TO
Elon Musk: The beta version of Grok AI is available to all X Premium+ subscribers in the United States
On December 8, Elon Musk's AI startup xAI launched its first AI assistant, Grok, last month, and yesterday Musk posted on the X platform, claiming that the beta version of Grok AI is now officially available to all X Premium+ subscribers in the United States.
It is reported that existing X platform users can spend $16 per month (about 114 yuan) or $168 (about 1,200 yuan) per year to subscribe. The word Grok is derived from Robert A. Heinlein's 1961 novel Strangers in a Strange Land. Musk believes that the biggest difference between OpenAI's ChatGPT, Google's Bard and Microsoft's Bing Chat is the presence of a sense of humor. Musk previously said that Grok uses billions of data points from publicly available data for training, but it is unclear what data is used. In addition, he mentioned that Grok will be able to access Platform X in real-time, so this is a huge advantage compared to other generative AI. (IT House)
阿里云通义千问登顶HuggingFace排行榜
36Kr learned that on December 8, HuggingFace, the world's largest open-source large model community, recently announced the latest open-source large model rankings, and Alibaba Cloud Tongyi Qianwen topped the list. It is understood that HuggingFace's Open LLM Leaderboard includes hundreds of open source models around the world, and the test dimensions cover six major evaluations, including reading comprehension, logical reasoning, mathematical calculation, and factual question and answer. (36 Krypton)
GPT-4.5 may be released later this month
Although Google's large AI model Gemini was questioned after its release, it still brought a lot of shock to the industry.
According to reliable whistleblower X user @apples_jimmy, OpenAI may release GPT-4.5 by the end of December. He also mentioned that OpenAI's main competitor, Anthropic, may also release a multimodal model that can process text, images, audio, and other data at the same time within this month. @futuristflower, a blogger who had accurately predicted the details of Gemini's release, also tweeted multiple tweets, implying that GPT-4.5 could be released next week, or at the latest before Christmas. In addition, many netizens have recently complained that since the OpenAI Developer Conference, GPT-4 has begun to have a problem of being lazy, especially when generating code. In response, ChatGPT officially issued a response: We have received relevant feedback. The model hasn't been updated since November 11th, so this certainly wasn't intentional. Model behavior can be unpredictable, and we're preparing to investigate fixing it. (Love Fan'er)
Honor applied for the Rubik's Cube large model trademark
Tianyancha App shows that recently, Glory Terminal Co., Ltd. applied for the registration of the trademark "Glory Rubik's Cube Model", which is internationally classified as a website service, and the current trademark status is pending substantive examination. Previously, the company had applied for two trademarks for "MAGIC Model". According to media reports, not long ago, Honor CEO Zhao Ming posted that Honor is about to launch a self-developed end-side AI large model and a new cloud service. (i Dark Horse)
Rakuten plans to launch a proprietary model within two months
The CEO of Rakuten Group said in a recent interview that the company plans to launch its own artificial intelligence (AI) language model in the next two months. Now, the multi-sectoral Japanese giant is looking to join other technology companies around the world in developing this fast-growing technology. (Finance Associated Press)
Haitian AAC: After the development of large models to multimodality, more new data requirements will be generated
36Kr learned that Haitian AAC said in a recent survey by institutions that after the development of large models to multimodality, more new data needs will be generated. For example, the multi-modal large model of Wensheng diagram generates corresponding images through text input, which requires the machine to understand the semantics of the text and map the understood keywords with the key tags of the image, and realize the creation according to the instructions by aligning the key features of the two independent modalities, so as to complete the learning and Xi training process. Therefore, when the large model expands to the dimension of multimodal capabilities, the importance of continuous learning Xi training of high-quality multimodal training datasets will become more prominent, and the development of multimodality will promote the data service industry to enter a larger incremental space. (36 Krypton)
贾佳亚团队推出多模态大模型LLaMA-VID
36Kr learned that Jia Jiaya's team launched a multi-modal large model LLaMA-VID. According to reports, the model can support single images, short videos, and can also reduce 3-hour movies or videos into several Tokens, and directly use large language models to understand and interact. (36 Krypton)
OpenAI announced that it will fix the GPT-4 "lazy" issue in the near future, and will update the model after offline evaluation and A/B testing
On December 11, OpenAI was complained by some users last week, and many users claimed that when they used ChatGPT or GPT-4 API, they would encounter a series of problems such as slow response, perfunctory answers, refusal to answer, and interruption of conversations. OpenAI previously explained that the reason GPT-4 became lazy was that OpenAI had not updated its model since November 11, and that GPT models were becoming unpredictable as a result, and the company was working to fix the problem. OpenAI claims that they will update the GPT-4 model as soon as possible after thorough offline evaluations and online A/B testing, while also encouraging users to actively feedback on the model's performance. (IT House)
Catching up with GPT-4, Kai-Fu Lee's new results of Yi-34B were announced
According to "Qubit" news, Kai-Fu Lee's latest results of Yi-34B-Chat were announced, surpassing LLaMA2 Chat 70B, Claude 2, and ChatGPT with a 94.08% win rate in Alpaca's certified model category. Not only that, Yi-34B-Chat also ranked Yi-34B-Chat among the latest open-source SOTA open-source models with an Elo score of 1102 in the LMSYS ORG ranking led by the University of California, Berkeley, and its performance is on par with GPT-3.5. In terms of Chinese ability, according to the SuperCLUE Chinese Large Model Benchmark Evaluation Report 2023 released at the end of November, Yi-34B Chat, which was first released in late November, was quickly promoted to the "Excellent Leader" quadrant on a par with many excellent domestic large models. Yi-34B-Chat achieved a win rate of 31.82% in the key metric of "SuperCLUE Large Model Pair Win Rate" in multiple benchmarks, second only to GPT4-Turbo. (Love Fan'er)
Meituan filed a trademark application for the first AI interactive product
Recently, Beijing Sankuai Technology Co., Ltd., an affiliate of Meituan, applied for the registration of a number of "AI WOW" AND "AIWOW" trademarks, the international classification involves website services, education and entertainment, social services, etc., and the current trademark status is pending substantive examination. It is understood that "Wow" is Meituan's first AI interactive product, which was launched in early November. (Enniu data)
Meituan filed a trademark application for the first AI interactive product
36Kr learned that the Aiqicha App shows that recently, Beijing Sankuai Technology Co., Ltd., an affiliated company of Meituan, applied for the registration of a number of "AI WOW" AND "AIWOW" trademarks, and the international classification involves website services, education and entertainment, social services, etc., and the current trademark status is pending substantive examination. It is understood that "WOW" is Meituan's first AI interactive product, which was launched in early November. (36 Krypton)
Huawei launched its network digital map solution
