Abstract: According to a report from the Financial Associated Press on September 20, today, the automotive metaverse base "Yuanchuang Bay" built by Shanghai Jiading in Anting was officially unveiled. According to the Anting Automotive Yuanverse industry planning goals, by 2025, the Automotive Yuanverse Town will begin to take shape and achieve an industry scale of more than 100 billion yuan; cultivate 2 leading companies with an output value of more than 10 billion yuan, and cultivate 2 leading companies with an output value of more than 2 billion yuan. There are 10 high-growth enterprises. In addition, special parks such as automotive chip industrial parks will be built, with a carrier area of 100,000 square meters.
Image source: Generated by Unbounded AI
Hot info
Shanghai Jiading Anting Automotive Metaverse Base was unveiled, with the goal of industrial scale exceeding 100 billion yuan in 2025
According to a report from the Associated Press on September 20, at the 2023 Second World Metaverse Conference held today, the automotive metaverse base "Yuanchuang Bay" built in Anting, Jiading, Shanghai, was officially unveiled. According to the Anting Automotive Yuanverse industry planning goals, by 2025, the Automotive Yuanverse Town will begin to take shape and achieve an industry scale of more than 100 billion yuan; cultivate 2 leading companies with an output value of more than 10 billion yuan, and cultivate 2 leading companies with an output value of more than 2 billion yuan. 10 high-growth enterprises. Special parks such as automotive chip industrial parks will also be built, with a carrier area of 100,000 square meters.
Meng Wanzhou: AI development is crossing an inflection point, and Huawei supports the "letting of a hundred flowers bloom" of large models in the intelligent era
According to a report by the Science and Technology Innovation Board Daily on September 20, at the Huawei Full Connect Conference, Huawei Vice Chairman, Rotating Chairman, and CFO Meng Wanzhou said, “Computing power is the core driving force for the development of artificial intelligence. Large models It requires a lot of computing power, which determines the speed of AI iteration and innovation, and also affects the speed of economic development. The scarcity and expense of computing power have become the core factors restricting the development of AI. Huawei is committed to building China's solid computing power A base, a second choice for building the world.”
In addition, Meng Wanzhou said, "Artificial intelligence neural network models with parameters exceeding 100 billion or even trillions are accelerating into thousands of industries, and AI development is also crossing an inflection point. From the small model era to the large model era, the practicality of AI technology A qualitative leap has occurred. In the past, different application scenarios required the development of different models. Now, large models absorb massive amounts of knowledge, and one model can adapt to multiple business scenarios, significantly lowering the threshold for AI development and application and shortening the technical process. to the application cycle, making AI move from workshop-style development and scenario-based customization to industrial development and scenario-based tuning, making it possible to solve industry problems on a large scale relying on large models. In this process, through the computing power base, AI platform, development With the opening of tools, Huawei supports the "letting of a hundred flowers" of large models in the intelligent era, and strives to make a "hundred gardens" of black soil. We support each organization to use its own data to train its own large models, allowing each industry to use its own We use our expertise to develop our own industry model.”
OpenAI announces open recruitment for red team network
Babbitt News, on September 19, OpenAI announced an open recruitment for the OpenAI Red Team Network, inviting field experts who are interested in improving the security of OpenAI models to rigorously evaluate their artificial intelligence models. OpenAI stated that all members of the red team network will be compensated for their contributions when participating in red team projects. Joining the network will not limit publishing research results or pursuing other opportunities, but any participation in red team and other projects will generally be subject to Non-Disclosure Agreement (NDA), or indefinite confidentiality.
DeepMind's new AI model is expected to solve human genetics problems
According to the "Kechuangban Daily" report, on September 19, local time, Google DeepMind announced that its research team created AlphaMissense based on the AlphaFold methodology - by using the protein sequence database and variant structure background, it can identify disease-causing missense mutations and The causative gene is unknown. AlphaMissense successfully predicted the pathogenicity of 216 million possible single amino acid changes in 19233 standard human proteins, resulting in predictions of 71 million missense mutations. Subsequently, AlphaMissense successfully predicted 89% of missense mutations, compared with only 0.1% of mutations that have been confirmed by human experts.
Xiaomi Wang Bin: In the future, large and small models will coexist, and general and special-purpose models will coexist.
