The emerging concept of metacosmity once exploded out of the circle, accompanied by the development of artificial intelligence and other technologies, triggering the accelerated layout of companies in various fields.
It's been four months since Zuckerberg proposed the metacosm, how are companies laying out the metacosm? What impact will the development of related technologies have on the lives of ordinary people? On March 6, Nandu, N Video and Alibaba's Security Department jointly invited three guests in the technical field to exchange views with netizens on the virtual world and AI risk governance.
The metacosm in the eyes of three technologists
In the interview, Li Bin, chief technology officer of Youzu Network, took the game "Minecraft" as an example and introduced that in the game close to the metaverse, netizens can freely build their own "world". In the game, the different city buildings are restored, "a lot of professional teams have joined the world, where a group of people can build some virtual, but artistic, personalized scenes." Li Bin said: "People can learn about different cultures, histories, geographies through these scenes, and interact together in this scene." He believes that more netizens will join the free creation of this open virtual world in the future, and more imaginative and creative works will be born.
Game Minecraft.
YunZhi, head of Taobao Buy+ VR products, described to netizens a more convenient life and work scenario in the future metaverse. For example, in the future office, people can not be bound by computer monitors, as long as they wear VR or AR glasses, different sizes of screens will be displayed in front of them, and even switch office scenes at will. When performing home decoration, you can use virtual reality equipment to project your favorite design scheme and intuitively experience the design effect. People can also be immersed in various anime scenes, close contact with anime characters, or have a more direct interaction with virtual partners.
An office scene with VR glasses on.
Lang Yining, senior algorithm engineer at Alibaba Security, said: "The metacosm is a higher-dimensional Internet, which has been upgraded from a two-dimensional screen to an immersive Internet space. "From now on, everything can be interconnected, the whole world is your screen; the logic of content production and distribution will also change, more ordinary people will participate in content creation; wearable devices will also make interaction between people more convenient."
"Visual deception" in the metaverse
The experiences that the virtual world brings can increase the richness of people's experiences and make people feel more. But there are also some security risks. If technology can achieve the exact same virtual person as a real person, how can I prove that "I" am me?
Last year, NVIDIA founder and CEO Jen-Hsun Huang used a 14-second "Easter egg" to demonstrate the technology of the 3D simulation platform in a keynote speech, making a perfect virtual replica of Huang Jenxun's online kitchen and his own digital clone, which deceived many netizens.
NVIDIA CEO Jen-Hsun Huang appeared in 14 seconds with a synthetic virtual person.
This security problem, known as "visual spoofing," is one of the risks that has emerged in recent years in the field of AI applications. "What is visual deception?" It is that what you think you see in your eyes is true, but it is not, but it is synthesized. Lang Yining, senior algorithm engineer of Alibaba Security, said.
At the risk of visual deception, everyone can face the problem of whether a video is true or false that the naked eye cannot recognize. At this stage, many Internet companies, including Alibaba, have a solution. According to Lang Yining, in the past, the program for deep fake video detection was mainly divided into two categories: frame-level detection and video-level detection. Frame-level detection methods are expensive to label, and frame-level predictions are converted into video-level predictions, which require high fusion technology and can easily lead to missed or false detections. Video-level inspection studies focus too much on time-series detection models, resulting in certain limitations on detection effects. The study of algorithms that rely only on video-level annotation to obtain good detection results is of great significance to ensuring the security of the future virtual world.
"Governing AI with AI" promotes the sustainable development of artificial intelligence
Some netizens in the live broadcast proposed that today's level of technological development is still a long way from the realization of the meta-universe. Will it be too early to raise the risk governance issue of the metacosm? Lang Yining said that the development of AI governance should not lag behind the development of science and technology.
In the future, people's senses of smell, touch, taste, and pain are at risk of being deceived. This future world, which integrates AR, VR, MR, AI and other technologies, is the product of a high degree of fusion of reality and virtual, and may also bring a series of security problems.
"Artificial intelligence has developed to the current stage, there are already many things that can be done, if we do not do some model constraints, do not increase robustness, then it is likely to have the risk of getting out of control in the future." How to better deal with the security risks, defeat magic with magic, and govern AI with AI is also the direction that Alibaba's Artificial Intelligence Governance and Sustainable Development Lab (AAIG) is studying.
"Achieving sustainable development of AI algorithms is the original intention of governance." Lang Yining said that AI algorithms have become an indispensable part of our lives, in order to avoid the undesirable factors produced in rapid development, to make it sustainable, there must be some governance technologies and concepts.
He believes that it is equally important to insist that technology is available, reliable and trustworthy. "Available" means that the functions implemented based on algorithms will bring a more convenient experience to people's lives; "reliable" means that the technology has stability and does not easily collapse in the face of massive data influx. The current development of open source technology, cloud computing, etc., so that the technology availability, reliability of the two levels can be achieved, and the "credibility" of technology is facing higher requirements, algorithm defense, fairness, etc. are the elements to achieve technical credibility.
"On this basis, we must also achieve a balance between governance and development through multi-party participation and collaborative governance." Lang Yining said that multi-party participation not only refers to the collaborative governance between different Internet companies, but also includes different fields such as academia and industry.
"I think the greatest value of AI governance is to make AI sustainable." Talking about the value of AI governance, Lang Yining said that he hopes to provide people with a more reliable and credible artificial intelligence. "That way, the day we actually walk into the metaverse, it's free and safe."
Written by: Nandu reporter Ao Yinxue