Academic frontiers
Wang Yuan|Privacy Risks and Legal Responses of Embodied Agents: A Case Study of Humanoid Robots
Embodied agents combine artificial intelligence and robotics, and it would be one-sided to discuss the threat to privacy of either technology in isolation. The embodiment, interactivity, and emergence-based characteristics of embodied agents demonstrate their powerful interaction and action capabilities, which bring unprecedented new challenges to privacy and data protection. On the one hand, embodied intelligent bodies can invade private spaces, record private activities, and imperceptibly collect and process private information. On the other hand, the combination of independent decision-making and autonomous action can cause real harm and infringe on the personal and property rights and interests of users. Existing privacy and data protection rules revolve around information control, and the generation and emergence of embodied agent data, combined with the ability to cause actual harm, not only leads to the failure of personal information control mechanisms, but also may make accountability impractical. In order to address these challenges, in addition to weakening the role of individual consent in existing data laws, it is also necessary to embed the concept of data protection in AI legislation, strengthen the responsibilities of regulators and designers, especially prohibit the market application of general-purpose embodied intelligence, and establish the principle of "privacy and data protection by design" to clarify the responsibility of designers. (The full text is published in Oriental Jurisprudence, Issue 3, 2024)
Wang Yuan is a lecturer at Southeast University School of Law
Xu Wei: Identification of the Principle of Infringement Attribution of Generative AI Service Providers
Whether privacy rules or personal information protection rules apply to private information is a matter of great controversy in the academic community, and there are also different approaches in the practical community. There are insurmountable shortcomings in the protection mode of distinguishing between private information and sensitive personal information, the protection mode of competition and cooperation rules, and the protection mode of aggregation rules. With the simultaneous development of the theory and practice of the commercialization of personality elements, there are many difficulties in applying the principle of fault liability to the infringement of existing generative AI service providers, and the basic concepts of the principle of fault liability for infringement of generative AI service providers are not compatible, and it is difficult to effectively solve the problem of proving fault by determining fault based on the "usual technical level of the industry" or "presumption of fault for illegality". The principle of presumption of fault liability adopted by generative AI service providers for infringement has a certain degree of reasonableness, which is in line with the current situation of unequal status, high technical complexity, and information asymmetry between service providers and victims, and in line with the practical needs of victims who should be strengthened for relief. However, the determination that the service provider is at fault is contrary to the objective fact that the service provider cannot foresee and control the generation of harmful content, which will impact the theoretical system of presumption of fault liability. Generative AI infringement is different from personal information infringement, and it is not appropriate to refer to the latter principle of attribution. As a new type of risk liability, generative AI infringement is justified in applying the principle of no-fault liability: service providers can benefit from service provision and have the ability to disperse losses, innocent victims should be given priority protection, and the principle of no-fault liability can help reduce AI risks, reduce costs caused by institutional uncertainty, and provide a predictable institutional environment for service providers. (The full text was published in Legal System and Social Development, Issue 3, 2024)
Xu Wei is a distinguished professor of Sheshan Scholars at Shanghai University of Political Science and Law
Gao Qinwei|Design and realization of the value of the rule of law in digital administration
Digital technologies such as artificial intelligence, machine learning and algorithms have rapidly spread in government operations, improving administrative efficiency, consistency and accuracy, but this has also given rise to questions such as the extent to which technology can protect or even promote the rule of law, and whether the rule of law can be designed as a technical system. and the challenge of directing automated systems to whether human intervention is required. As digital administration continues to evolve, there may be more problems and technical solutions can close some of the gaps, but interdisciplinary research between legal and technical experts is still needed to enhance the alignment between the theoretical expectations of digital government and administrative jurisprudence. Only by integrating the construction and operation of digital government with the value of the rule of law, and improving the norms of public-private cooperation, can the efficiency of digital technologies such as cloud computing, big data, artificial intelligence, and blockchain be fully utilized, and only then can the rule of law fully demonstrate its due meaning of democracy, transparency, participation, and accountability. (The full text was published in Comparative Law Studies, Issue 2, 2024)