According to a report by IT House on September 20, Wang Bin, director of Xiaomi Group's AI laboratory and chief scientist of natural language processing (NLP), said in an interview with Shengdonglively that the future must be a state of coexistence of large and small models, and the coexistence of general and special-purpose models. . Wang Bin said that unlike other companies, Xiaomi has a large number of devices, and the computing power of these devices has been greatly improved. For example, the computing power of the chips on mobile phones is much higher than before. Secondly, Xiaomi is a company that focuses on end-users. If it wants to use a large amount of cloud computing power, the cost will be relatively high. In addition, user privacy and network conditions when users use mobile phones are also factors we consider. However, if the parameters of a large model are too small, it will also affect the user experience. Therefore, Xiaomi needs to strike a balance between the model scale and the hardware threshold.
NFT market LimeWire acquires AI image generation platform BlueWillowAI
According to a report by Decrypt on September 20, NFT market LimeWire announced today that it has acquired the AI image generation platform BlueWillowAI. This deal was reached after LimeWire launched LimeWire AI Studio last month. BlueWillow says it has more than 2.3 million active members and has created more than 500 million images to date. LimeWire plans to integrate BlueWillow's proprietary text-to-image and image-to-image AI models into its AI Studio. LimeWire co-CEO Julian Zehetmayr said the acquisition of BlueWillow and the integration of its artificial intelligence capabilities will help shape the future of content creation and creativity.
Shanghai AI Laboratory releases XTuner, a large model training toolbox, lowering the threshold for large model training
Babbitt News, recently, Shanghai Artificial Intelligence Laboratory (Shanghai AI Laboratory) released XTuner, an open source toolbox for large model training, which once again lowered the threshold for large model training. It is reported that XTuner focuses on the fine-tuning process, providing a lightweight fine-tuning framework for various open source large models, once again consolidating the practical tool attributes of the full-chain open source system. XTuner supports the adaptation of multiple levels of hardware. Developers only need to use a minimum of 8 GB of consumer-grade video memory to train an "exclusive large model" suitable for specific demand scenarios.
Tencent’s open source StableDiffusion plug-in LightDiffusionFlow can save all workflow data with one click
Babbitt News, according to a report on Tencent Open Source’s official public account on September 19, Tencent recently launched an open source plug-in called LightDiffusionFlow based on the AI painting open source platform StableDiffusion webUI (hereinafter referred to as SD), which has now been launched on GitHub. The plug-in is said to "help users save all workflow data with one click. Next time they use it, they can just drag in the Flow file and quickly reproduce the entire workflow, just like using Photoshop's PSD mockup file."
Huawei releases Ascend AI computing cluster Atlas 900 SuperCluster, which can support large model training with over one trillion parameters
According to Jiemian News, on September 20, at the Huawei Full Connection Conference, Huawei announced the launch of the Ascend AI computing cluster Atlas 900 SuperCluster with a new architecture. It is reported that the new cluster can support large model training with over one trillion parameters. It uses the new Huawei Galaxy AI intelligent computing switch CloudEngine XH16800. With its high-density 800 GE port capability, the two-layer switching network can achieve 2250 nodes (equivalent to 18,000 cards) ultra-large-scale non-convergence cluster networking. The new cluster also uses an innovative super-node architecture to improve large model training capabilities.
Zuckerberg: Will build an AI GPU cluster with more than 1,000 H100 GPUs for life science research
According to the Science and Technology Innovation Board Daily, on September 19, local time, Meta founder and CEO Zuckerberg posted a blog stating that his charity project plans to build an AI GPU cluster and use AI systems for life science research. The foundation of Mark Zuckerberg and his wife Priscilla Chan is funding the construction of a large-scale computing system for medical research. The system will be equipped with more than 1,000 GPUs and will use the top-end H100. New computing system allows researchers to use generative AI to model healthy and diseased cells. Over time, this could help them develop new ways to treat the disease.
Investment and Financing
“Zhipu AI” completed the B-4 round of financing with participation from Tencent and Alibaba, with a valuation reaching approximately US$1 billion
According to 36Kr’s report on September 19, the large model company “Zhipu AI” recently completed the B-4 round of financing. Among them, Tencent Venture Capital, Alibaba Venture Capital and many other institutions participated in the investment. After the completion of this round of financing, Zhipu AI's valuation reached approximately US$1 billion. Previously, Zhipu AI completed the B-2 round of financing of several hundred million yuan, exclusively invested by Meituan War Investment.
Tsinghua University’s large-scale AI model “Bohan Intelligence” received nearly 100 million yuan in Series A+ financing
According to reports from the investment community on September 20, Bohan Intelligent, a large manufacturing AI model manufacturer, recently completed an A+ round of financing of nearly 100 million yuan. The investment was jointly led by Gimpo Investment and CITIC Capital, and followed by Rongyu Venture Capital and Zhuoyuan Capital. The funds raised in this round will be used for business research and development and market expansion in areas such as large model implementation and intelligent manufacturing. Founded in 2019, Bohan Intelligence focuses on end-to-end large model basic platforms for artificial intelligence and provides machine learning and deep learning products and solutions for scenarios such as intelligent manufacturing and autonomous driving.