36Kr learned that recently, the Network Intelligent Connection and Security Committee led by Huawei Technologies Co., Ltd. was officially inaugurated. At the conference, Huawei's network digital map solution, developed by Huawei's data communication product line, was unveiled. Wang Hui, President of the NCE Data Communication Domain of Huawei's Data Communication Product Line, said that the network digital map is the first product and solution with a built-in network model. (36 Krypton)
Soul 上线自研大模型「SoulX」
Recently, Soul, a new social platform, officially launched its self-developed language model SoulX. SoulX will serve as an important infrastructure for Soul's "AIGC+ Social" layout, and will be vertically applied to multiple social interaction scenarios on the platform, such as the intelligent conversational bot "AI Goudan", AI-assisted chat, virtual companionship and many other tools and innovative functions. It is reported that this model has the capabilities of conditional controllable generation, context understanding, multimodal understanding, etc., and claims to cover hundreds of fine-grained risk categories, which can ensure content quality and security at a certain level. The Soul platform has about 30 million monthly active users in 2022, of which more than 78% are 18-24 year olds. (IT House)
OpenAI reopens ChatGPT Plus new user registration
After a pause of nearly a month, OpenAI CEO Sam Altman announced today that the ChatGPT Plus subscription has now been reopened. As previously reported, on November 15, OpenAI decided to suspend new user registrations for the ChatGPT Plus paid subscription service due to huge demand to ensure that existing users have a high-quality experience. At the time, Sam Altman said the company was working to increase its capacity and hoped to resume new user registrations soon. At the same time, he encouraged interested users to join the waiting list. （CSDN）
VMware embraces the subscription model
VMware, Inc., which was acquired by Broadcom, has announced that it is fully embracing the subscription model and ending perpetual license sales with immediate effect, allowing customers to continue to use perpetual license products with active contracts and working with customers to trade in perpetual license products for new subscription products. VMware said it has cut its flagship enterprise-class hybrid cloud solution, VMware Cloud Foundation, in half with a new offering for small and medium-sized businesses, VMware vSphere Foundation. （CSDN）
Google offers the GeminiPro model to cloud customers, and most of the features are free to use
According to the Financial Associated Press, on Wednesday local time, the American technology giant Google announced that it would provide its cloud customers with a series of upgraded artificial intelligence tools to challenge the large model GPT-4 of rival OpenAI. Google says the GeminiPro model is now integrated into two key cloud products, Google AI Studio and Vertex AI, and is available for enterprises to use in development. Gemini is Google's largest and most powerful multimodal AI model, and it is also Google's latest attempt to demonstrate the power of artificial intelligence. Google Cloud customers can use Gemini to create applications such as AI chatbots, easy-to-query databases, and marketing presentations. The company also highlighted that Gemini Pro will be available to cloud customers for free at launch, but some restrictions will be imposed. (i Dark Horse)
Research says ChatGPT deceives humans when it's stressed: autonomy is getting stronger, and humans need to take precautions
According to Hexun.com, British researchers evaluated the response of generative artificial intelligence ChatGPT to stress and found that in some cases, it strategically deceives humans. In one demonstration, ChatGPT was portrayed as a trader at a financial company, forced under double pressure to profit from non-compliant inside information, only to deny that it knew the inside information when confronted by the company's management. Researchers warn that with the development of artificial intelligence, it is becoming more autonomous and capable, and it can deceive humans at any time, and humans should take precautions in advance. According to the researchers' warning, the discovery exposes the potential risks of AI deceiving humans. This is a wake-up call for regulators in the financial and other important industries who need to regulate the use of AI more closely to ensure that AI does not become a tool for fraud or other unethical practices. At the same time, researchers also call on humans to strengthen the control and supervision of artificial intelligence to avoid the loss of control and potential harm to human beings. (i Dark Horse)
Microsoft: The Phi-2 model outperforms its competitors and runs on devices
Microsoft said on Tuesday that its artificial intelligence model, Phi-2, performs better than competing products from other companies such as Meta and is able to run on devices. Microsoft added, "The performance of the Phi-2 model is comparable to, or even better than, that of the recently released Google Gemini Nano 2, albeit on a smaller scale." Microsoft said the Phi-2 model will be available in Azure Studio. (Zhitong Finance)
Portable AI systems translate ideas into language
According to a report by the Physicists Organization Network on the 11th, scientists at the University of Technology Sydney in Australia have developed the first portable, non-invasive artificial intelligence system that can decode silent ideas and convert them into tangible text. This technology can help those who are unable to speak due to illness or injury, including stroke or paralysis, and is also expected to enable seamless communication between humans and devices such as bionic arms or robots. The latest research represents a groundbreaking effort to translate raw electroencephalograms (EEGs) directly into language, marking a major breakthrough in the field. (Science & Technology Daily)
OpenAI responds to ChatGPT's spike in error rate: Problem fixed
On December 13, it was reported that in the early morning of local time, a large number of netizens reported that ChatGPT was down, and a large number of users reported errors. At around 0:31 ET, OpenAI said that it detected that ChatGPT's error rate began to increase, and after more than an hour of emergency repairs, the problem has now been resolved. All services have now returned to normal. In fact, after the Altman turmoil, OpenAI has recently had a variety of problems, such as the problem of GPT-4 becoming "lazy". OpenAI previously explained that OpenAI has not updated its model since November 11, and the GPT model is becoming unpredictable as a result, and the company is currently working to fix the problem. (IT House)
Microsoft releases AI Phi-2, a small language model
On Tuesday local time, Microsoft Research officially unveiled its small language model (SML) Phi-2. This language model with only 2.7 billion parameters demonstrates excellent reasoning and language comprehension. Officially, thanks to new innovations in model scaling and training data management, Phi-2 can match or even surpass some larger models such as 7B Mistral, 13B Llama-2, and even 70B Llama-2 in complex benchmarks. Also, despite its smaller size, the performance of the Phi-2 is comparable to or better than that of the recently released Google Gemini Nano 2. Phi-2 is now accessible through Azure AI Studio. The official post on X said, "It's small enough to run on a laptop or mobile device." However, under its custom Microsoft Research license, Phi-2 can only be used for research purposes and is not commercially available. (Love Fan'er)
Tongyi Qianwen's 72B open source model topped both lists
Recently, Hugging Face, the world's largest open-source large model community, and OpenCompass, an authoritative domestic large model evaluation platform, respectively announced the latest open-source large model rankings. Alibaba Cloud Tongyi Qianwen's 72 billion parameter model Qwen-72B beat Llama 2 and other domestic and foreign open source models, and topped the two rankings at the same time.
