Deep synthesis application trends report released, experts: Everything could be faked in the future

2022-07-02 0 By

Olympic AI sign language anchors, virtual idols, face-changing stars…How advanced are the deep synthesis technologies behind these applications?What is the relationship between the fire “metasverse” and deep synthesis?What are the technical and ethical challenges of deep synthesis regulation?On February 18, the second Great Wall Engineering Technology Conference “Artificial Intelligence security control and ethical governance” sub-forum, the “Deep Synthesis application of ten trends report (2022) was released.Around the ethical issues brought by deep synthesis, several experts discussed at the meeting.1. With the rapid growth of deep synthesis content and the continuous upgrading of relevant researches, deep synthesis technology refers to the technology of making text, image, audio, video, virtual scene and other information by using generative synthesis algorithms represented by deep learning and virtual reality.In 2017, a user named “Deepfakes” brought deep synthesis to the masses by sharing pornographic videos of tampering with faces on the Us website Reddit.The report shows that in recent years, the production and distribution of deeply synthesized content has grown rapidly.The number of new deeply composited videos released in 2021 has increased more than tenfold from 2017.In addition, the attention of deeply synthesized content is also increasing exponentially. Taking the data of likes and likes of videos as an example, the number of likes of newly released deeply synthesized videos in 2021 has exceeded 600 million.Data description: In 10 domestic and foreign platforms (iQiyi, Tencent Video, Youku, Bilibili, Douyin, Kuaishou, Weibo, YouTube, Twitter, TikTok), 10 Chinese and English keywords such as “Deepfakes” were searched, and the data results were obtained after the URL was removed.The research results are the underlying driving force for the spread of deep synthetic content.Among them, the University of Montreal proposed generative adversarial Network (GAN) in 2014, which pushed the fidelity of data to a new level and greatly lowered the threshold of deep synthesis.The number of papers in the field of deep synthesis continues to grow each year, according to the report.These papers in the field of in-depth synthesis include the technical research on the synthesis of different modes such as image, speech and text, among which the research on image generation accounts for the highest proportion (64%).Audio and text accounted for 24% and 12%, respectively.In addition to paper research, open source projects in the field of deep synthesis are also on the rise.Open source projects promote the continuous upgrading and iteration of deep synthesis methods in terms of synthesis quality and production efficiency.In terms of application scenarios, the report shows that in-depth synthesis of application scenarios, such as the restoration of historical photos, AI sign language anchor, virtual idols, and other applications.Especially in the field of film and television production, in recent years, deep synthesis technology has become a tool to save the works from the bad behavior of a few artists.In addition, more and more organizations are using deep synthesis technology to provide products and services to the public.The situation varies by field. According to the report, images and videos were most commonly used in the early stages of deep synthesis applications, but their number gradually decreased after regulations were put in place, due to the uneven quality of the products and the infringement of users’ privacy.In audio, speech synthesis has become an important part of human-computer interaction, and is widely used in intelligent hardware, intelligent customer service, voice navigation, audio books, robots, voice assistants, automatic news reporting and other scenarios.In terms of text, deep synthesis has been applied more and more in news report, poetry creation, chat and question and answer, and has shown great creative efficiency and future potential.With automated data generation, full-body synthesis, 3D modeling and other technologies taking shape, the report predicts that a new world of human existence will unfold based on deep synthesis.The meta-universe is a future human virtual digital space based on deep synthesis technology. It “completes multiple copies and extensions of real space and time, breaks away from the limitations of traditional physical space, and provides a new world where virtual people, natural persons and robots merge and grow close to and beyond reality”.According to the report, when the technology of deep synthesis penetrates into all areas of social life, the negative risks of deep synthesis content continue to intensify and cause substantial harm.With the opening and open source of deep synthesis technology and the increase of deep synthesis products and services, the technical threshold of deep synthesis content production is getting lower and lower, realizing the “popularization” of technology.Through the deep synthesis technology to create false video, false audio frame, slander, fraud, extortion and other illegal acts have been common.Deep synthesis technology will also have a more profound impact on the dissemination of information.The report analyzes that human communication activities are gradually stepping into a “deep post-truth” era due to deep synthesis technology.First of all, “deep forgery” has a profound impact on the record of the truth of news, and the difficult screening of false content affects the effectiveness of fact checking.Secondly, at the nodes of major social emergencies or political events, if deep synthesis technology is maliciously used, false information will spread virally on the Internet with the help of social media.Thirdly, in the information release and tracking of