Gao Qinwei is a professor at Sun Yat-sen University School of Law
Huo Junge|Research on the Typological Protection Model of Criminal Law for Health and Medical Data
The current criminal law's control-based protection model for health and medical data is not only difficult to eliminate secondary leakage, but also causes loopholes in the protection system, and is not conducive to the development and application of health and medical data. After comparing and analyzing its sharing attributes, it can be seen that health care data can be divided into two types: non-sharing and open sharing. In terms of security protection requirements, the former pays more attention to the protection of data confidentiality under the criminal law, while the latter focuses more on the protection of data integrity and availability under the criminal law. Based on this, the criminal law shall adopt a typed protection model, strengthen the control-type protection of health and medical data that is not to be shared, and promote the utilization-type protection of open and shared health and medical data. The illegal use of non-shared health and medical data involving personal privacy and the undermining of open and shared health and medical data should be included in the scope of the criminal law, and the principle of reasonableness should be regarded as a cause for illegal obstruction of the processing of open and shared health and medical data. (Full text published in Contemporary Jurisprudence, Issue 2, 2024)
Huo Junge is an associate professor at the School of Artificial Intelligence Law, Southwest University of Political Science and Law
Transformation of legal rules for personal data protection in digital government
The construction of a digital government will inevitably lead to a large amount of personal data entering the public database controlled by the government, and the personal data in these public data needs to be shared within the digital government and circulated externally, so as to promote the increase of social public interests. This will inevitably lead to the fact that the previous legal rules on personal data protection, which were mainly based on "notification and consent", cannot be adapted to the application of personal data in digital government. The digital technologies adopted by digital governments have an impact on the expression of personal intent and have led to a degree of transparency in the privacy of individuals. There is a duality between personal data and its rights in the legal nature of personal data and its rights, and the public nature of personal data should be expanded after personal data has been entered into government public databases through legal procedures. In the digital government, there should be personal data protection rules that are different from those of private law, mainly including a system of rules for the use of personal data that emphasizes the sharing and circulation of personal data to increase social value; a system of protection rules emphasizing state regulation to protect personal data; A relief system that emphasizes precautionary measures. (Full text published in Law Science, Issue 4, 2024)
Peng Tao is a professor at the School of Administrative Law, Northwest University of Political Science and Law
The Brookings Institution released a series of articles on the U.S.-China Government Dialogue on Artificial Intelligence
In November 2023, the leaders of China and the United States announced the establishment of a new bilateral consultation channel on AI at their meeting in California, aiming to reduce the risks of AI application and intend to reach a consensus around AI governance. Although the two sides have yet to reach a consensus on some specific issues, there is a policy background and basis for interaction. The multifaceted nature of AI means that the dialogue covers a wide range of topics, including key topics such as military use and data sharing, and China and the United States need to carefully choose the areas to be discussed. Against this background, the Brookings Institution has released articles such as "A roadmap for a US-China AI dialogue" and "Laying the groundwork for US-China AI dialogue", which sort out the background and foundation of the Chinese and US governments' AI dialogue, and focus on the military application of artificial intelligence. The importance of fostering positive cooperation and maintaining focus is essential for the dialogue to achieve tangible results. (The full text is available on the official website of the Brookings Institution; The tweet of the official account of "Frontiers of Cyber Law Theory and Practice" on May 8, 2024 contains a compilation)
