Babbitt | Metaverse daily must-reads: Shanghai releases an action plan for key technologies of Metaverse, focusing on immersive technology and Web3 technology, AIGC, digital human, spatial computing, etc. are also key words

Abstract: On June 13, the Shanghai Municipal Science and Technology Commission issued the "Shanghai "Metaverse" Key Technology Research Action Plan (2023-2025)". The plan proposes to focus on immersive technology and Web3 technology as the two main directions, with independent innovation and open collaboration as the promotion path, and strives to improve the self-reliance and self-improvement of science and technology in the "Metaverse" field. Create new heights in key technical fields such as blockchain.

Image credit: Generated by Unbounded AI tools

Hot info:

Hangzhou Asian Games Organizing Committee: Intelligent preparation of the Asian Games throughout the process, launching the first digital torchbearer in the history of the Asian Games

According to a report by the Financial Associated Press on June 14, the State Council Information Office held a press conference today on the preparations for the Asian Paralympic Games in Hangzhou. Hangzhou Mayor Yao Gaoyuan said at the meeting that 100 days later, the 19th Asian Games will be grandly opened in Hangzhou. This is the third time that a Chinese city has hosted the Asian Games after Beijing and Guangzhou. Smart preparations for the Asian Games throughout the process. Iteratively upgrade the digital competition platform "Asian Games Online", create a one-stop digital viewing service platform - "Smart Asian Games One-Stop", launch the first digital torchbearer in the history of the Asian Games, launch the Asian Games Pass, implement automatic driving, smart buses, etc. Application, use digital means to improve the efficiency of preparation work and service guarantee efficiency, and fully demonstrate the achievements of "Digital China" construction.

AMD announced the new MI300X AI chip, the number of GPUs required for AI model training will be less

According to a report by the Financial Associated Press on June 14, AMD announced the new MI300X AI chip; the new chip means that the generative AI model no longer needs so many GPUs; the company also announced the MI300A chip, which is shipping in small batches. To be used in the "El Capitan" supercomputer in the United States.

Xueersi responds to the incident of "Bishen Composition Claiming Its AI Large Model Steals Data": The calls are in line with the contract requirements

According to a report from IT Home on June 14, yesterday, Bishen Composition issued a statement on the infringement incident of the "Xueersi" AI large model. Bishen Composition believes that Xueersi uses "crawlers" to train with the data of the Penshen Composition App server, and requires Xueersi to pay 1 yuan in compensation, publicly apologize and delete the obtained data. Xueersi made a reply, claiming that "the data calls are in line with the pre-agreed contract requirements". As previously reported, Bishen Composition claimed that the Xueersi AI model had stolen its data.

Economic Daily: To build an artificial intelligence large-scale model industry ecology should focus on solving core problems such as uncontrollable high-end computing power technology

Babbitt News, on June 14, the Economic Daily published an article "Building an Artificial Intelligence Large-scale Model Industrial Ecology". The article stated that to create a good artificial intelligence large-scale model industry ecology, efforts should be made to solve the core issues such as uncontrollable high-end computing power technology, few large-scale high-quality data sets, and weak algorithm reliability and interpretability faced by the development of large-scale artificial intelligence models in my country. question. First, promote the construction of new computing power ecological infrastructure. Implement a major special support plan for the research and development of high-end computing power chips, and encourage leading companies to join hands with high-level universities to "unveil the list and take command". Second, build a large-scale and high-quality data ecosystem. Thirdly, create a stable and reliable algorithm open source ecosystem. Finally, strengthen the strategic planning and in-depth governance of large artificial intelligence models. Formulate a development plan for the artificial intelligence large-scale model industry, and provide corresponding policies, funds, and talents for large-scale model technology research and development and industrial applications.

The Japanese Actors Union proposes AI legislation, calling for the establishment of "sound portrait rights"

According to IT House's report on June 14, the Japan Actors' Union, which is committed to protecting actors and seiyuu (voice actors), announced yesterday a "proposal on the use of generative artificial intelligence technology." The proposal specifically calls for industry and government to review the operation of copyright law and develop rules to establish "sound likeness rights".

The content of the proposal mainly includes the following points: 1. Actively exchange opinions with domestic and foreign countries, and formulate relevant laws in Japan with reference to EU artificial intelligence regulations; 2. Review the operation of copyright laws; 3. Clearly mark AI-generated works; 4. Formulate Specific rules when applying AI to the field of expression; 5. Establishing sound portrait rights.

The Japan Actors Union also stated: "It is desirable to promote the development of human society through the development of new technologies. We strongly hope that new guidelines and laws will be formulated based on the concept that 'human beings are the main body, and AI will be used for help humanity'". Currently in Japan, AI-generated images of actresses and female voice actors have appeared, and there are even unofficially authorized voice changers for AI voice actors for sale, so relevant legislation needs to be improved urgently. In Japan, "sound rights", as a concept similar to voice portrait rights, were mentioned and discussed during the period of rapid development of speech synthesis technology.