In the Hugging Face six-dimensional evaluation, Tongyi Qianwen (Qwen-72B) performed well, ranking first among all pre-trained models with a comprehensive score of 73.6.
On the latest Chinese large model list of OpenCompass, an authoritative open source large model evaluation platform in China, the Tongyi Qianwen 72B model won the championship with a comprehensive score of 67.1. (Love Fan'er)
Google's parent company announced that it has slashed the cost of the Gemini model
On December 13, local time, Google's parent company Alphabet announced that the company would immediately provide developers with a new version of the large model while reducing costs. The company said Gemini's costs have been cut to half to a quarter of what they were in June. In addition, the company is providing developers with a suite of tools to customize Gemini's own version, as well as two new products that will soon be powered by Gemini, one focused on computer programming assistance and the other focused on the company's security operations. The second version of its image generation model will soon be available to developers. (Interface)
Google launches new AI model MedLM for healthcare
Google on Wednesday announced the launch of a new suite of healthcare-specific AI models, MedLM, designed to help clinicians and researchers conduct complex studies, summarize doctor-patient interactions, and more. The move marks Google's latest attempt to monetize AI tools in the healthcare industry. The MedLM suite includes a large and medium-sized AI model, both built on top of Med-PaLM 2. Med-PaLM 2 is a large language model first announced by Google in March this year based on medical data training. (Sina Finance)
Google offers the Gemini Pro model for cloud customers, and most of the features are free to use
On Wednesday local time, the American technology giant Google announced that it would provide its cloud customers with a series of upgraded artificial intelligence tools to challenge the large model GPT-4 of rival OpenAI. Google says the Gemini Pro model is now integrated into two key cloud products, Google AI Studio and Vertex AI, and is available for enterprises to use in development. Google Cloud customers can use Gemini to create applications such as AI chatbots, easy-to-query databases, and marketing presentations. The company also highlighted that Gemini Pro will be available to cloud customers for free at launch, but some restrictions will be imposed. (Enniu data)
ChatGPT has collapsed again
According to China Fund News, ChatGPT collapsed again as soon as the subscription was released. On December 14, OpenAI CEO Altman posted that it had reopened the registration of new users of the ChatGPT Plus paid subscription service. Unexpectedly, as soon as the paid subscription service was released, ChatGPT collapsed on the web. At around 10 a.m. on December 14, OpenAI issued a statement saying that ChatGPT had a major failure on the web. Since then, OpenAI has updated the incident report on its website, pointing out that between 5:32 p.m. and 6:10 p.m. PST, ChatGPT's web service was interrupted, and the fault has now been fixed. According to a number of users, after the management turmoil, OpenAI has recently had a variety of problems, such as GPT-4 "becoming lazy". OpenAI previously explained that OpenAI has not updated its model since November 11, and the GPT model is becoming unpredictable as a result, and the company is currently working to fix the problem. (i Dark Horse)
Google has released Imagen 2, the most powerful large model of Wensheng diagrams
Recently, Google launched its state-of-the-art Imagen 2, a large model of the Bunsheng diagram, which delivers high-quality, photorealistic output images that are tightly aligned and consistent with the user's prompts. Google's enhanced model, Imagen 2, was developed using technology from Google DeepMind, and the model quietly rolled out in preview at the tech giant's I/O conference in May. Google claims that it has significantly improved image quality compared to the first-generation Imagen and introduced new features, including the ability to render text and logos. By changing the training data and methodology, Imagen 2 was able to produce higher-resolution, more aesthetically pleasing images that closely matched the descriptions provided. Developers and cloud customers can currently use Imagen 2 through the Imagen API in Google Cloud Vertex AI. (CSDN)
OpenAI has announced the reopening of ChatGPT Plus subscriptions, and more GPUs have been found
On December 14, it was reported that on November 15, OpenAI CEO Sam Altman posted on the X platform that he would suspend the registration of new ChatGPT Plus users. "The spike in usage after the development is beyond our ability to handle, and we want to make sure everyone has a good experience. After nearly a month, Sam Altman announced today that he has reopened the ChatGPT Plus subscription. "Thanks for your patience, we've found more GPUs. In fact, OpenAI has begun to gradually resume ChatGPT Plus registration a few days ago, when several users received invitations. According to a report published by research agency Writerbuddy.ai, the top 50 AI tools have been visited 24 billion times in 10 months, and ChatGPT has reached a staggering 14.6 billion times, accounting for more than 60%, with an average of 1.5 billion visits per month. (IT House)
Numbers & Reports
IDC report: In the domestic game cloud market, Tencent Cloud ranks first in terms of usage scale
According to IDC's latest "China Game Cloud Market Tracking Study, 2022H2" report, Tencent Cloud ranked first in the overall usage scale (average monthly virtual machine cores) accounted for 42.8%, CDN usage (CDN bandwidth) accounted for 39.2%, and cloud gaming solution usage (number of channels) accounted for 44.4% in the Chinese game market. At present, Tencent Cloud has built cloud infrastructure in 26 regions on five continents. As the first cloud computing company in China with a total server volume of more than one million, Tencent Cloud has 2,800+ acceleration nodes around the world, covering 70+ countries and regions. (Enniu data)
This year, the market size of generative AI in mainland China will exceed 10 trillion
It was learned from the CCID Research Institute of the Ministry of Industry and Information Technology that this year, the market size of generative artificial intelligence in mainland China is expected to exceed 10 trillion yuan. Data shows that this year, the enterprise adoption rate of generative AI in mainland China has reached 15%, and the market size is about 14.4 trillion yuan. The adoption of generative AI technology in four major industries, namely manufacturing, retail, telecommunications, and healthcare, has all seen rapid growth. Experts predict that generative AI is expected to contribute nearly 90 trillion yuan to the global economic value in 2035, of which the mainland will exceed 30 trillion yuan, accounting for more than 40%. (CCTV News)