The Brookings Institution is a well-known think tank in the United States
New Regulation Express
Notice of the Ministry of Finance on Doing a Good Job in the Pilot Work of Digital Transformation Cities for Small and Medium-sized Enterprises in 2024
On April 30, 2024, the website of the Ministry of Finance announced the Notice on Doing a Good Job in the Pilot Work of Digital Transformation Cities for Small and Medium-sized Enterprises in 2024. The "Notice" makes it clear that about 35 cities will be selected to carry out the second batch of pilot work in 2024, which will focus on the needs of enterprises, accelerate the digital transformation of small and medium-sized enterprises, and give priority to the inclusion of specialized, special and new small and medium-sized enterprises with urgent digital transformation needs and a certain digital foundation into the pilot scope, strengthen digital empowerment, and promote the optimization and upgrading of industrial and supply chains. Focus on promoting the application of artificial intelligence large models, big data, blockchain and other technologies in R&D and design, visual quality inspection, parameter optimization, energy consumption management, intelligent sorting and other scenarios. (Ministry of Finance)
Netherlands Data Protection Authority|Guidelines for Data Scraping for Individuals and Private Organizations
On May 1, 2024, the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) published the Guidelines for Data Scraping for Individuals and Private Organizations. The guidelines set a strict framework for data scraping by private entities and individuals, while also noting that government scraping activities are also fraught with privacy risks. The guidelines emphasize the basic principle that personal data scraping is generally illegal, refute the common misconception that public data can be freely scraped, and make exceptions for cases such as small-scale scraping for personal projects and targeted scraping for specific corporate purposes on the basis of general prohibitions, in line with the GDPR's principles of data minimization and purpose limitation. (Netherlands Data Protection Authority)
State Administration for Market Regulation|Interim Provisions on Anti-Unfair Competition on the Internet
On May 6, 2024, the State Administration for Market Regulation (SAMR) promulgated the Interim Provisions on Online Anti-Unfair Competition, which will come into effect on September 1, 2024. The "Provisions" have a total of 43 articles, which provide for online unfair competition, supervision and inspection, etc., refine the constitutive elements and identification factors of anti-unfair competition acts, comprehensively sort out and enumerate online acts of unfair competition, emphasize the responsibility of platform entities, and further optimize the provisions on law enforcement and case handling procedures in online anti-unfair competition cases. (State Administration for Market Regulation)
Tianjin Pilot Free Trade Zone released the 2024 version of the negative list for the management of data exports
On 9 May 2024, the Cyberspace Administration of Tianjin (CAC) released the Administrative List for Cross-border Data Transfer (Negative List) in the China (Tianjin) Pilot Free Trade Zone (2024 Edition). The Negative List sets out the circumstances under which an enterprise in the Tianjin Pilot Free Trade Zone is required to apply for a security assessment for data export, enter into a standard contract for the export of personal information, and pass the personal information protection certification for providing data overseas. Enterprises in the Tianjin Pilot Free Trade Zone are exempted from applying for security assessment of data export, entering into standard contracts for personal information export, and passing personal information protection certification if they provide data outside the Negative List overseas. Data involving state secrets, core data, and government affairs data are not included in the management of the Negative List, and the export of relevant data is to be carried out in accordance with relevant laws, regulations, and provisions. (Netinfo Tianjin)
The U.S. Congress announced a bill to restrict the export of AI models
On May 8, 2024, Michael McCaul, chairman of the Foreign Affairs Committee of the U.S. House of Representatives, John Murnar, chairman of the U.S. and Chinese Communist Party's Special Committee on Strategic Competition, and Raj Krishnamoorthi and Susan Wild, top Democratic Party members, jointly introduced the "Enhancing National Frameworks for." Overseas Restriction of Critical Exports Act (ENFORCE). The ENFORCE Act makes three changes to the U.S. Export Control Reform Act of 2018, authorizing the U.S. Secretary of Commerce, in consultation with the Secretaries of State, Defense, and Energy, to impose export controls on certain "artificial intelligence systems" and "certain emerging and foundational technologies critical to U.S. national security," and restricts U.S. persons from collaborating with foreign nationals to develop such systems and technologies. (Foreign Affairs Committee of the House of Representatives, U.S. Congress)
