Today's Financing Bulletin
AI startup Cohere raises $500 million in a bid to compete with OpenAI
AI developer Cohere has raised $500 million in a new funding round, making it one of the world's most valuable startups in the space, which also strengthens the Canada company's ability to compete with competitors like OpenAI and Anthropic.
The funding round brings Cohere's valuation to $5.5 billion. Founded in 2019 by former Google researchers, the company is now competing fiercely with bigger rivals for lucrative contracts with companies vying to incorporate AI into their businesses.
Oak Deer Technology, a stir-fry robot company, received nearly 200 million yuan of strategic investment from JD.com
Oak Egret Technology is a stir-fry robot company, and on December 14, 2023, Oak Egret Technology received tens of millions of yuan in financing from Jingdong Group. Recently, it has received nearly 200 million yuan of strategic investment from JD.com and reached a comprehensive strategic cooperation, and the two will jointly promote the high-quality development of the "robot +" industry.
The world's leading vision-based AI stir-fry robot built by Oak Deer Technology will be officially launched in 2025.
AI cloud platform Nscale acquires AI data center solution Kontena
Nscale, a fully vertically integrated AI cloud platform, announced the acquisition of Kontena to strengthen its capabilities in high-performance computing and AI infrastructure, as well as provide faster and more cost-effective deployment options. Kontena is a leader in high-density modular data center and AI data center solutions. The acquisition aims to accelerate the development and deployment of cutting-edge generative AI data center solutions.
(Welcome to add WeChat AIyanxishe2 to learn more about AIGC, financing, and chat with like-minded friends about the latest AI products)
Today's big factory news
Yuan Jinhui's company added two new co-creators of silicon-based flow, and was invested by Hubble Zhipu AI and others
AI Technology Review exclusively learned that two new co-founders, Zhang Shi and Zeng Hua, have recently added two co-founders of Silicon-based Mobility, who are responsible for the commercialization strategy, business operation and management of Silicon-based Mobility, both of whom have rich IT and management experience. Silicon-based mobility was founded by Yuan Jinhui, whose previous venture, OneFlow, focused on technology research and development and open source ecology, while silicon-based mobility began to seek commercialization from the beginning of its establishment.
In addition, Beijing Silicon Technology Co., Ltd. underwent industrial and commercial changes, the corporate name was changed to Beijing Silicon-based Mobile Technology Co., Ltd., and the new shareholders were Shenzhen Hubble Technology Investment Partnership (Limited Partnership) under Huawei, Beijing Zhipu Huazhang Technology Co., Ltd., an affiliate of Zhipu AI, and Beijing Qihoo Technology Co., Ltd., a subsidiary of 360, etc., and the registered capital of the company increased from about 1.239 million yuan to about 1.376 million yuan. Recently, Silicon Technology announced the completion of the Pre-A round of financing, which was jointly participated by 360, Zhipu AI, Shuimu Tsinghua Alumni Fund, Yaotu Capital, etc.
Nvidia is reportedly preparing a flagship version of the AI chip for the Chinese market
Nvidia is developing a new flagship artificial intelligence chip for the Chinese market that complies with United States' current export controls, the sources said. Nvidia released the "Blackwell" chip family in March this year, which will be mass-produced later this year. In the series, the B200 is up to 30 times faster than its predecessor on certain tasks, such as providing answers from a chatbot. Nvidia will work with its Chinese distributor partners to launch and distribute the chip, tentatively named "B20," the sources said. (Interface)
Shanghai AI Lab open-sourced GRUtopia to simulate an interactive 3D world
GRUtopia is an open-source project called "Taoyuan" open-sourced by Shanghai AI Lab, which provides a comprehensive simulation environment and benchmarks for embodied AI, including 100,000 interactive, annotated scenes covering 89 scene categories, and an LLM-driven NPC system to simulate social scenarios and assign tasks for embodied AI. A virtual world of robots has been created to accelerate the robot's ability to learn and adapt in complex environments.
?https://github.com/openrobotlab/grutopia
Baidu Intelligent Cloud Launches Financial Services Intelligent Twins Application "Zhijin"
Baidu Intelligent Cloud released the intelligent twins application of large models in the financial industry - Baidu Intelligent Cloud Intelligent Gold, and launched four intelligent twins application products for subdivided scenarios: "Asset Intelligent Evaluation", "Transaction Intelligent Inspection", "Compliance Intelligent Judgment" and "Wealth Intelligent Advisory". (36 Krypton)
腾讯云推出TencentOS Server V3,大模型训练效率 2 倍提升
Tencent Cloud has released TencentOS Server V3, a new domestic server operating system. After 14 years of technology accumulation, this operating system has become the largest Linux operating system deployed in China, with a deployment scale of nearly 10 million sets.
In the era of large AI models, TencentOS Server V3 provides qGPU components, which can virtualize and divide a GPU card into multiple cards, greatly improving resource utilization. At the same time, the operating system also integrates large model training and inference acceleration frameworks, achieving a 2x increase in efficiency. In addition, TencentOS Server uses offline hybrid technology, which improves server deployment density and CPU utilization, and reduces operating costs. (AI Smart Push Star)
Meta launches the JASCO music generation model, which uses conditional inputs such as chords or beats to improve the controllability of music generation
Meta introduces the JASCO Music Generation Model, a text-to-music generation model that generates music based on global text descriptions and fine-grained local control. JASCO is able to extract information relevant to specific controls through information bottleneck layers and temporal ambiguity techniques, allowing for the integration of both symbolic and audio-based conditions in the same text-to-music model. The model is characterized by its ability to improve the controllability of music generation by inputting conditional information such as chords or beats.