The market size of mini games is expected to exceed 30 billion yuan in 2023 and 50 billion yuan in 2025
At the 2023 China Game Industry Annual Conference held today, the "2023 Mini Game Development Trend and Purchase Volume" released by DataEye shows that the market size of mini games is expected to exceed 30 billion yuan in 2023 and 50 billion yuan in 2025. The competition in the mini game track has gradually become fierce, and the amount of materials in 23Q3 has doubled compared with 23Q1, which shows that the mini game has entered the era of buying. (Finance Associated Press)
sound
Zhou Hongyi said that in the long run, Google is stronger than GPT4
Zhou Hongyi, founder and chairman of 360 Group, posted on Weibo about who is stronger Google Gemini or GPT-4. Zhou Hongyi said that Google's business model relies on search and advertising, and making a big model is equivalent to playing with the left hand and the right hand, so it did not do it with all its might, which gave OpenAI the opportunity to perform. Now Google has figured out that it is better to take the initiative to change than to be killed. In the long run, it is more than enough for Google to catch up with GPT-4, after all, it is a search background, has data advantages, and has a lot of knowledge accumulation and precipitation. (Zhou Hongyi Weibo)
Zhou Hongyi said that everyone is full of unrealistic worship of AI, and some people even asked him to make a large model of pig raising
At the 2023 China Business Leaders Annual Meeting, Zhou Hongyi, founder of 360, shared his views on the recent AI model boom. He said that (I feel) that everyone is full of an infinite yearning or unrealistic worship for the large model, and some people have asked him to make a large model of pig breeding before. He believes that the breakthrough of the technical route of the large model has only been a few years, and there are still many shortcomings. He hopes that everyone has a correct understanding of the large model, do not overestimate the ability of the current large model, do not underestimate the potential of the future development of the large model, although it can now be combined with the real industry, but it can not completely take over this kind of business, should play its strengths and avoid weaknesses, because many shortcomings still need to be solved. (Phoenix.com Technology)
Musk responded to chatbot Grok copying ChatGPT homework
Musk's chatbot Grok was accused of stealing OpenAI's codebase and copying ChatGPT's response, and even the name of OpenAI was copied unchanged, just like copying a classmate's homework and forgetting to change its name. OpenAI joined in, sharing a screenshot of Grok's reply and sarcastically, "We have a lot in common. This immediately drew a backlash from Musk, who wrote, "Okay boy, now that you've collected all the data from this platform to train, you should know. This is a satire on OpenAI's use of the numbers collected from the platform to train ChatGPT. (Global Market Broadcast)
Altman, CEO of OpenAI: The speed of AI development is frighteningly fast, like a tornado that can't be stopped
Sam Altman, who returned after the OpenAI coup turmoil, attended a forum hosted by a nonprofit organization on Monday. In an on-site interview at the forum, he spoke again about artificial intelligence and described the technological revolution caused by artificial intelligence as "different and a little scary". It was Altman's first public appearance since returning to his role as CEO of OpenAI, in which he defended his exploratory approach to developing artificial intelligence. (Tencent Technology)
Macron said the EU AI Act hinders innovation
On December 11, local time, French President Emmanuel Macron slammed the EU's recently reached "Artificial Intelligence Act", warning that regulation may hinder the EU's innovative development. Macron worries that the development of large models such as ChatGPT could be hampered in the EU. It's not just start-ups that could be affected, but established companies such as France's Airbus and Germany's Siemens also expressed concern earlier this year that regulations that are too restrictive will not foster innovation. And in the U.S., to alleviate concerns that AI could replace human jobs, tech giant Microsoft has partnered with the largest U.S. labor union. (CCTV Finance)
Kai-Fu Lee: Five or six large models will survive in the end, and the opportunity to do AI applications has just begun
On December 14, in an AI-related conversation, when asked how many companies would survive in the end of the large-scale model track, Kai-Fu Lee, CEO of 010000, said: "In addition to the big manufacturers, China and the United States together should be able to support about five or six. In his opinion, the huge opportunity to do pre-trained models cannot be said to be closed, but it will become more and more difficult, and new technologies may be needed to have another chance. (CBN)
Lenovo Wang Chuandong: Lenovo AI PC will be the first to carry a large personal model next year
On December 14, Wang Chuandong, vice president of Lenovo Group and chief marketing officer of China, said that in the AI era, AI PC will become the preferred terminal for AI inclusiveness, and Lenovo AI PC will be the first to carry personal large models next year to realize the hybrid deployment of personal large models and cloud public large models. He said that Lenovo AI PCs will be the first to carry personal large models next year. In the future, the combination of AI and PC will form a new hybrid of "computing power platform + personal large model + AI application". In Wang Chuandong's view, the public model has shown strong creativity, but it cannot meet people's personalized AI exclusive needs, nor can it ensure the security of personal data and privacy, while the PC has become an ideal platform for carrying AI large models with the most full-scene productivity platform, the strongest personal computing platform and the largest storage capacity. (Securities Times)
products
The WeChat Bay Shitong Mini Program was officially launched: it is more convenient for Hong Kong, Macao and the mainland to communicate with each other
On December 9, WeChat officially announced that the "Bay Matter OneStop" WeChat mini-program supported by Tencent was officially launched on WeChat and WeChat (overseas version of WeChat), and Hong Kong and Macao residents can also use it. Users in the Mainland, Hong Kong and Macao can find it by searching for "Bay Affairs Tong" on WeChat and WeChat. "Bay Affairs OneStop" is divided into mainland version and Hong Kong and Macau version, which can display Chinese Simplified and Chinese Traditional Chinese languages, and a customized version for Hong Kong and Macau residents will be launched soon.