Supreme People's Procuratorate|"Opinions on Strengthening Procuratorial Technology in the New Era"
On May 9, 2024, the Supreme People's Procuratorate issued the "Opinions on Strengthening Procuratorial Technology in the New Era". The "Opinions" pointed out that the role of procuratorial technical functions should be brought into full play in procuratorial supervision and case handling. It is necessary to improve the procuratorial technical work system, make overall plans for the use of procuratorial technical resources, strengthen organizational leadership, and increase the level of specialization and standardization of procuratorial technical work. Strengthen inspection and appraisal work, strengthen the evidentiary support role of inspection and appraisal in procuratorial investigations, and give play to the role of inspection and appraisal in the investigation and verification of public interest litigation and the judgment of litigation claims. Strengthen the application of technologies such as rapid detection, big data, and remote sensing in the discovery and screening of clues in procuratorial oversight and case handling. (Supreme People's Procuratorate)
The U.S. House of Representatives has submitted legislation related to cryptocurrencies
On May 10, 2024 local time, the U.S. House of Representatives said it would consider the 21st Century Financial Innovation and Technology Act (FIT21), clearing the way for a subsequent congressional vote. The House Rules Committee said the 21st Century Financial Innovation and Technology Act is an important first step toward regulatory clarity for digital assets, providing strong, time-tested consumer protections and the regulatory certainty necessary for the digital asset ecosystem to thrive in the United States. The Act sets out clear and practical federal requirements for the digital asset market. It provides strong consumer protections and the regulatory clarity necessary for the digital asset ecosystem to thrive in the United States. (U.S. House Financial Services Committee)
Action Plan for Digital Transformation of Manufacturing Industry
On May 11, 2024, the executive meeting of the State Council deliberated and approved the Action Plan for Digital Transformation of the Manufacturing Industry. The meeting pointed out that the digital transformation of the manufacturing industry is an important measure to promote new industrialization and build a modern industrial system. According to the diversified and personalized needs of the manufacturing industry, typical scenarios should be explored in different industries and fields. Accelerate the research of core technologies and the promotion and application of achievements, and do a good job in equipment networking, mutual recognition of protocols, standard formulation, platform construction, etc. It is necessary to increase support for the digital transformation of small and medium-sized enterprises, organically combine it with large-scale equipment renewal actions and the implementation of technological transformation and upgrading projects, improve public service platforms, and explore the formation of a long-term mechanism to promote the digital transformation of small and medium-sized enterprises. (Xinhua News Agency)
Applicable Law
The Croatian Agency for the Protection of Personal Data imposed fines totaling 16,000 euros on seven illegal video surveillance entities
On April 22, 2024, the Croatian Agency for the Protection of Personal Data (AZOP) published a summary of its decision, in which it imposed fines totalling €16,000 on seven unnamed entities for violating the General Data Protection Regulation (GDPR) and the Personal Data Protection Implementation Law. AZOP discovered that all seven entities had video surveillance set up on the premises, but none of them had notified the subjects under the video surveillance, i.e., the notification was not visible when it entered the recording range, and/or the notification did not contain all relevant information. (Croatian Agency for the Protection of Personal Data)
Beijing Intellectual Property Court Releases White Paper on Judicial Protection of Competition in the Data Industry
On April 25, 2024, on the occasion of the 24th World Intellectual Property Day, the 2024 Zhongguancun Forum Global Intellectual Property Protection and Innovation Forum was held at the Zhongguancun International Innovation Center. At the forum, the Beijing Intellectual Property Court released the White Paper on Judicial Protection of Competition in the Data-related Industry (in both Chinese and English). Song Yushui, member of the Party Leadership Group and Vice President of the Beijing Intellectual Property Court, said: "The Beijing Intellectual Property Court attaches great importance to the judicial protection of the digital economy with data as the core, and has concluded a number of competition cases involving the data industry, and released ten typical cases involving data against unfair competition. The White Paper on Judicial Protection of Competition in the Data-related Industry released this time is a summary and refinement of the judicial practice of the Beijing Intellectual Property Court related to data competition, and is also a useful exploration for judicial authorities to protect data rights and interests and maintain the order of fair competition. (Intellectual Property Beijing)
