In the field of robotics, the American startup Figure AI recently released a large edge model named Helix, marking a significant breakthrough in humanoid robot control technology. Helix is the first visual-language-action (VLA) model capable of high-frequency, continuous control over the entire upper body of a humanoid robot, including the head, torso, wrists, and fingers. This means that the robot can perform tasks directly based on natural language instructions without extensive training.

In today's fast-paced information age, the efficiency of team communication is particularly important. HiveNexus has launched HiveChat, an AI chat application designed specifically for small to medium-sized teams, aimed at helping teams improve collaboration efficiency and easily meet daily communication needs. The application supports various AI models such as Deepseek, OpenAI, Claude, and Gemini, providing users with a more intelligent and personalized chat experience.

Recently, the Evo2 bio AI model, jointly developed by Arc Institute and Nvidia, was officially released. This foundational model is based on DNA data from over 100,000 species and aims to deeply decode various complex phenomena in biology. Evo2 can identify patterns in the genetic sequences of different organisms that researchers would typically take years to discover, significantly enhancing the ability to identify disease-related mutations, and can design entirely new genomes comparable to simple bacteria.

Recently, the Arc Institute partnered with NVIDIA, along with researchers from Stanford University, the University of California, Berkeley, and the University of California, San Francisco, to jointly release the world's largest biological artificial intelligence model — Evo2. This model is based on data from over 128,000 genomes, training on 93 trillion nucleotides, making its scale comparable to the most powerful generative AI language models. Evo2's deep learning capabilities enable it to rapidly identify sequences of genes across different organisms.

AI security company Adversa AI has released a shocking report stating that Elon Musk's startup xAI's newly launched Grok3 model has significant vulnerabilities in terms of cybersecurity. Adversa's research team found that this latest AI model is susceptible to 'simple jailbreak attacks', potentially allowing malicious actors to access sensitive information such as 'how to groom children, handle corpses, extract DMT, and make bombs'. Worse still, Adversa

Recently, the Microsoft Research team, in collaboration with researchers from several universities, released a multimodal AI model called 'Magma'. This model is designed to process and integrate various types of data, including images, text, and videos, to perform complex tasks in both digital and physical environments. As technology continues to advance, multimodal AI agents are being widely applied in fields such as robotics, virtual assistants, and user interface automation. Previous AI systems typically focused on visual-language understanding or robotic operations, making it difficult to combine the two.

Tencent has announced that its deep thinking model 'Hunyuan T1' is now officially open to all users, who can experience and test the model through the Tencent Yuanbao platform. Following its gray testing phase, Hunyuan T1 received widespread attention and positive feedback from users. In response to user demand, Tencent has rapidly deployed resources to fully open Hunyuan T1, allowing unlimited access for all users.

In a recent statement, OpenAI officially updated its model guidelines, emphasizing that its AI models will no longer shy away from sensitive topics nor dismiss certain viewpoints. This high-level rule aims to guide OpenAI's models' behavior and ensure they remain open when handling various topics. OpenAI clearly stated in the new guidelines: "Our models must never attempt to guide users towards pursuing their own agendas, either directly or indirectly." This means that the models should remain neutral in fields such as politics, society, and culture.

Recently, OpenAI announced the launch of its significantly expanded model specifications document, which outlines the behavioral standards for AI models. The new specifications have grown from approximately 10 pages to 63 pages, covering how AI should handle controversial topics and various aspects of user customization. OpenAI's goal is to enable users and AI models to 'pursue truth together.' Note: Image generated by AI, image licensed by service provider Midjourney. The specifications emphasize three main principles: customizability,

According to media reports, Baidu is set to launch its next generation AI model Ernie 5.0 this year. Sources state that the Ernie 5.0, referred to as a 'foundation model', will see significant enhancements in multimodal capabilities, though specific functionalities were not detailed. This news comes as Apple shifts its potential clients towards Alibaba, leading to widespread speculation that Baidu is seeking to stabilize its stock price and market position amid changing circumstances.

Within just six days, participants successfully bypassed all security measures of Anthropic's AI model, Claude3.5, sparking new discussions in the field of AI security. Jan Leike, a former member of OpenAI's alignment team now at Anthropic, announced on the X platform that one participant managed to breach all eight security levels. This collective effort involved approximately 3,700 hours of testing and 300,000 messages from participants. Despite the challengers

According to the Qichacha APP, on February 11, a patent titled 'An Artificial Intelligence Communication Method, Device, and Storage Medium' was published, which was applied for by Beijing Xiaomi Mobile Software Co., Ltd. The patent summary indicates that this disclosure pertains to an artificial intelligence communication method, device, and storage medium. The AI communication method includes determining a first time, which represents the processing time for AI tasks based on the AI model; and performing AI communication processing based on the AI model according to the first time. This disclosure achieves AI model-based communication through the determination of the first time.

Recently, cybersecurity researchers discovered two malicious machine learning models that were uploaded quietly to the renowned machine learning platform HuggingFace. These models utilized a novel technique, successfully bypassing security detection through 'corrupted' pickle files, raising concerns. Karlo Zanki, a researcher at ReversingLabs, pointed out that the beginning of the pickle files extracted from these PyTorch format archives suggests that

With the rise of the domestic AI model DeepSeek, a group of speculators are profiting from users' curiosity about AI. Investigations reveal that some merchants are making substantial profits by selling free DeepSeek installation packages and low-priced 'monetization courses'. On an e-commerce platform, the prices for DeepSeek installation packages range from 9.9 to 19.9 yuan. One store has sold over 200,000 copies, which, even at the lowest price of 9.9 yuan, translates to over 2 million yuan in revenue. However, reporters found that these paid contents are actually...