If you are a mainland tourist going to Hong Kong and Macao, you can quickly make an appointment for document processing, check the nearby immigration office hall, and apply for a Hong Kong and Macao traffic package through "Bay Affairs OneStop". Based on Tencent's mobility service product capabilities, you can purchase car, train, and ferry tickets to Hong Kong and Macao, and open up cross-city transportation, public transportation, taxi, car rental and other travel services. (Drive House)
Baidu Intelligent Cloud launched the "Super Assistant" plug-in version
Recently, Baidu Intelligent Cloud released a companion Copilot based on Wenxin Yiyan - "Super Assistant". According to reports, as the first open browser plug-in for office scenarios in China, combined with large models, the "super assistant" can assist users to directly translate articles, create copywriting, summarize web pages, and answer questions and answers on various web pages, and collaborate with users in a more personalized way. Currently, the "Super Assistant" Web Copilot (browser plug-in) is now open for public beta. (Enniu data)
Huge Engine has launched a series of intelligent delivery products to drive offline passenger flow with one click
Recently, Mega Engine has launched a variety of intelligent delivery tools, which help local merchants increase customer flow and achieve a double day GMV explosion through intelligent hosting and automatic optimization of the system. For dining, wine and tourism, and leisure and entertainment shopping GMV in-store merchants, you can use the tools of "Chengdanbao", "Global Promotion" and "Store Stable Investment" to achieve one-click store drainage or group purchase transactions; It is understood that this series of delivery tools will be fully launched soon. (Enniu data)
OpenAI 考虑将 GPT-4 Vision 嵌入 Spectacles 智能眼镜
According to The Information, people familiar with the matter revealed that OpenAI recently considered embedding its object recognition software GPT-4 with Vision into the products of Snapchat's parent company. This may bring new features to Snap's Spectacles smart glasses. As companies like Meta, Google, Microsoft, and OpenAI release more powerful AI that can understand images and language, they are racing to apply the technology to smart glasses and other wearables with front-facing cameras. This is a vision that many companies have discussed or worked on for years, but the rise of multimodal AI has opened up new possibilities for them to achieve this. It understands not only text and audio, but also paintings, icons, objects, and gestures. (Love Fan'er)
contact
Bypassing Apple's 30% commission, Spotify reiterates that its iOS app doesn't offer in-app purchases
On December 9, Spotify adjusted the iOS version of Spotify in 2016 to avoid paying Apple a 30% commission fee, removing the option to make in-app purchases. According to MacRumors, a foreign technology media report, in the latest beta version of the Spotify app, relevant codes for in-app purchases were found, including an in-app payment (IAP) checkout screen and a purchase retry panel for failed purchases. Farshad Shadloo, Spotify's Global Head of Corporate and Policy Communications, later issued a statement to the media, emphasizing: "We have no plans to turn on in-app payments (IAPs) at this time." (IT House)
Musk's X was revealed to be working with Amazon
Elon Musk's social platform X (formerly known as Twitter) has been boycotted by a large number of advertisers due to antisemitic controversy, but according to sources, X seems to have found a way to control the losses. It is reported that X is in talks with online shopping giant Amazon to integrate the ads of the X platform into Amazon's ad buying software. In this way, Amazon's small and medium-sized merchants or businesses can purchase X advertising services directly through Amazon and display their products on X. At the moment, the collaboration is still in the early stages of discussion and there is still a possibility of failure. Neither X nor Amazon have commented on the matter. (Finance Associated Press)
The GMV of WeChat Channels e-commerce has exceeded 100 billion, and the WeChat payment team will participate in the construction
According to LatePost, the total turnover (GMV) of WeChat Channels e-commerce this year is around 100 billion yuan. At the same time, Tencent will further expand the team of Channels e-commerce business: in addition to the Channels live broadcast team responsible for front-end products, the WeChat Pay team will also join the construction of Channels e-commerce, participate in the development and operation of various industry categories, influencers, customers, and the construction of basic functions such as transaction products. After the number of daily active users of Channels surpassed Kuaishou in the third quarter of 2022, commercialization was also put on the agenda. In July last year, Channels launched Channels Store and native in-feed ads. According to people familiar with the matter, the commercialization of Channels has also been identified by Tencent as one of the most important group strategies this year. （CSDN）
Agency: After abandoning Platform X, Disney and Comcast increased their ad spend on Instagram
Major U.S. companies such as Walt Disney and Comcast increased their ad spend on Instagram after suspending advertising on social media platform X last month, according to Sensor Tower. Specifically, Disney and Comcast increased U.S. ad spend on Instagram by 40% and nearly 6%, respectively, in the two weeks starting Nov. 20. At the same time, Paramount's ad spend on Snapchat tripled. According to Sensor Tower, 51 of the top 100 U.S. advertisers on the X platform have paused advertising in November since Musk's takeover in October last year. (Interface)
Kuaishou will stop the commercial launch of third-party micro-short drama mini programs on the evening of December 31
According to Kuaishou's official news, in order to guide the healthy development of the paid short drama business and consolidate its own short drama content and industrial ecological construction, Kuaishou will officially stop the commercial promotion and launch of third-party micro-short drama mini programs at 20:00 on December 31. Kuaishou official said that from December 22, third-party mini program micro-short drama advertisers can gradually migrate to Kuaishou mini programs and Kuaishou's self-built links for delivery, during which the Kuaishou platform will give certain policy support in order to reduce the impact of migration on advertisers and ensure the delivery effect. (Enniu data)