daily events, deeply forged information will also cause the constant reversal of public opinion in the public opinion field and intensify the contradictions among different social groups.What needs to be vigilant is that malicious forged content of deep synthesis technology usually caters to the public’s curiosity and has a strong ability to shape consciousness.The report also points out that the identification of deeply synthesized content is facing technical challenges.With the emergence of new forgery methods and the structural defects of detection algorithms based on deep neural network, anti-deep forgery detection technology also faces “strong antagonism”, which requires continuous updating and iterative optimization.This is similar to the “cat and mouse game”, where deep composition and detection evolve themselves as they learn how to attack and defend, circumventing the antagonistic techniques of the previous generation.At present, both academia and industry have invested a lot in the research and development of identification detection technology, and many domestic and foreign scientific research institutions and technology enterprises have launched detection products.As the negative effects of deep synthesis appear, the establishment of regulatory mechanisms in the world has become a trend.The European Union tends to bring deep synthesis into the existing legal framework for regulation.In the US, some states have passed formal laws regulating “deep forgery”, such as California, Virginia and Texas;Singapore has also enacted special laws to clarify the responsibilities of the parties and platforms.Released in January 2021 in our country the administration of network audio and video service provisions specifically mentioned the great advantage of deep learning technology production and spread false news and information, in January of this year, national network office issued a letter “Internet information service depth synthesis management regulations (draft)”, is a systematic, targeted and operability of the special provisions for the administration.3 expert: Everything of value may be forged in the future. Facing the current challenge, how to standardize the application of deep synthesis technology and reduce the negative impact of technology?A number of experts expressed their views from the perspective of ethics and governance.Xue Hui, head of The Security perception and Cognitive Intelligence department of Alibaba, believes that the difficulty lies in two aspects. One is that deep synthesis technology has great commercial value, so we should not ban it all, but adopt an “inclusive and prudent” attitude. But how to determine the boundary of supervision is a problem.Another problem is that deep synthesis is faced with continuous attack and defense and game, in which the attacker often finds a point to break through, but the defense is relatively backward.Tao Jianhua, a researcher at the Institute of Automation, Chinese Academy of Sciences, pointed out that at present, the connotation and extension of the concept of deep synthesis are not clear, resulting in regulatory difficulties.”Is it deep synthesis when you use deep learning?I think that’s debatable.”Moreover, he argues, the users of deep synthesis should be managed more effectively, rather than constraining its developers too much.A lot of technology in ARTIFICIAL intelligence is two-sided, he compares technology to knife, the impact of technology depends on how it is used.Many of the earliest researchers engaged in deep synthesis were motivated by entertainment and improvement of people’s lives. For example, some people wanted to make the machine learn the mother’s voice autonomously and read aloud to the baby, which was a way to improve people’s lives.But do not rule out some malicious attack tools.Therefore, the regulation of technology should be open.Ren Kui, dean of the School of Cyberspace Security at Zhejiang University, raised the problem of insufficient data sets.He explained that the current deep synthesis technology is mainly aimed at people, so training the deep synthesis detection model requires a large amount of face data, but face data and audio data are highly sensitive personal information, which is difficult to obtain.He recommends that credible nonprofits comb through the data and engage qualified research institutions to maximize the value of the data “and use it in a positive way.”However, future deep composited scenarios are likely to be more complex.In his view, the future of deep synthesis will not stop at simple audio, images and video, but will be used for all kinds of forgery, and not just forgery in the digital space. Forgery in the physical space can be even more deceptive and deadly.”From the point of view of key scenarios, like autonomous driving, where I might fake a scenario, it might be digital, it might be a way of merging with the physical world.If we think a little bit further, like the idea of the metasomes, it’s not necessarily about falsifying information about people, anything of value can be falsified, and there’s a lot of room for imagination and exploitation and attack in deep synthesis.”Tian Tian, CEO of Beijing Ruilai Intelligent Technology Co LTD, believes that the essential problem with deep forgery is a lack of transparency.In this technology, the traditional “seeing is believing” is challenged, so it is important to raise people’s awareness of deep synthesis technology.”For the general audience, there needs to be a lower threshold for the question, a recognition of what is deeply synthetic, or a simple tool to determine that it is synthetic.Only when the threshold is lowered to the extent that all audiences can recognize, discuss and understand this issue under a common framework, can it be a relatively healthy and benign development, and its application can be expanded in a larger range.”He said.Reporting: Nandu reporter Li Yaning