After attracting record guests and participants at home and abroad, comprehensively displaying the latest development achievements and trends of artificial intelligence, and promoting the signing of 32 major industrial projects with a total investment of nearly 30 billion yuan, the 2023 World Artificial Intelligence Conference with the theme of "Intelligent Connected World Generates the Future" ended yesterday at the Shanghai World Expo Center.
As one of the most influential events in the field of artificial intelligence in the world, the World Artificial Intelligence Conference has been successfully held in Shanghai for five consecutive times. This year's conference continues to play an important role as a "science and technology vane, application showcase, industry accelerator, and governance hall", bringing together top scholars, industry leaders, government officials, and heads of relevant enterprises, investors, entrepreneurs, etc., to build a world-class cooperation and exchange platform for them, discuss development and win-win opportunities.
This World Artificial Intelligence Conference is wonderful and fruitful. In just 3 days, a total of 133 forums were held, with more than 1,400 guests. The guests spoke freely, focusing on the hot big model and generative artificial intelligence wave, and deeply discussed the prospects and paths of new AI to empower thousands of industries; Focusing on the huge demand for computing power of various intelligent applications and products, it emphasizes the construction of infrastructure and intelligent computing ecology with upstream and downstream connectivity, equal emphasis on software and hardware, and cloud-edge-device integration. Focus on hot areas such as scientific intelligence, embodied intelligence, brain-like intelligence, metaverse, and intelligent robots, emphasize the comprehensive consolidation of industrial foundation, and activate future development momentum; Focusing on the potential impact of rapid iteration of new technologies on the economy, society, culture, etc., we call for improving the relevant governance systems of artificial intelligence, especially large models, while adhering to inclusiveness and prudence and encouraging innovation.
The exhibition of artificial intelligence technology and applications held during the conference attracted a record number of visitors. This year, the total exhibition area exceeded 50,000 square meters, the largest ever; As of 3 p.m. on the closing day, the total number of offline visitors exceeded 177,000, the total network traffic exceeded 1.07 billion, and the total network exposure reached 6.41 billion, all reaching record highs.
At the closing ceremony, the relevant person in charge of the Shanghai Municipal Economic Information Commission briefed on the upcoming "Big Model Innovation and Development Policy". According to reports, the policy focuses on four directions: innovation ability, innovation elements, innovative application, and innovation environment, and will focus on building 3 plans and 5 platforms to break the bottleneck that restricts the development of this field. The reporter learned that the so-called three plans, namely the big model innovation support plan, the intelligent computing power acceleration plan, and the demonstration application promotion plan, and the five basic support platforms focused on building include the big model test verification and collaborative innovation center, the large model corpus data alliance, the intelligent computing power scheduling platform, the software and hardware adaptation evaluation center, the large model open source community and the collaboration platform.
Yesterday's closing ceremony invited a group of world-renowned business leaders and innovation pioneers to send messages for the development of the industry. Su Zifeng, academician of the National Academy of Engineering, chairman and CEO of AMD, Red Hoffmann, partner of top venture capital firm Grelock, founding investor of OpenAI, Brudermüller, chairman of the executive board of BASF Group, Wende Fengde, CEO of Syngenta Group, and Imad Mostark, founder and CEO of Stability AI, respectively elaborated their views on new AI and new business, and expressed their enthusiasm for the intelligent future. Representatives of Shanghai Law Society, Shanghai Artificial Intelligence Industry Association, SenseTime, Fudan University, Daguan Data, etc. jointly launched the "Model" Initiative, proposing to better release the new momentum of artificial intelligence represented by large models.
Yesterday, a number of new artificial intelligence projects were signed on site, covering different fields such as computing power, data, large models, hardware, and software, which will help Shanghai continuously strengthen its innovation sourcing capabilities, continuously condense new momentum for industrial development, constantly refresh new benchmarks for intelligent applications, and continuously optimize the new ecology of openness and cooperation. According to statistics, during the World Artificial Intelligence Conference, a total of 32 major industrial projects were signed, with a total investment of 28.8 billion yuan; More than 200 upstream and downstream enterprises have completed the docking, reaching the intended purchase amount of 11 billion yuan.
Vice Mayor Liu Duo attended.
New AI injects "soul" into digital humans, but it also comes with risks
In the branch venue of Zhangjiang Science Hall, the audience is experiencing an intelligent health detection system. Photo by Yuan Jing, reporter of this newspaper
Jin Wenyun, founder of 3D technology company Canbeing, customized many highly realistic digital humans for customers. These intelligent twins with high appearance, high emotional intelligence and moving voices in the virtual network space have begun to become gold medal shopping guides in online stores, gold medal hosts on news platforms, or gold medal teachers in after-school tutoring. But as his career became more and more successful, Jin Wenyun became a little confused about the future of digital humans.
As a science fiction fan, Kim keeps in mind the "three laws of robotics" proposed by American science fiction writer Isaac Asimov to ensure that artificial intelligence (AI) does not harm humans. At the 2023 World Artificial Intelligence Conference "AIGC Digital Human Forum" held yesterday, Jin Wenyun said that we need to ensure that digital humans do not cross the "three laws".
Like other new technologies, digital humans test the principle that we can "solve development problems with development solutions". Yesterday's forum actually gave everyone a rare opportunity to sort out the two hot concepts of AIGC (generative artificial intelligence) and digital human, and especially think about what kind of "chemical reaction" will occur when the two are combined.