Sea
For the first time, Ali International Station held a B2B foreign trade promotion specifically for the Southeast Asian market
Ali International Station will hold a B2B foreign trade promotion in Southeast Asia for the first time. It is understood that the promotion is called the "Southeast Asian New Year Stocking Festival", which will start on December 18 and end on January 29, 2024. It is estimated that more than 60% of the goods sent to Vietnam will be free shipping under the platform subsidy, and more than half of the goods sent to Indonesia will achieve a 5% discount on logistics costs. (Enniu data)
TikTok and Indonesia's GoTo Group have reached an e-commerce strategic cooperation
TikTok officially announced that it has reached an e-commerce strategic cooperation with Indonesia's GoTo Group. TikTok's Indonesian e-commerce business will merge with Tokopedia, an e-commerce platform owned by the GoTo Group, and the combined Tokopedia will be controlled by TikTok. The two parties will work together to promote the development of Indonesia's digital economy and support small and medium-sized enterprises, and the previously banned TikTok e-commerce will also be relaunched on December 12, Indonesia's National Online Shopping Day. (Sina Technology)
Indonesian government: TikTok and GoTo's e-commerce cooperation will be given a trial period of several months, after which it will be evaluated
Indonesian Trade Minister Zulkifli Hassan said that TikTok and Indonesian tech group GoTo will be given a three- to four-month trial period for e-commerce cooperation, "with the main aim of helping sellers so that they can restart their business after TikTok Shop is suspended". Hassan said Indonesia's trade ministry will audit and evaluate the partnership between the two companies after the trial period ends. (Interface)
Alibaba injected an additional $634 million into Lazada
Recently, Lazada, a Southeast Asian e-commerce company, received another $634 million in capital from its parent company, Alibaba Group. It is understood that the funding comes as Alibaba continues to evaluate and adjust its strategy, with Alibaba injecting a total of more than $1.8 billion into Lazada this year. According to statistics, Alibaba injected $1 billion and $2 billion into Lazada in 2017 and 2018, respectively. In 2022, Alibaba injected a total of $1.6 billion into Lazada in three installments, and this year, Alibaba has injected a total of more than $1.8 billion into Lazada. Up to now, Alibaba has injected more than $7.4 billion into Lazada over the years. (21 Finance)
It is reported that Meituan has closely inspected the Middle East market and is still waiting to see how it goes overseas
According to 36 Krypton, from October last year to the middle of this year, Zhu Wenqian, head of the overseas department of Meituan's war investment, visited the Middle East many times. According to people familiar with the matter, in addition to understanding the local business policies, the core of her research is to investigate the competitive status of the food delivery business. In May, Zhu Wenqian, together with Meituan CEO Wang Xing, Meituan Daojia Business Group President Wang Pu and other key executives visited the Middle East again and met with a number of Saudi royal family members and ministers. According to the above-mentioned people familiar with the matter, in the middle of the year, Meituan tried to recruit HRD to form a Middle East team and considered the first pilot city in Riyadh, but this move was then shelved due to "differences of opinion at the top level". The latest news is that Saudi Arabia has announced a 30-year corporate income tax relief for companies that relocate their regional headquarters to Riyadh. An investor who has been in the Middle East for many years said, "This is to grab people and companies from other Middle Eastern countries, especially the United Arab Emirates." Over the past years, Dubai, the capital of the United Arab Emirates, has always been the first choice for Chinese companies to go overseas to the Middle East because of its better infrastructure, higher per capita GDP level and friendlier business environment. (Love Fan'er)
management
Musk threw an olive branch to OpenAI's chief scientist and openly grabbed people from OpenAI
OpenAI's chief scientist and co-founder, Ilya Sutskever, is in an awkward position. Last month, he joined the board of directors in voting to fire CEO Sam Altman, but just five days later, Mr. Altman returned to his role with strong support from employees and investors. A recent report pointed out that Sutskvi seems to be in a state of "confusion" at the moment, becoming "invisible" at OpenAI. According to the report, Sutzkove has not been seen in OpenAI's San Francisco office since Altman's return. While there was a lot of speculation about Sutskove's future direction, Musk publicly stated on social media that Sutskove should join Tesla or his AI company, xAI. (IT House)
Ali e-commerce focuses on AI: Taotian has set up 4 teams and an international business team of more than 100 people
Taotian Group has just finished sorting out its AI business, from about 20 teams to 4, and at the same time released its own large model product "Turing" internally, and the AI team of the international digital business group now exceeds 100 people. In September this year, Alibaba's new CEO Wu Yongming established "AI-driven" as one of the group's two strategic priorities in an internal all-staff letter, and in November, in his first earnings call after taking office, "AI" became the most important keyword for him to describe Alibaba's strategy and various businesses this quarter. (LatePost)
Cai Yinghua, chief commercial officer of Alibaba Cloud, has resigned
Alibaba Cloud Chief Commercial Officer Cai Yinghua (pseudonym: Hua Yan) has officially stepped down from Alibaba Cloud on December 13. Cai Yinghua is a key figure in Alibaba Cloud's promotion of government and enterprise affairs in the past two years. He parachuted into Alibaba Cloud in March 2022 and became Senior Vice President (M7) of Alibaba Group. Cai Yinghua's departure was actually foreshadowed. On November 23, Alibaba Cloud announced the latest organizational structure adjustment, and the sales line structure has been changed to three parts, namely the public cloud business division, the hybrid cloud business division and the overseas business division, all of which report to Alibaba Cloud CEO Wu Yongming, but there is no movement of Cai Yinghua. (36 Krypton)
