Today's Financing Bulletin
Groq, an artificial intelligence chip startup, raised $640 million in funding to challenge Nvidia
Groq, a startup that develops chips to run generative AI models faster than traditional processors, said Monday that it had raised $640 million in a new funding round led by Blackrock. Neuberger Berman, Type One Ventures, Cisco, KDDI, and Samsung Catalyst Fund are also involved.
The funding brings Groq's total funding to more than $1 billion and the company's valuation to $2.8 billion, which is a major win for Groq, which was initially reportedly looking to raise $300 million at a slightly lower valuation ($2.5 billion).
Yann LeCun, chief AI scientist at Meta, will serve as technical advisor to Groq, and Stuart Pann, former head of Intel's foundry business and former CIO of Hewlett-Packard, will join the startup as COO.
Youdi Robot was changed to a joint-stock company and increased its capital to 350 million
Youdi Robot (Wuxi) Co., Ltd. underwent industrial and commercial changes, the type of market entity was changed from a limited liability company (Hong Kong, Macao and Taiwan investment, non-sole proprietorship) to a limited liability company (Hong Kong, Macao and Taiwan investment, unlisted), the name was changed to Youdi Robot (Wuxi) Co., Ltd., and the registered capital increased from about 17.079 million yuan to 350 million yuan. According to the official website, Youdi Technology focuses on the applied research and development and commercialization of the core technology of delivery robots.
ROBOTICS STARTUP DELIVERS. AI is valued at $36 million after a new round of funding
DELIVERS. AI's autonomous mobility platform uses advanced, AI-driven, low-emission road delivery robots and vehicles to make last-mile logistics affordable and sustainable.
DELIVERS. AI was valued at $36 million and raised a new round of funding, the exact amount of which was not disclosed. The Warwick, United Kingdom-based company received funding from Japan Post Capital, the Turkey Development Fund, Bulgaria's Push Capital and Istanbul Technical University. Previous investors include Driventure, Arz Portföy, StartupFON, Plug and Play Ventures, Inveo Ventures, StartersHUB, and Kalyon Ventures.
AI-powered jewelry company Stepin received 10 million yuan in angel round financing
Stepin is a cross-border jewelry brand, focusing on AI jewelry cross-border e-commerce, with the help of AI technology and China's supply chain advantages, to achieve a rapid new and high turnover model of "small orders and quick returns". Stepin completed an angel round of financing of 10 million yuan, led by Xinyue Capital and 99 Capital, with the participation of other industrial institutions and angel investors, and Inspur Capital as the exclusive financial advisor. At present, the application of AI within Stepin is mainly to generalize ideas and generate relatively rudimentary 3D models for designers to fine-tune. (36 Krypton)
(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
传 ResNet 作者张祥雨加盟阶跃星辰
Qubit news, post-90s AI bull Zhang Xiangyu joined StepLeap Star, he is one of the four authors of ResNet, Sun Jian's first deep learning doctoral student, and is the winner of the Future Science Award. Xiangyu Zhang received his bachelor's degree in software engineering from Xi'an Jiaotong University and completed ResNet with Jian Sun, Kaiming He, and Shaoqing Ren at Microsoft Research Asia, which won the CVPR Best Paper Award in 2016 and the Mathematics and Computer Science Prize of the Future Science Prize in 2023.
In addition to Zhang Xiangyu, Stepleap Star also recruited two other heroes, Tencent Yu Gang and MSRA Duan Nan. Yu Gang is an undergraduate student from Shandong University, a master's degree from Shanghai Jiao Tong University, and a Ph.D. from Nanyang Technological University. Nan Duan is a senior principal investigator at MSRA, mainly engaged in natural language processing and other research, and has worked at MSRA for 17 years and 9 months, and his research results have been applied to a number of Microsoft AI products.
OpenAI admits it is working on ChatGPT text watermarking, but faces challenges
OpenAI has developed a tool that can recognize ChatGPT-generated text with high accuracy, but it has not been released for a long time. In response, OpenAI acknowledged that it was working on text watermarking technology, but said that there are still many challenges to this technology.
OpenAI envisions weaving an invisible "digital fingerprint" — a text watermark — — between the lines by subtly tweaking the word choices in ChatGPT-generated text. The beauty of this design is that in the future, specific tools will be able to easily identify and verify the original source of text, opening up new avenues for copyright protection and content traceability. Text watermarking is only one part of OpenAI's diverse solution matrix. They also work on classifier technology and metadata strategies in parallel, aiming to build a comprehensive and multi-level text identity authentication system to ensure that the source of information is clear and traceable.
Figure previews the second generation of new humanoid robots, with stronger human flavor and stronger hardware
Figure released the trailer for Figure 02 and said that it will officially release the product on August 7, Beijing time. Compared with the video demonstration of Figure 01 equipped with Open AI GPT4, the focus of this exhibition is on hardware, and it is expected that the hardware capabilities will be greatly improved. Founder CEO Rett Adcock confidently stated: Figure 02 is the best humanoid robot on the planet.