Startup Afterparty has completed US$5 million in financing and plans to launch an AI digital human
According to a report by Science and Technology Innovation Board Daily on September 20, American blockchain startup Afterparty recently raised US$5 million in a new round of financing. The company is using the funding to launch a new platform called Afterparty AI, which creates AI-powered digital beings of creators that can individually respond to fan messages. Using open source technology, Afterparty built a proprietary large-scale language learning model that is trained on hours of content from creators to capture their personalities and the way they speak.
AI reading startup Edsoma completes US$2.5 million in seed round financing, led by NBA superstar O'Neal
According to TechCrunch on September 20, AI reading and education startup Edsoma raised US$2.5 million in a seed round led by NBA superstar Shaquille O'Neal, with a post-financing valuation of US$14 million.
Kyle Wallgren, the company's founder and CEO, said Edsoma started with only 300 paying users, which later grew to 1,000 and 9,000 in the second month. At present, he has considered expanding the scale, raising Series A financing, and even expanding the business to other languages. He plans to launch a Series A round of funding in a few months, targeting $10 million to $15 million.
HiddenLayer completes US$50 million in financing to further strengthen the defense capabilities of enterprise AI models
According to a report by VentureBeat on September 19, network security startup HiddenLayer announced that it has received US$50 million in Series A financing to further strengthen the defense capabilities of artificial intelligence models adopted by enterprises. The round was led by M12, Microsoft's Venture Fund and Moore Strategic Ventures, with participation from Booz Allen Ventures, IBM Ventures, Capital One Ventures and Ten Eleven Ventures. Currently, HiddenLayer has helped multiple Fortune 100 companies protect the AI/ML models they use in fields such as finance, government, defense, and cybersecurity.
AI startup Darrow completes US$35 million in financing, led by Georgian
According to TechCrunch, on September 19, AI startup Darrow announced the completion of US$35 million in financing. This round of financing was led by B2B expert Georgian, with participation from F2 and previous backers Entrée Capital and NFX. Including this latest round, Darrow has raised nearly $60 million from investors including Y Combinator and R-Squared Ventures. The new funding will be used to hire more engineers and business development talent; add more focus areas to its search and analytics tools; and invest in expanding its large language models and other technology assets.
It is reported that Darrow has developed an artificial intelligence-based data engine that can draw from large amounts of public documents and search for the possibility of class action lawsuits in areas such as data privacy violations and environmental pollution. Darrow said the total number of claims currently being filed as a result of its data insights is about $10 billion.
Recommended Reading of the Metaverse
"Large model To C has been launched, and the champagne is opened at halftime for AI commercialization?" 》
There has indeed been a buzz about the concept of large models, but at present, the bustle has faded and the market does not seem to buy it. On August 15, the "Generative Artificial Intelligence Service Management Measures (Draft for Comments)" came into effect, and the road to AIGC product compliance was clear. As a result, domestic large models were intensively released, including Baidu, iFlytek, SenseTime, etc. Players began to grab the ToC market. So can the "first large model application" on people's mobile phones become the "first entrance" to AI generation? It’s worth looking into.
https://www.8btc.com/article/6833382
"How can enterprises achieve AI transformation?" After studying hundreds of companies, we found the answer》
An article in the Harvard Business Review “What makes a company successful in using AI? "What Makes a Company Successful at Using AI?" provides a "Guide to Successfully Crossing the River." The article adopts a study by McKinsey and MIT’s Machine Intelligence in Manufacturing and Operations Initiative (MIMO), which tracks the progress of 100 companies (involving various industries such as automobiles and mining) in digitalization, data analysis and machine intelligence (MI) technologies. Based on the performance of goals, actions, and results, it can be concluded that leading digital intelligence enterprises have certain commonalities in five aspects: governance, deployment, partners, personnel, and data availability.
https://www.8btc.com/article/6833388
"Google version of ChatGPT is fully integrated with its email, videos, maps, etc.! 》
On September 20, Google announced the launch of Bard Extensions on its official website. With this extension, users can use Bard in Google's Gmail, Google Docs, Netdisk, Google Maps, Video and other products.
https://www.8btc.com/article/6833392
Babbitt Park is open for cooperation!
Chinese Twitter: https://twitter.com/8BTC_OFFICIAL
English Twitter: https://twitter.com/btcinchina
Discord community: https://discord.gg/defidao
Telegram channel: https://t.me/Mute_8btc
Telegram community: https://t.me/news_8btc
