current position:Home>The next outlet after Web3.0, AIGC will become the future content trend?

The next outlet after Web3.0, AIGC will become the future content trend?

2022-11-23 13:53:44Potential of the AI



图片来源:由无界版图 AI 工具生成

AI,人工智能,From the beginning of this year to now, it has become the most popular on the Internet、One of the most discussed technological concepts among netizens.

今年4月,一款名为DALL-E 2的AIA painting model is born,In just a few months, it has taken the world by storm,Be the hottest this yearAI绘画模型,Known as the strongestAI画手.

Whether it is Renaissance style、polaroid style,or edward·Hope style、Monet style、数字艺术,or retro style、World War II Historical Photography Style and More,DALL-E 2都能轻松实现,just give keywords,大约60can be generated in seconds10images for user to choose.

DALL-E 2部分作品(图源:DALL-E 2)

Brain hole is infinite、写实、易操作、More accurate command understanding...The collection of various elements has produced many eye-catching creative works,DALL-E 2Works are loved by many users on social media.

在DALL-E 2Before the Internet became popular,Enter text to generate images,This trend of things that exist in the imagination becoming reality has begun to prevail.在今年年初,AI绘画工具Disco Diffusion、Midjourneyartist、favored by fashion pioneers,Not just for entertainment,Can also be seen in commercial useAI的身影.

Before winning first prize in the Colorado State Fair Art Creation Competition event,今年六月《经济学人》The journal cover is fromMidjourney之手.(图源:《经济学人》)

AI绘画平台Midjourney在Discordhas over a million subscribers.


从Disco Diffusion到MidJourney,再到DALL-E 2Internet-wide popularity,Text-to-Image(文字生成图像)From niche development community fire to social media into the public eye,AIThe creative way of generating images has amazed users at home and abroadAIdevelopment to such an extent!

输入“mountain of bones”、“宗教化”、“火焰”、“Indifferent people”Four cards obtained by four key phrasesAIGC图像(图源:VR陀螺)

最近,AIAfter the enthusiasm for painting subsides,AIWriting a thesis helps againAI破圈.It comes from an articleReddit上的帖子,发帖人Urdadgirl69Indicating that his thesis was written byAIghostwriting completed,And his teacher didn't notice anything unusual about the homework.


More impactful is,MetaText-to-Video released in early OctoberMake-A-Video功能,意味着AIAnother breakthrough in image production,In the case of realistic and rich image content,Towards the Dynamic Video Domain.

输入“A dog flying in the sky wearing a superman red cape”获得的视频(图源:Meta)

从“人均AI画师”到“我让AIhelp me write a thesis,不仅拿A,还能赚钱”,AIThis keyword is frequently out of the circle.无论是AI生成图像,还是AI生成文本,或是AI生成视频,These are collectively referred to asAIGCcontent generation model.

AIGC,人工智能生成内容,As the name suggests, content created by artificial intelligence-powered tools,This content production mode is not only in painting、The field of writing becomes a hot trend,Modeling the game scene、数字人、AI聊天、科研AI for Science、AI换脸、Music and other fields have also made achievements,可以看到,This model is becoming the new mainstream.


Benefit from frequent exposure to the public“刷脸”,AIGCThe popularity and prospects of the industry have attracted more and more capital to focus on it.

今年9月,An article officially published by Sequoia Capital《Generative AI: A Creative New World》》中认为AIGCwill represent a new paradigm shift(Cognitive transfer)的开始.


近日,Rely on text generationInstagram标题、Tiktok视频脚本、advertising marketing text、e-mail etc.,less than two years oldAIGC 初创公司 Jasper declares itself as15亿美元的估值获得 1.25亿美元 A轮融资,摇身一变成为AIGCEmerging Powerhouses of the Circle.


10月17日,英国开源人工智能公司 Stability AI 宣布获得 1.01 亿美元融资,估值高达 10 亿美元,跻身独角兽企业行列.

10月21日,根据澎湃新闻,《华尔街日报》Says Google is negotiating,Intended for artificial intelligence start-upsCohere投资至少2亿美元.Coherewhose business includes developing natural language processing software,Includes programs such as chatbots that understand human speech and text.

此外,Microsoft is facingOpenAIA new round of investment in late-stage negotiations.OpenAI推出了DALL·E 2项目,Allow users to generate images from text,即AI作画.

不仅在海外AIGCEnterprises have multiple financing events,甚至出现独角兽企业,在国内AIGCCompanies such as capital favor.

今年1月,Hyperparameter Technology Announced Completion1亿美元B轮融资.

10月,虚拟内容 AIGC Technology service provider Huiye Technology completed tens of millions of RMBPre-A+ 轮融资,Led by Hillhouse Capital and Shunwei Capital.

同样发生在10月,成立于2021production of yearAI平台TIAMAT宣布完成数百万美元天使轮融资.据了解,TIAMAT近4Months of the entire network exposure has been reached5000万,并且这个数据还在不断攀升.