It's not just fake
First, we need to clarify the concept of digital humans.
Tang Yujia, a senior researcher at the AIGC Digital Human Research Institute, said that the world's first digital human is the heroine of the cartoon "Chrono Fortress". According to the arrangement of the filmmaker, she "walked out" of the cartoon and published a music album in the real world, which allowed people to see the vigorous vitality of virtual IP.
Limited by technology, the first generation of digital humans mainly relied on hand-drawing. By the beginning of this century, with the advancement of motion capture technology, 3D digital characters appeared in film and television works, and digital humans entered the second stage. In the last 10 years, because of the breakthrough of deep learning algorithms, the third generation of AI-generated virtual characters has come out, and they look closer to real people (and may have some CG flavor), and can do things according to scripts.
The Daily Economic News is a user of such technology. They introduced two digital hosts in their live news shows. What is somewhat legendary is that because it is so realistic and the "identity" has never been officially announced, it is more than two months after the broadcast that the audience realized that the two anchors turned out to be virtual. Later, tireless digital people gradually became the darling of e-commerce. Some media reported on the new generation of digital "model workers" - "The company is empty in the early morning, and all the people who bring goods are digital people".
If the evolution of digital humans ends here, it will not make Jin Wenyun have any entanglements. But the popularity of ChatGPT made the situation start to change. "AIGC gives the digital human soul." Tang Yujia said.
Deeply driven by AI
Limited by technology, previous generations of digital humans lacked a complete "personality", if not "worthless". According to Tang Yujia's research, there are two types of driving for digital people: one is real person drive, that is, let the camera capture the actions of real people, and then apply them to the image of digital people with the help of software, which is roughly equivalent to covering real people with a "digital mask"; The second is AI-driven, that is, the digital human is not exactly the shadow of the real person, he can move and speak independently from the real person.
Before the birth of large models such as ChatGPT, even if it was AI-driven, the performance of digital humans was still not in place.
CEO Jiang Yahong revealed that they have developed a set of algorithms and a special camera, as long as they use it to take pictures of real people, they can quickly model and generate digital avatars close to real people. At present, they have implemented their business, such as embedding tourists' digital twins into the beauty of West Lake and generating "metaverse tour Hangzhou" videos; Or let digital people do things that real people simply can't do - backflips, big slam dunks... Upload on social media to get liked.
Although such a digital human is AI-driven, it can only follow a library of pre-recorded actions, which seems a bit blunt and not smart enough to talk to people. However, the shortcomings may soon be remedied. Jin Wenyun said that AIGC will definitely create an action generation model, so that digital humans can get rid of the pre-recorded action library and generate rich body language by themselves. What's more, large models also allow digital humans to completely get rid of scripts when speaking.
The phenomenal mobile game "Against the Water and Cold", which was officially tested at the end of last month, introduced a GPT-like model for the first time. In the game, NPCs (non-player game characters) can talk to players and make actions on their own, which amazes many people. In fact, this technology can be fully transplanted to digital humans, so that digital people not only look realistic, but also behave closer to reality.
Stay calm in the heat
Human-like appearance, human-like body language, but digital humans with the blessing of AIGC have more ambitious goals.
Chuangyi Technology is one of the most influential digital human companies in China, and has created the top Internet celebrity digital human "Liu Yexi". CEO Liang Zikang said yesterday that digital people are the key to the meta-universe era, and to break through the boundary between virtual and real, it is necessary to let digital people have emotions close to real people, "digital people without emotions, can only be e-commerce anchors, no one will watch at all." ”
At present, Chuangyi Technology has done a lot of work in emotional simulation. Liang Zikang revealed that in the next 3 years, their goal is to create a metaverse digital human IP matrix.
More immediate business models are already emerging. Li Xingye, vice president of SenseTime, said that they recently released an app that allows ordinary people to take a five-minute video of themselves with their mobile phones to generate digital humans who can speak more than 150 dialects and foreign languages. In addition to satisfying ordinary users, SenseTime is also running a "Variety of Influencers" program that will create digital twins for 100 opinion leaders.
With the empowerment of AIGC, digital humans are expected to land in the following scenarios - becoming entertainment idols, e-commerce anchors, or playing professional roles in professional fields such as news, healthcare, education, and law. Overall, this market is expected to reach tens of billions of yuan in 10 years. Even, Tang Yujia, who thinks further, believes that with the prosperity of large models and AIGC tools, everyone will have their own digital avatar in the future, and "everyone can achieve digital immortality."
Jin Wenyun's entanglement began with "digital immortality". A client asked him to develop a product with the goal of replicating a digital twin for a deceased loved one. After careful consideration, Jin Wenyun declined, "There may be ethical issues in this." Facing the audience, he raised two questions: Should digital humans be seen as emotional and conscious subjects in the real world? What is the status of digital humans in society and legal systems?
Indeed, the technological and business prospects of digital humans are exciting, and their security and ethical risks are also worth paying attention to. Fortunately, both the government and manufacturers have paid attention to this problem. Li Xingye revealed that in accordance with regulatory requirements, SenseTime's digital human technology is embedded with a trusted security module, and the generated digital human pictures and videos cannot be imported into other content, so as not to be used by criminals.
The development of AI may be accompanied by controversy for a long time, and digital humans are no exception. Maybe this will be a norm in the future.
Author丨Zhang Yi Editor丨Lu Lina