?https://arxiv.org/abs/2406.10970
Arcee AI发布开源语言模型Arcee-Nova:基于Qwen2-72B,性能接近GPT-4
Based on the Qwen2-72B-Instruct model and another custom-tuned model, Arcee-Nova is widely used in a variety of fields such as customer service, content creation, software development, and education. For example, it can be used to improve customer service; Ability to generate high-quality marketing materials in terms of content creation; In software development, code generation and quality checks can be helped; In terms of data analysis, business insights can be provided; In the field of education, it can be used in personalized learning systems. It is reported that the performance of Arcee-Nova has almost reached the level of GPT-4. (AIbase Base)
NexusflowX 开放权重聊天模型 Athene-70B
该模型是基于 Meta AI 的 Llama-3-70B 进行微调,Athene-70B 在 Arena-Hard 评测中取得了 77.8% 的得分,这一成绩接近了一些顶级的专有模型,如 GPT-4o 和 Claude-3.5-Sonnet。 HuggingFace?https://huggingface.co/Nexusflow/Athene-70BDiscord?https://discord.com/invite/rtaKdn3ghr
The AI video repair speed has been increased tenfold, and Meitu UCAS has a new algorithm
Researchers from Meitu Imaging Research Institute, the University of Chinese Academy of Sciences, and Sichuan University have proposed a new AI video restoration algorithm, BlazeBVD. The algorithm is able to automatically remove flickering from videos and processes up to 10 times faster than existing methods. BlazeBVD is suitable for a wide range of videos without knowing the type or degree of video flickering beforehand. Experimental results show that BlazeBVD scores high on PSNR and SSIM metrics, and low scores on Ewarp, delivering excellent qualitative and quantitative results and inference 10 times faster than state-of-the-art models. The paper has been accepted by ECCV 2024 and published on arXiv.
?https://arxiv.org/abs/2403.06243v1
Today's product news
Product Hunt 热榜,AI Memory,记住一切的个人知识库
Developed by Flot AI, AI Memory is a tool that helps users build a personal knowledge base seamlessly. Users can highlight text, links, screenshots, and files, save information to AI memory, and recall it anytime, anywhere with a simple shortcut to the AI. With the goal of solving the problem of forgetting important details in modern fast-paced life, AI Memory provides a comprehensive information management platform that enables users to easily save and deeply engage with their saved data. The tool supports cross-platform use and is available for both Windows and Mac systems.
? https://flot.ai/AI-memory?ref=producthunt
Recommended by the developer
1. Signals: Efficient analysis and intelligent filtering of large amounts of call data to improve product development efficiency
BuildBetter.ai's Signals tool analyzes thousands of calls in seconds with astonishing accuracy, and can quickly find call clips on specific topics, such as user research, sales, onboarding, and support calls, through smart filters just like you would with quantitative data.
Signals helps product teams save time and is able to extract critical information that helps drive the product forward, such as feature requests, bug reports, and user feedback. In addition, Signals is capable of in-depth analysis based on multiple dimensions such as topic, severity, role, emotion, and impact, and can even create usable documents or user stories. Users can get all relevant customer calls, support messages, and survey content by asking specific questions, such as "billing errors in the past month." Signals also supports integration into all of your organization's communication channels, including phone, email, team meetings, and slack channels
?https://buildbetter.ai/
2. StockBot: A chatbot powered by Lama3-70BStockBot is a real-time stock data and news chatbot powered by Lama3-70B, running on Groq that quickly displays query results in the form of a UI through Vercel's generative UI technology. Users can view daily market performance heatmaps, stock financial data, price history, candlestick charts, headlines, stock screeners, and more through StockBot. In addition, StockBot not only supports the analysis of the stock market, but also covers the forex, bond, and cryptocurrency markets, providing comprehensive market analysis.
GitHub? https://github.com/bklieger-groq/stockbot-on-groq?https://groq-stockbot.vercel.app/
Special attention
MIT professors predict that only one-quarter of AI-related tasks will be cost-effectively automated in the next decade.
Daron Acemoglu, a professor at the MIT Research Institute, said in a Goldman Sachs podcast that only 25% of AI-related tasks will be cost-effective and automated in the next decade. Acemoglu believes that even if AI makes a major breakthrough, its impact will take time to materialize. Acemoglu expressed skepticism about the current architecture of large language models, arguing that these models could run into serious limitations, and was unsure whether increasing GPU capacity would accelerate AI to reach its goals. He stressed that higher quality data will be needed in the future, but it is not yet clear how such data will be obtained.
Jim Covello, global head of equity research at Goldman Sachs, noted that in order for investments in AI to pay off in the AI space at companies such as Nvidia, Microsoft, Google, Meta, Amazon, Supermicro, etc., AI must solve complex problems. Covello argues that the problems currently solved by AI are not complex, that there is a lack of cognitive reasoning, and that there is a misunderstanding of the role of AI.
Stay tuned for updates tomorrow!
The AI Intelligence Bureau is recruiting intelligence partners to gather exclusive value clues! If you can provide information about AI's latest financing & industry insider & unique products, please add the operation WeChat ID: AIyanxishe2 to note the industry position.
Leifeng.com