Meta releases "human-like" AI model I-JEPA that can analyze and complete unfinished images more accurately than existing models

According to news from IT House on June 14, social media giant Meta (formerly Facebook) announced on Tuesday that it will open up to researchers some components of a new "human-like" artificial intelligence model that can be better than existing models. Accurately analyze and complete unfinished images. The model, called I-JEPA, uses background knowledge of the world to fill in missing parts of images, rather than just inferring from nearby pixels like other generative AI models. The approach, which employs a style of human-like reasoning championed by Meta's chief AI scientist Yann LeCun, helps avoid common mistakes in AI-generated images, such as having an extra finger.

Former JD Vice President Mei Tao founded HiDream.ai, aiming to surpass Stable Diffusion in basic models by the end of the year

According to the news from the Heart of the Machine, Jingdong AI expert Mei Tao resigned from the position of vice president of Jingdong Group before the Spring Festival this year, and is determined to devote himself to the large-scale model business. One month later, HiDream.ai was established, and the company's direction was determined as the basic model and application of generative visual multimodality. It is reported that HiDream.ai's model can not only support Wensheng map, but also support Wensheng video, picture video and Wensheng 3D content. Among them, Wenshengtu products support input to generate pictures, and can further edit the parts that you want to modify in any picture. For example, it is required to modify the background, subject, etc. of the generated picture. The initial goal of Mei Tao's team is to surpass the latest version of Stable Diffusion in terms of basic models before the end of the year, and to surpass MidJourney in terms of products.

Before starting the business officially, Mei Tao, Zhou Bowen, and He Xiaodong were the leaders of JD.com's AI department, responsible for the research and development of JD.com's photo shopping, multi-modal content review and generation, multi-modal digital humans, and intelligent robotic arms. Earlier, Mei Tao worked at Microsoft Research Asia for 12 years, supporting AIGC projects such as Microsoft Image, Video Search and Summary, and Office Graphics and Text Mixing.

Google will delay the first launch of AI chat tool Bard in the EU market due to regulatory concerns about privacy

Google will delay the launch of its artificial intelligence chatbot Bard in the EU after EU regulators raised privacy concerns, Ireland's Data Protection Commission said, Politico reported. Ireland's Data Protection Commission said Google did not share enough about how its artificial intelligence tool protects the privacy of Europeans to justify its rollout in the EU. The issue is under an "ongoing investigation" by the Irish Data Protection Commission.

Investment and Financing:

British start-up company Synthesia's latest financing is about 90 million US dollars, and it is promoted to AIGC unicorn

According to the Financial Associated Press, on June 13, local time, Synthesia, a digital media platform headquartered in London, UK, announced that it had obtained approximately US$90 million in the latest round of financing, with a valuation of US$1 billion, and was officially promoted to "Single Horn". beast" enterprise. It is understood that Synthesia was established in 2017. The main business of the platform is to help users create videos using artificial intelligence technology. The virtual avatars in the videos can make corporate introductions, training videos, etc. in more than 120 different languages.

French artificial intelligence startup Mistral AI raises 105 million euros

French artificial intelligence startup Mistral AI has raised 105 million euros ($113 million) in its first round of funding, making it a "new leader" in the field of artificial intelligence, the French news agency reported on June 14. global players". It is reported that the company is a new company formed by former DeepMind and Meta researchers, and will try to compete with OpenAI and Google to develop generative artificial intelligence infrastructure. Mistral’s funding is one of the largest seed rounds ever raised by a European generative artificial intelligence company.

Daily must-reads in Metaverse:

"Shanghai released <Shanghai "Metaverse" key technology research action plan (2023-2025)>, focusing on two main directions of immersive technology and Web3 technology"

On June 13, the Shanghai Municipal Science and Technology Commission issued the "Shanghai "Metaverse" Key Technology Research Action Plan (2023-2025)". The main direction of the program includes immersive technology and Web3 technology. Aiming at the immersive, open, sustainable real-time, human-centered and other cross-border composite technical characteristics of the "Metaverse", it focuses on the content, storage, calculation, transmission and Terminal and other technical aspects, combined with domestic and foreign industrial development and the city’s research and development foundation, focusing on the two main directions of immersive technology and Web3 technology, and building a new heights.

"Artificial intelligence helps Silicon Valley usher in a "singularity" moment? Human intelligence can be increased by 1 million times in 2045"

Kurzweil, the biggest advocate of the singularity theory, predicts that by 2030, computer systems will pass the Turing test and be indistinguishable from humans. Fifteen years from now, the real singularity will come, when "computing will become part of us and our intelligence will increase by a million times." Critics of the opposition argue that even the astonishing progress of large language models is a far cry from the immortal, long run of world wisdom promised by the singularity theory.

"OpenAI has been updated heavily, and it has become stronger again! The API has a terrifying function call capability, can handle longer contexts, and the price has dropped by 75%! "

On June 13, OpenAI's official website suddenly released the latest capability update of ChatGPT, including the addition of a new function call capability in the Chat Completions API, the model update can handle longer text, and the price of the most advanced embedding model has been reduced by 75% %wait.

"How strong is China's generative AI? Analyzing the Gap between China and the United States"

How strong is China in generative AI? This is the title of an article published by the British political magazine "The Economist" recently. The article compares the strengths of the two countries based on the four basic levels of the number of papers, the number of systems, computing power, and chip hardware. It is believed that China’s large-scale models currently lag behind the United States by two to three years. There are gaps in hardware such as chips and scientific and technological talents, but the distance is not insurmountable, and the real competition lies in applications.