Qingdao police cracked a particularly large case of illegal use of information networks
News on April 26, 2024. In the "Clean Network 2024" special operation, the public security organs in Qingdao, Shandong Province cracked a particularly large case of illegal use of information networks. In October 2022, the Qingdao Internet Police found during their daily inspection that in Huangdao District, Qingdao City, someone operated a QQ account to provide fan attraction and drainage services for overseas fraud gangs, causing many people to be deceived. At the same time, the police also received a number of police calls, and the victims were defrauded of amounts ranging from more than 2,000 yuan to more than 100,000 yuan. Qingdao police arrested more than 110 people involved in the case, smashed more than 10 fan attraction and drainage studios, and seized more than 2,400 sets of criminal equipment, comprehensively cracking down on the "black and gray industry" criminal chain of attracting fans and draining overseas fraud groups. The case is currently at the prosecution stage. (Cyber Security Bureau, Ministry of Public Security)
Artificial intelligence cannot be used as a patent inventor
On April 26, 2024, the State Intellectual Property Office announced the top 10 patent reexamination invalidation cases in 2023. Among them, the rejection decision of the patent "food container and device and method for attracting and enhancing attention" designed by DABUS was upheld after the reexamination procedure. The collegial panel pointed out that the purpose of the patent application approval process is not only to examine whether the invention-creation should be granted, but also to determine the rights related to the invention-creation. Only civil subjects who meet the requirements of the Civil Law can be the right holder of the relevant civil rights of the inventor. The collegial panel of the Chinese Patent Office held that DABUS is an "artificial intelligence system", which is not one of the three types of civil subjects stipulated in the Civil Law, and cannot exercise rights and perform obligations as a civil subject, so it cannot be identified as an inventor in the patent administrative procedure. (Enterprise Patent Watch)
The European Commission said that it will launch an antitrust lawsuit against Apple this year
On April 29, 2024 local time, the European Commission designated iPadOS as a "gatekeeper" service under the rules of the Digital Markets Act. After eight months of investigation, the European Commission believes that iPadOS is an important gateway for business users to reach end users, and that Apple enjoys an entrenched and enduring position in iPadOS. Apple now has six months to ensure full compliance with its Digital Market Access Act obligations applicable to iPadOS, including allowing European users to download apps from outside the App Store, uninstall apps pre-installed on iPads, and select default services such as browsers from select screens. If Apple fails to comply, the company could face fines of up to 10% of its global revenue, and up to 20% if it repeats violations. （The Verge）
The Federal Communications Commission (FCC) penalized four telecommunications carriers for illegally sharing users' location data
On April 29, 2024 local time, the Federal Communications Commission of the United States imposed a total fine of $200 million on AT&T, Verizon, T-Mobile, and Sprint for illegally sharing users' location data without their consent. The FCC filed a lawsuit against the U.S. carriers in 2020 and is currently enforcing the fine judgments. The FCC said the operators sold access to customer location data to aggregators, who in turn sold the information to third parties, "in an attempt to shift the obligation to obtain customer consent to downstream recipients of location information," meaning that in some cases valid customer consent was not obtained. （CNET）
Industrial development
Japan's Fujitsu Corporation and others jointly announced the development of 6G equipment
On April 11, a number of Japanese telecommunications companies, including DOCOMO, NTT, NEC, and Fujitsu, jointly announced the development of the world's first high-speed 6G wireless device. Its data transfer speed is as high as 100Gbps per second, which is 10 times the peak speed of 5G and more than 500 times the current download speed of ordinary 5G smartphones. It is reported that DOCOMO, NTT Corporation, NEC Corporation and Fujitsu have been developing this device since 2021. Although the transmission distance is currently limited and the test is only in the range of 100 meters, it is expected that as technology advances, the size and cost of the device will decrease. （Medium）
Meta AI et al. French research team released new multi-token prediction technology