Google Gemini API 大降价 比 GPT-4o mini 价格便宜一半
The Gemini 1.5 Flash model halved the input cost by about 85 percent, followed by a significant reduction of about 80 percent. This means that the cost of using the Gemini API is now almost 50% lower than that of its main competitor, GPT-4o mini. Gemini Flash's new pricing costs just $0.075 per million tokens for input and $0.3 for output. Gemini 1.5 Flash and Gemini 1.5 Pro now support over 100 languages, and Google has also introduced innovations like contextual caching and a batch API.
Baichuan Intelligence and Chinese Renmin University established a "large model joint laboratory"
Chinese Renmin University and Baichuan Intelligent jointly established the "Large Model Joint Laboratory" to promote the innovation and development of large model technology. The establishment of the joint laboratory marks that the two parties will carry out in-depth cooperation in cutting-edge technologies such as large model pre-training, alignment, retrieval enhancement, agents, and multimodality. Chinese University will use its talent and technical advantages in large-scale model research to combine with Baichuan Intelligence's strength in engineering and product research and development to jointly promote the research and application of related technologies.
Ali launched Tora, the "trajectory controllable version of Sora" to make video generation more in line with the laws of physics
Tora is the first trajectory-oriented DiT architecture that integrates textual, visual, and trajectory conditions simultaneously to generate video. Tora is designed to fit seamlessly with DiT's scalability, allowing for precise control of video content with different durations, aspect ratios, and resolutions. Numerous experiments have proven that Tora excels at achieving high motion fidelity while meticulously simulating motion in the physical world.
Meta is said to be in talks with Hollywood movie stars about AI voice projects, and will provide millions of dollars to obtain licensing
According to media outlets, citing sources, negotiations between Meta and some of the actor's representatives have undergone multiple interruptions and restarts, as the two parties have been unable to agree on the terms of use of the actor's voice. Meta is accelerating negotiations to allow enough time to develop AI tools, which are expected to be released at the Connect conference in September. It is unclear how Meta will use these voices, most likely acting as digital assistants. For example, users can chat with chatbots that have an awkwafina voice.
Today's product news
Product Hunt 热榜，Avatar Architect
Avatar Architect is a blend of artificial intelligence and notion designed to help entrepreneurs, marketers, and product developers improve their marketing strategies and sales performance by gaining a deep understanding of their target customer segments, while providing a range of tools and guides to build and manage customer personas.
Avatar Architect's strengths lie in its AI-driven efficiency, in-depth insights into the target market, and detailed customer information to guide product development. The system is suitable for solopreneurs, marketers, and product developers, especially those who want to better understand their customers and improve their marketing strategies. Users can customize the functionality of the system according to their business needs, and customer data should be updated regularly for best results.
? https://gcproductivity.gumroad.com/l/avatararchitect/ProductHunt?ref=producthunt
Recommended by the developer
1. Simple prompts to easily create LlamaCoder Mini Program code
LlamaCoder is a platform based on the Llama3.1405B model, which enables developers to quickly generate complete React applications and components with simple instructions through powerful automation capabilities. The platform features a modern technology stack, including popular technologies like Next.js and Tailwind, to provide an interface that is both beautiful and easy to use. LlamaCoder's features are designed with comprehensiveness in mind, including code sandboxing, Helicone integration, and the use of Plausible tools to improve development efficiency and product optimization. The barrier to entry is low, and users can start the project with the npm command by simply cloning the codebase and setting the API key. It is currently open source.
? https://llamacoder.together.ai/
2. Supermemory personal knowledge base project
The project allows users to save content such as web pages, tweets, and notes, and use its built-in artificial intelligence capabilities to efficiently search and ask questions. Organize information in the form of a two-dimensional canvas to help users better understand and relate knowledge points. AI-assisted writing based on saved data. Supports integrations with platforms such as Telegram, Twitter, and more.
github?https://github.com/supermemoryai/supermemory网站?https://supermemory.ai/onboarding
Special attention
Jim Fan: Amplification of robot data is a key problem in solving the development of robotics
Jim Fan, Senior Research Scientist, Head of Embodied Intelligence, and Head of Embodied AI (GEAR Lab) at NVIDIA, released an update on Project GR00T, proposing a systematic approach to scaling up robot datasets to solve data bottlenecks in robotics by 1,000x or more through human demonstrations on real robots, combined with simulation techniques.
Apple Vision Pro technology gives a human-operator first-person control of a humanoid robot. Vision Pro's ability to interpret human hand posture in real-time and map it onto the robot's hand makes the operator feel like they are immersed in another body, collecting small amounts of high-quality data despite slower remote operations.
Use RoboCasa, an open-source generative simulation framework, to scale a single piece of real-world presentation data to a wide variety of environments by changing the visual appearance and layout of the environment, resulting in a large amount of diverse presentation data. This allows the data of a physical kitchen to be extended to an infinite number of simulated kitchen scenarios.
MimicGen technology is applied to generate a large number of new motion trajectories based on raw human demo data and filter out failed attempts to form a larger, richer data set.
With this approach, starting with a human trajectory, RoboCasa can generate data for N different visuals, and MimicGen further enhances data for NxM different actions. This approach solves the problem of expensive human data collection through GPU-accelerated simulations at the cost of computing power, breaking down the traditional barriers to data collection confined to the atomic world.
?https://x.com/DrJimFan/status/1818302152982343983
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
![OpenAI Launches ChatGPT's New Interface, Canvas, More Efficient Writing and Coding, A New Era of AI Collaboration![fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)