AIGCIn the investment circle, it has become the focus of capital attention,在科技界,Major technology companies also regard it as a key strategic layout link.

Important strategic layout of large factories AIGCis the future content trend 

With the continuous upgrading of the Internet,从Web 1.0到现在的Web 2.0,再到未来Web 3.0时代,From the media era dominated by portal websites to the self-media era、The Rise of Social Platforms,Users' dependence on the Internet is gradually increasing,Content production methods are also changing.

在Web 3.0时代中,The digital world will become clearer,AIIt has always been regarded as one of the key technologies of the Metaverse、An important tool to help digital twins.从“只读模式”的PGC(专业生产内容)to the spontaneous participation of usersUGC(用户生产内容)再到AIGC(AI生产内容),Barriers to Content Creation Are Gradually Lowering,Content productivity will undoubtedly get a higher release.


不妨想象一下,在虚拟世界里,Whether it is to create an exterior that compares to the real scene,Or shape the digital creature“思想”内在,Need to be more efficient、More real-time content creation,This also highlights the ability to break through human limitationsAI技术的重要性.

in-game intelligenceNPC为例,若每个NPCExist in the form of an agent,Then it can not only recognize changes in the state of the environment,And can adjust the behavior strategy according to the state of the environment,act in accordance with one's own personality traits.在任何场景下,Every interaction between the player and the agent,Agents can be based on the player state、The parameters of the environment state and its own state make corresponding behaviors,And then generate sudden dynamic content in the process of interacting with players,This will be in the digital world of meaning is self-evident.

电影《失控玩家》self-conscious intelligenceNPC男主(图源:网络)

Step to the next generation Internet is major technology giants、互联网大厂,Now under the blessing of the metaverse,A new round of acceleration in the digital age,AIGCproductivity behind--Artificial intelligence techniques have been incorporated into the important strategic layout of giants.

去年12月,Meta已将其AI团队并入Reality Labs部门,to develop the Metaverse.

今年1月24日,Metasaid in a blog post that its company's research team is building a new artificial intelligence supercomputer,and is expected to impact in the second half of this year“全球最快AI超算”宝座.

实际上,早在2017年,Facebook released the first generationAI超级计算机.据Meta而言,Compared with the first generation supercomputer,New AI Superclusters(RSC)Improved speed in running computer vision workflows20倍,运行英伟达集体通信库(nccl)的速度提高了9倍多,Natural language processing capabilities more than tripled,and can learn from trillions of examples,在数百种语言中运作,分析文本、Is there any problem with the content of images and videos.

Meta AI团队的研究人员Kevin Lee和Shubho Sengupta在博文中表示,RSCWill play an important role in the next computing platform Metaverse.


今年2月,在Meta AI: Inside the Lab活动上,Meta展示了其AIGC向的“exploratory tool”原型Builder Bot.在Meta官方发布的视频中,随着扎克伯格“order”,公园、海、沙滩、岛屿、clouds etc.,甚至BGMgradually form a complete scene.


这一切都是由AIDriver Generated Content,Build by voiceVR场景,Metasays the tool will“Fostering Creativity in the Metaverse”.

可以预见的是,If this technique is successful,There is a high probability that otherVR世界和平台产生影响,打造VRScenarios will become easier.如果将Meta的Builder bot技术整合,Yuan content ecological will also get a new growth of the universe,例如,With the game metaverse platform that has begun to test the voice functionRobloxCollaborate to create scenes,To enrich the ecological content, etc..

去年4月,英伟达打造的“工业元宇宙”创作平台OmniverseOut of the circle with the virtual Huang Renxun who is difficult to distinguish between true and false,11月份,英伟达推出AIAvatar creation platformOmniverse Avatar说,今年CES 2022上,英伟达宣布Omniverse三项AI新功能,助力AIGC.

从Omniverse到Omniverse Avatar,Nvidia is helping users synthesize the digital content they need more quickly,Including digital scene、Avatar化身、Easily make even a single track3DFacial animation, etc..

英伟达中国区OmniverseBusiness Development Manager He Zhan once said,“从UGC过渡到AIGC,无论是设计师、Startup and many of the big platform companies have obvious trend.因为UGC已经发展了很长时间,Now we want a platform for rapid industrial production,so that everyone canAIBonus for better benefits.”


Not only overseas manufacturersAI,尤其是对AIGCThis new content model“给予厚望”,今年7月,Baidu, a major Internet company in China, usedAITechnology restored《富春山居图》残卷,The core technology behind isAIGC.

在大会现场,Li Yanhong, head of Baidu, also said that:“AIGC是PGC、UGC之后,全新的内容生产方式.它不仅会提升内容生产的效率,也会创造出有独特价值和独立视角的内容.”



Transition from 2D to 3D ,理想中的AIGC究竟有多难?