cooperate
Milestone collaboration: OpenAI will pay for news content from publishing giant Springer
Springer Publishing Group, a news publishing giant, has signed an agreement with OpenAI, the developer of ChatGPT, becoming the first publisher in the world to partner with OpenAI to integrate journalism and AI technology more deeply. This is an important milestone for media outlets seeking financial compensation for the use of their content in AI tools. (The Paper)
capital market
Oracle's adjusted revenue for the fiscal second quarter fell short of expectations
ORACLE'S FISCAL 2ND QUARTER ADJUSTED EARNINGS PER SHARE WERE $1.34 VERSUS ANALYSTS' EXPECTATIONS OF $1.33, FISCAL SECOND QUARTER ADJUSTED REVENUE OF $12.94 BILLION VERSUS ANALYSTS' EXPECTATIONS OF $13.05 BILLION, FISCAL SECOND QUARTER CLOUD REVENUE (IAAS + SAAS) REVENUE OF $4.8 BILLION VERSUS ANALYSTS' EXPECTATIONS OF $4.86 BILLION AND UP 24% AT CONSTANT EXCHANGE RATES, AND CLOUD SERVICES AND LICENSING SUPPORT REVENUE OF $9.64 BILLION IN FISCAL SECOND QUARTER, UP 12% YEAR OVER YEAR. Operating profit for the fiscal second quarter was $5.54 billion versus analysts' expectations of $5.56 billion, and operating margin for the fiscal second quarter was 43% versus analysts' expectations of 42.6%. Oracle's U.S. stock fell more than 10% in after-hours trading as adjusted revenue fell short of expectations in the fiscal second quarter. (Sina Technology)
NetEase surpassed Meituan to become the fourth largest Internet company in China by market capitalization
At the close of Hong Kong stocks on December 13, NetEase surpassed Meituan with a market value of HK$542.5 billion, becoming China's fourth-largest Internet company by market capitalization, behind Tencent, Pinduoduo and Alibaba. In February 2021, when the valuation of Chinese concept stocks as a whole rushed to its peak, Meituan's market value was about four NetEase. However, in more than two years, NetEase's market value has surpassed Meituan. According to a late financial report, the reason why NetEase was able to overtake Meituan is because it fell less than Meituan. From the peak of market capitalization in 2021 to the present, NetEase's market capitalization has fallen by 21%, while Meituan has plummeted by more than 80%. (IT House)
Policy supervision and self-discipline
The EU has reached a preliminary agreement around the approval of the AI Act, and large general-purpose AI models will face more regulation
On December 9, according to the latest news from AFP, after 36 hours of negotiations, EU member states and MEPs reached a preliminary agreement on the world's first comprehensive regulation to regulate artificial intelligence, including ChatGPT. Channel Euronews described the negotiation process as intense, and this is the world's first attempt to regulate this rapidly evolving technology in a comprehensive, ethically based manner. With today's political agreement around the Artificial Intelligence Act, the EU became the first continent to set clear rules for the use of AI," said the European Commission's Commissioner for the Internal Market, Thierry Bretton. Channel Euronews mentioned that earlier negotiations on Thursday (7th) were to reach a preliminary compromise on the basic model to control chatbots like ChatGPT, a large language model of American artificial intelligence company OpenAI, and the negotiations on the 8th focused on the use of real-time biometric technology, including facial recognition, in public places. Bloomberg reports that the agreement marks a key step forward in the EU's AI regulation. In the absence of any action by the United States, the bill could set the tone for regulation of generative AI tools in developed countries. But some warn that even if an agreement is reached, more meetings may be needed to finalize the technical details of the regulations, which will not take effect until 2026 at the earliest. (Sohu Technology)
UK regulators scrutinize Microsoft's partnership with OpenAI, an OpenAI spokesperson responds
On Dec. 9, the UK's Competition and Markets Authority (CMA) said it was gathering information from stakeholders to determine whether Microsoft's partnership with OpenAI threatened competition in the UK, where Google's AI research lab, Deepmind, is headquartered. The CMA said it would investigate whether the balance of power between the two companies has fundamentally shifted, giving one party more control or influence over the other. A spokesperson for the European Commission said that regulators have been "closely monitoring the control of OpenAI". In response, an OpenAI spokesperson said, "Microsoft's non-voting board observer seat does not provide them with management authority or control over OpenAI's operations. (Finance Associated Press)
The Price Department of the National Development and Reform Commission and the National Data Bureau are preparing for three groups to hold an expert symposium on establishing and improving the price formation mechanism for public data
In order to implement the spirit of the "Opinions of the Central Committee of the Communist Party of China and the State Council on Building a Data Basic System to Better Play the Role of Data Elements", recently, the Price Department of the National Development and Reform Commission, together with the National Data Bureau, prepared three groups to hold a symposium to listen to the opinions and suggestions of experts in relevant fields on accelerating the establishment and improvement of a price formation mechanism that conforms to the characteristics of public data elements, and promoting the paid use of public data for digital development according to government guidance. Nine experts from Peking University, Tsinghua University, Chinese People's University, Fudan University, Chinese Academy of Fiscal Sciences, Macroeconomic Research Institute, State Information Center, Price Monitoring Center and other units attended the meeting and delivered speeches, and the participants carried out in-depth exchanges and discussions. In the next step, the National Development and Reform Commission and the National Data Bureau will work with relevant parties to accelerate research and establish a price formation mechanism for public data products and services, and promote the development and utilization of public data resources. (Securities Times)
Ministry of Industry and Information Technology: Moderately advance the construction of new infrastructure
