Avatar launched by Nvidia is an interactive AI product based on technologies such as voice, machine vision, natural language processing, etc. It integrates video rendering capabilities (OmniVerse), speech recognition and interaction (Riva, Maxine), natural language processing (NeMo Megatron), and AI recommendation (Merlin). It can effectively form three-dimensional portraits and conduct human-computer dialogues, and can be applied to fields such as artificial intelligence assistants.
AIGC Future commercial value
technology will eventually serve business. As the next hotspot for exploration, AIGC has stimulated a large number of industry demands and is creating more and more realistic value.
Artificial Intelligence Generated Content (AIGC) Application View
From the perspective of application value, AIGC will be expected to become a new engine for the innovative development of digital content, injecting new impetus into the development of the digital economy.
From the current development stage of AI technology, AIGC is closely related to game narratives, which not only shapes a broader interactive narrative category, but also brings new inspiration to social gameplay and business models.
For example, AI Dugeon developed an AI model to deal with multiple players through AI technology, which can provide feedback on the interactions of different players. In terms of business model, AI Dungeon provides more advanced AI models as value-added services to players, such as smarter monster AI models. In traditional RPG games, the pets obtained by krypton gold are numerically advantageous over free player pets, while in "AI Dungeon", it is reflected in higher intelligence and stronger interactiveness.
In the field of art, AI's learning and creative ability is subverting our cognition and giving the public a greater room for imagination about the integration and innovation of technology and art. In early June this year, the digital collection of AI paintings created by the fledgling "AI painter" Du Xiaoxiao sold for a high price of more than 170,000 yuan. The four paintings she sells can be completed in just a few seconds on average.
It is worth noting that AI content generation technology has been implemented in various explicit business scenarios . At the human level, digital employees have natural advantages in reducing labor costs, improving work efficiency, and reducing personnel flow risks. At the level of goods , the content display of some e-commerce platforms is more three-dimensional, and will present the products customers want to buy from various angles. on the site level , through 3D online space, participants can be more immersive.
The first digital virtual idol to realize AIGC, Xigaga
Although the development and revolution of technology have brought prosperity to the dissemination and creation of intellectual property content to a certain extent, the legal issues such as the ownership of relevant ownership that affect capital confidence and industrial development have not been confirmed.
In February this year, the Copyright Review Board of the U.S. Copyright Review Board once again rejected the request for the reconsideration of the copyright of the "Paradise Entrance" written by Stephen Thaler, a work created by Abbott, and reiterated that according to the provisions of the US Copyright Act, the work requires that the identity of a human author is included. Therefore, this work "a recent entrance to paradise" created by artificial intelligence cannot obtain copyright authorization.
In fact, since artificial intelligence technology began to be applied in the fields of news writing, painting, poetry writing, etc., The copyright issues related to artificial intelligence products have been plaguing the academic and practical circles, and there are many controversies.
At present, the development of AIGC should be considered to have surpassed the general weak artificial intelligence standard, infinitely reaching the strong artificial intelligence stage, but it has not achieved the arrival and surpassing the strong artificial intelligence, or general artificial intelligence standards.
Oxford philosopher and well-known artificial intelligence thinker Nick Bostrom defines super intelligence as "much smarter than the smartest human brain in almost all fields, including scientific innovation, general knowledge and social skills." In the super-artificial intelligence stage, artificial intelligence has crossed the "singularity", and its computing and thinking capabilities are far beyond the human brain. Artificial intelligence at this time is no longer something that humans can understand or imagine.
Regarding the time of completion, Melanie Mitchell, an artificial intelligence expert at the Santa Fe Institute , had tangible discussions and controversies with Elon Musk (Elon Musk). The focus of their controversy is that the time of completion is 2029. What kind of surprises will AIGC bring to us at that time and what serious challenges will it face? Let's wait and see.
Editor: Yue Qingzhi
Producer: Li Hongmei
Reference:
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-