According to the China Academy of Information and Communications《人工智能生成内容(AIGC)白皮书》,AIGCThere are three stages of development,The first is the early concept stage, which is limited by the technological development,到1990Precipitation and accumulation stage after years,再到2010Year-to-date rapid development,特别是从2014年起,生成式对抗网络(GAN)The proposal and iterative update,以及AIAlgorithm upgrades allowAIGCUshering in the era of content with a hundred flowers blooming.

但目前AIGCThe coverage is more in the image、视频、In relatively two-dimensional scene applications such as voice,And back to the vision of the metaverse,3D application scenarios are the key,AIGCThe application of technology will undoubtedly promote the transition from the two-dimensional Internet to the three-dimensional Internet,但理想的AIGC的难度可想而知.


理想很丰满,现实很骨感,This sentence describes the metaverse,In some ways it can be said that it is very close.And to realize metaverse content productionAIGChow hard is it?需要什么?

AIGCThe improvement of technology is backed byGAN和AI算法的支持,According to the research findings of China Academy of Information and Communications Technology,The upgrade of deep neural network is the driving forceAIGCAnother major reason for the rapid development,实验证明,The learning ability of deep neural networks is positively correlated with the model size,With the increase of the number of model parameters,The ability of the corresponding deep neural network will generally be greatly improved.

Even though model architectures vary,但可以肯定的是,The production of these deep learning network logic often requires a large amount of data to support.Light is a breakthrough in natural language models,The domestic and foreign enterprises and universities have invested a lot of manpower、算力,据悉,The number of parameters of the model has grown from tens of millions to hundreds of billions.

However, the data support of the 3D world is relatively limited.,To be realized in the MetaverseAIGC生产方式,First you need to large three-dimensional material library.And the scanning of 3D material,Data collection is obviously more troublesome than two-dimensional,In today's lidarLiDARWhen this kind of depth camera is not yet popular,The material volume of 3D material is obviously not enough to support an idealAIGC模型训练.



Relative to the two-dimensional plane,Three-dimensional space relatively increases many dimensions,In addition to speech semantic recognition、Beyond computing and image recognition,and spatial recognition such as distance、大小、Spatial location, etc.,Therefore, the upgrade of computing power and model architecture is essential.

Take Nvidia's autonomous driving demonstration case as an example,为了保证安全性,The data sources of the collector on the self-driving vehicle need a lot,激光雷达、雷达、摄像头等.

The data obtained by these samples is not a single,but diverse,包括地图、环境位置、车内状况、emergencies, etc.,Therefore, the amount of data computed per second by an autonomous vehicle-level computer is very large..To achieve high-precision 3D scene independent construction,The upgrade of computing power is also an extremely important part.


Idealized yuan scene is generated according to individual be fond of the universe is in line with the user's own 3 d scene、Characters, etc.,目前,3D建模AIGCcan already be seen in many cases,无论是Deepfake换脸技术,or automatically generated3Dmodel software etc..but still limited toXR技术的发展,目前3DThe optimized scenes and models can only be used in2Dsee in plane.

3DModeling figures(图源:机器之心)

并且,User personalized data requires a huge user base,But user data leaks、Security issues such as illegal collection have also been criticized,There are also many cases of security controversies in the virtual space.


Consciousness here does not mean self-awareness,Instead, it can be based on different semantic information、环境、creative consciousness,eg some intelligenceNPC、Application of virtual idols, etc..

According to the China Academy of Information and Communications Technology,According to the development process of technology and the form of practical application,The ability to create digital content can be divided into two categories: imitation-based creation and concept-based creation.

在未来,With the rapid development of deep neural network,Artificial intelligence will be constantly refreshed,Accompanied by artificial intelligence's stronger ability to understand and generate,Digital content will also reach new heights.


The moves of the technology giants undoubtedly indicate that artificial intelligence has been placed in the core layout of development.,在未来,AI will play a key role.在AI的加持下,Contents of the internet world,Both diversity and quantity will once again reach new heights,毫无疑问,AIGCWhat a key role it will play in the iteration of Internet content creation.

But as one of the most controversial technologies,Artificial intelligence is constantly developing、In the process of integration, new ideas are constantly being brought,new controversy,同时AIInvasion of academia、教育界、Events in the art world and others have created a sense of crisis among the general public.

水能载舟,亦能覆舟.如同MidJourney的创始人大卫·Holtz(David Holz)说的一般,“People see AI as a tiger,a dangerous、The tiger will eat people.but like water,danger in water,But people can swim、造船、use water to generate electricity,It is dangerous,the driving force of civilization.这是一个机会,它没有意志,We may be drowning in the water,But that doesn't mean water should be banned.Artificial intelligence is a new water source,As long as can let the human change well used.”

虽然AIGCSignificant to the scalability of information production,But at the same time, this need a computer、数学、心理学、社会学、The intersection of many disciplines such as philosophy,Only by making progress together can mankind leapfrog the next step of industrial production,This is also the direction that many technology giants are still exploring and striving for..



copyright notice
author[Potential of the AI],Please bring the original link to reprint, thank you.

Random recommended