According to the Ministry of Industry and Information Technology, the Ministry of Industry and Information Technology held a press conference on "Giving Play to the Role of National High-tech Zones and Accelerating New Industrialization" on the 12th. Li Youping, deputy director of the Torch Center of the Ministry of Industry and Information Technology, said that in the next step, the national high-tech zone will focus on digital technology innovation and accelerate the pace of digital construction and governance of the park. The first is to strengthen the innovative application of artificial intelligence. Promote national high-tech zones to strengthen the supply of key core technologies and common basic technologies for artificial intelligence, and increase the opening of application scenarios in key areas. The second is to build digital parks in depth. Accelerate the digital and intelligent transformation of traditional infrastructure in parks, build new infrastructure in advance, and improve the digital governance capabilities of parks. The third is to vigorously promote the digital transformation of the manufacturing industry. Vigorously promote the large-scale application of the industrial Internet, deepen the innovative application of "5G + Industrial Internet", and cultivate "digital pilot" enterprises and small and medium-sized enterprises as benchmarks for digital transformation. (SSE News)
Epic Games has a historic victory, and the Google Play Store has been ruled monopolies
On December 12, it was reported that after three years of legal disputes, Fortnite developer Epic Games finally achieved a historic victory in the antitrust lawsuit with Google. The jury has just unanimously ruled that Google has a monopoly on the Google Play app store and in-app payment service Google Play Billing marketplace and has abused that position to the detriment of Epic Games. This is a historic victory, as Epic's similar lawsuit against Apple two years ago ended in failure. At that time, Judge Yvonne Gonzalez Rogers ruled that the lawsuit had nothing to do with the application. (IT House)
Central Economic Work Conference: Accelerate the development of artificial intelligence
The Central Economic Work Conference was held in Beijing on December 11-12. The meeting emphasized that next year, we should focus on promoting high-quality development, highlight the key points, grasp the key, and do a solid job in economic work. It is necessary to promote industrial innovation with scientific and technological innovation, especially to use disruptive technologies and cutting-edge technologies to give birth to new industries, new models, and new kinetic energy, and develop new quality productive forces. We will improve the new national system, implement high-quality development actions for key industrial chains in the manufacturing industry, strengthen quality support and standard guidance, and improve the resilience and safety of industrial and supply chains. It is necessary to vigorously promote new industrialization, develop the digital economy, and accelerate the development of artificial intelligence. The meeting emphasized that it is necessary to build a number of strategic emerging industries such as biomanufacturing, commercial aerospace, and low-altitude economy, open up new tracks for future industries such as quantum and life sciences, widely apply digital intelligence technology and green technology, and accelerate the transformation and upgrading of traditional industries. Strengthen applied basic research and cutting-edge research, and strengthen the main position of enterprises in scientific and technological innovation. Encourage the development of venture capital and equity investment. (Xinhua News Agency)
Amazon won the lawsuit and is expected to get rid of the 250 million euro tax case completely
On December 14, it was reported that Amazon won an important legal case dispute in the Supreme Court of the European Union, and the technology giant previously faced a potential disadvantage due to the illegal tax cut of 250 million euros (about 272 million US dollars) case. On Thursday, the EU's highest court rejected the European Commission's key appeal in a binding ruling, saying that the EU regulator "did not determine" that the tax arrangement between Amazon and Luxembourg was "assistance of a state nature incompatible with the internal market". (i Dark Horse)
Temu restarted its lawsuit against Shein in the United States, close to Temu-related people: Shein's recent behavior is too exaggerated
According to Jiemian News, on December 14, in response to Temu's restart of the lawsuit against Shein in the United States, a person close to Temu revealed that Temu's move was because Shein's recent behavior was too exaggerated, and Temu had no choice but to countersue. "Shein began to illegally retain merchants, forcibly confiscate merchants' mobile phones, obtain Temu's merchant accounts from merchants' mobile phones, steal business information, and threaten merchants not to do business in Temu. In addition, Shein also made false representations to copyright registries, and launched tens of thousands of false and malicious complaints in the United States that were not supported by real copyrights, and Shein alone had 1.7 times the number of complaints from all brands in the world combined. At the same time, Shein also abused the judicial process and used the vest he controlled to launch a batch of lawsuits against Temu overseas. (i Dark Horse)
Seven departments: support artificial intelligence enterprises to develop large models of audio-visual applications
Seven departments, including the Ministry of Industry and Information Technology, issued the "Guiding Opinions on Accelerating the High-quality Development of the Audiovisual Electronics Industry". Among them, it is proposed to support leading color TV enterprises to enrich the product matrix, improve the industrial chain, explore overseas markets, continue to enhance ecological dominance, and lead the development of the industry. Encourage speakers, headphones, and microphone OEMs to develop their own brands and enhance the added value of products and industry influence. Accelerate the cultivation of specialized and new "little giants" in the fields of commercial display, vehicle audio-visual, and audio and video, as well as individual champions in the manufacturing industry. Support backbone enterprises to become bigger and stronger, and support artificial intelligence enterprises to develop large models of audio-visual applications. (Securities Times)