According to a paper published in arXiv on April 30, researchers from Meta, ParisTech Pont University and Université Paris-Saclay have proposed a way to improve the accuracy and speed of AI's large language models (LLMs) by predicting multiple characters (tokens) at the same time. This is contrary to the classical structure of autoregressive language models, which are designed to predict one token at a time. However, multi-token prediction is not suitable for all types of model and language tasks, but it offers significant advantages in some areas, with a 3x increase in inference speed and better performance on generative tasks. While there is still room for improvement, this technology could be a powerful tool for some LLM applications. （arXiv）
OpenAI｜下周将发布ChatGPT搜索引擎
On May 3, Lior S, a former researcher at Mila in the artificial intelligence laboratory and a lecturer at MIT, broke the news that according to OpenAI's latest SSL certificate logs, OpenAI has created a search.chatgpt.com subdomain (currently inaccessible) and will enter the search engine to comprehensively challenge Google, the global hegemon in this field. At present, OpenAI has updated the homepage of its website to add a search box for "ask ChatGPT anything". The launch of OpenAI's search engine may pose a challenge to Google's search engine, especially at the user experience and technical level. (Reuters)
Researchers at the University of Texas have developed a stretchable electronic skin
On May 3, researchers at the University of Texas at Austin announced the development of an innovative stretchable electronic skin (e-skin) that imparts robots and other devices with similar softness and tactile sensitivity to human skin. This technology breaks through the problem of the existing technology to reduce the sensing accuracy when the material is stretched, and the pressure response of the electronic skin can remain stable no matter how it is stretched, ensuring accurate perception. Currently, the research team is working with computer science experts to develop a robotic arm equipped with an electronic skin and has filed a provisional patent for the technology. They are also actively looking for opportunities to partner with robotics companies to bring this technology to market. (University of Texas at Austin)
The National Institute of Standards and Technology will establish a new chip manufacturing research institute
On May 5, the U.S. National Institute of Standards and Technology (NIST) announced that it is creating a new U.S. Chip Manufacturing Institute that will leverage a digital twin approach to dramatically accelerate the development and validation of next-generation semiconductor manufacturing, advanced packaging, assembly, and test processes. NIST said it expects to provide up to $285 million in funding for the institute for a period of up to five years. NIST has issued a notice of funding opportunities for the project and is inviting industry to attend the May 8 briefing. （NIST）
China and France issued a joint statement on AI
On May 5-7, the heads of state of China and France reached a consensus on the basis of the consensus reached in the Sino-French Joint Statement on April 7, 2023, especially on artificial intelligence. China and France emphasize that AI must serve the public interest, and that the development and use of AI by all countries must be consistent with the purposes and principles of the UN Charter. The two sides emphasized strengthening international cooperation to bridge the digital divide and improve AI capabilities in developing countries. Both sides recognize that in order to achieve the Sustainable Development Goals, AI must be actively developed. These should include sustainable development, climate and biodiversity conservation, agricultural production, education and health for all. (Ministry of Foreign Affairs)
Apple has launched the iPad Pro with the new M4 chip
On May 7, Apple unveiled its latest iPad series at a virtual launch event. On the new iPad Pro, Apple uses a new Apple chip called the M4. It is reported that Apple's M4 chip is based on the second-generation 3nm process, with a new display engine, using 4 large cores + 6 small core CPU, claiming that the CPU speed is up to 50% higher than M2, and ray tracing is supported for the first time. In addition, Apple's M4 chip is also equipped with a 16-core neural network engine, with a speed of 38 trillion operations per second. (Apple)
MITRE|, a U.S. nonprofit, has announced that the AI sandbox it has created will be available to U.S. federal agencies
On May 7, the U.S. nonprofit MITRE announced that the AI sandbox it built is expected to be available to U.S. federal agencies by the end of this year. MITRE, which operates a U.S. federally funded R&D center, announced at the National Competitiveness AI Expo that it expects the AI sandbox to be applied in areas of "national security, healthcare, transportation, and climate." The AI sandbox will be powered by NVIDIA's AI data center infrastructure. （MITRE）
Apple is developing AI chips for its data centers
On May 7, it was reported that Apple is developing artificial intelligence chips for its data centers. This strategic move could give Apple a dominant position in the intensifying AI race. According to insiders, Apple's server project is internally codenamed the "Apple Chips in Data Center" (ACDC) project, and Apple may use the chip for its own servers. The ACDC project has been in research and development for several years, but the specific launch time of the new AI chip has not yet been determined. (Wall Street Journal)
OpenAI｜开发DALL-E 3生成图像检测工具
On May 7, OpenAI announced the launch of a new tool designed to detect whether digital images were created by DALL-E 3, which it developed. OpenAI says that in internal tests, the tool was able to correctly identify images created by DALL-E 3 about 98% of the time, and can handle some common modifications such as compression, cropping, and saturation changes, among others. OpenAI also plans to add tamper-evident watermarks to mark digital content such as photos or audio, and these marks are difficult to remove. （OpenAI）
Google released AlphaFold 3, an artificial intelligence model
On May 8, Google launched its revolutionary AI model, AlphaFold 3. AlphaFold 3 is able to predict the structure and interactions of all living molecules. AlphaFold 3 is at least 50% more accurate than existing methods in predicting protein interactions with other molecules, and even 1x better for some interaction classes. This makes AlphaFold 3 the first AI system to go beyond a prediction tool based on real biomolecular structures. This leap could impact everything from the development of biorenewable materials and more resilient crops, to the acceleration of drug design and genomics research, bringing about a revolution in bioscience research. (Google)
68 software manufacturers signed the U.S. Cybersecurity and Infrastructure Security Agency's "Security by Design" commitment
On May 8, 68 of the world's leading software manufacturers, including Amazon Web Services, Cisco, Google, Hewlett Packard Enterprise (HPE), IBM, Microsoft, and top cybersecurity companies including Cloudflare, CrowdStrike, Fortinet, and Palo Alto Networks, signed the U.S. Cybersecurity and Infrastructure Security Agency's (CISA) "Secure by." Design)" commitment. CISA Director Jen Easterly said the move follows the U.S. National Cybersecurity Strategy, which calls for shifting responsibility for cybersecurity from individuals, small businesses, and local governments to more capable entities such as technology manufacturers. （CISA）
Alibaba Cloud officially released the Tongyi Qianwen version 2.5 model
On May 9, Alibaba Cloud officially released Tongyi Qianwen 2.5, and the model performance has fully surpassed GPT-4 Turbo and become the strongest Chinese large model on the surface. At the same time, Tongyi Qianwen's 110 billion parameter open-source model has achieved the best results in multiple benchmarks, surpassing Llama-3-70B and becoming the most powerful model in the open source field. (Caixin)
Artificial intelligence startup Hugging Face open-sourced its robotics codebase, LeRobot
On May 10, it was reported that artificial intelligence startup Hugging Face announced the open source of its robot codebase LeRobot, which marks a major positive for the field of artificial intelligence robot research and development. Previously, Hugging Face has successfully attracted former Tesla scientist Remi Cadine to lead this new open source project. Remi Cadene said the next step in AI is to apply it to the physical world, and the goal of the LeRobot project is to push the community to work together to build AI robots that will be open to all. He also demonstrated some of the capabilities that LeRobot can give robots, including training robots to grasp objects and navigate in unknown spaces. （Medium）
The UK Institute for AI Safety has launched an AI testing and evaluation platform
On May 10, the UK's AI Safety Institute launched an AI testing and evaluation platform to the global AI community. According to the UK's AI Safety Institute and the UK's Department for Science, Innovation and Technology (DIST), the platform, called "Inspect", aims to facilitate security innovation in AI models. The UK AI Safety Institute said the platform will accelerate global AI safety assessment efforts and improve the quality of security testing to develop safer AI models. (Artificial Intelligence Safety Institute, UK)
富士通等|发布大语言模型Fugaku-LLM
On May 10, a team of researchers from Tokyo Institute of Technology, Fujitsu and other companies announced the development of Fugaku-LLM, a large language model with advanced Japanese processing capabilities, using the Japanese supercomputer Fugaku. With 13 billion parameters, Fugaku-LLM is the largest large language model in Japan. Currently, Fugaku-LLM is publicly available on GitHub and Hugging Face, allowing external researchers and engineers to use the model for academic and commercial purposes in compliance with the license agreement. （Japan Times）
Shanghai Lingang|The section from the main urban area of the new area to Pudong Airport has been approved for road test of intelligent taxi autonomous driving
On May 10, the high-speed elevated autonomous driving road test permit from the main urban area of Shanghai Lingang to Pudong Airport was officially issued, and SAILCO was the first to obtain the test license, and will carry out long-distance and large-scale road test activities to continuously explore new scenarios for the commercial application of autonomous driving. (Shanghai Lingang)