开源路线
Search documents
印度AI峰会献媚美国,用中国机器狗冒充自研,科技展成了站队会
Sou Hu Cai Jing· 2026-02-22 21:10
印度的这场AI峰会,翻车了。原本一个应该展示科技创新的舞台,硬生生被办成了一个政治站队秀,场面尴尬而失控。莫迪政府口口声声说要独立发展 AI,然而背后却是紧紧拥抱美国,显然并未考虑与中国展开任何合作。中国的AI发展路线,强调开源、低成本与普惠,恰恰最适合像印度这样的发展中国 家。单拿DeepSeek那款开源模型来说,印度如果采用,将大幅降低成本,甚至能解决因电网老化而频繁崩溃的问题,尤其是数据中心的绿色能源技术,对 印度来说,无疑是一个理想的解决方案。但即便如此,印度政府选择继续与美国巨头合作,甘愿付出巨额代价,让谷歌(150亿美元)、微软(175亿美元) 和亚马逊(350亿美元)来为印度建设数据中心。 在技术方面,印度并不傻,心里清楚什么是好的。就拿一个典型的例子来说,某些参展的印度大学,竟然把中国宇树科技的机器狗贴上自家研发的标签,公 开展示。明明知道中国的技术好,为什么不与中国合作呢?原因很简单,印度心里明白自己无法独立发展AI,只能站队。站对了队,科技的好坏就不再重 要,立场成了最重要的事。莫迪政府的算盘打得很好,但显然计算错了。 印度的玩法似乎挺精明。它意识到,自己根本无法在基础模型研发上与中美抗衡 ...
LeCun创业首轮估值247亿!Alexandre当CEO
量子位· 2025-12-19 01:01
这家名为 Advanced Machine Intelligence Labs (AMI Labs)的新公司,计划于明年一月正式亮相,目标估值 30亿欧元 (约247亿人 民币)。 克雷西 发自 凹非寺 量子位 | 公众号 QbitAI LeCun在Meta的Last Day还没来,新公司又被曝出更多细节。 前脚LeCun本人在播客当中宣布了新公司名称,现在融资和估值目标就被《金融时报》曝光了。 AMI Labs的研究方向,就是LeCun一直主推的"世界模型",而且将走开源路线,老东家Meta也将与其保持合作。 另外,曝料也透露了AMI Labs的 CEO人选并非LeCun本人 ,而是他的一位老部下。 LeCun不当CEO 新公司AMI Labs定于2026年1月在巴黎正式启动,在Meta逐渐转向封闭生态的背景下,LeCun选择了他在学术界一贯坚持的开源路线。 而且在技术层面,AMI Labs选择了比主流的LLM更具挑战性的道路—— 死磕"世界模型" 。 因为在LeCun看来,基于自回归机制的LLM存在根本性的逻辑缺陷,它们只是在统计概率上预测下一个字符,并不真正理解物理世界的运行规 律。 为此,新公司将通过 ...
AI圈大变天!阿里千问硬刚ChatGPT,开源免费掀翻行业规则
Sou Hu Cai Jing· 2025-11-21 08:48
Core Viewpoint - Alibaba's Qianwen app has launched its public beta, directly competing with ChatGPT by adopting an open-source and free strategy, aiming to disrupt the global AI market [1][4][19] Group 1: Product Launch and Strategy - On November 17, Alibaba announced the public beta of the Qianwen app, targeting ChatGPT with a product based on the high-performance open-source model Qwen3, emphasizing a free strategy [4] - Qianwen integrates Alibaba's ecosystem, allowing users to compare prices, plan trips, and generate meeting minutes, transforming AI from a chat tool to a practical assistant [4] Group 2: Long-term Investment and Development - Alibaba has invested 380 billion yuan in AI infrastructure and has been iterating on the Qwen open-source model, focusing on building a solid foundation before launching [7] - The Qwen3-Max model has surpassed GPT-4 and ClaudeOpus, ranking among the top three globally, while the open-source Qwen series has seen significant adoption, with downloads surpassing Llama [9] Group 3: Competitive Landscape and Market Dynamics - A Stanford report indicates that the performance gap between top AI models in China and the U.S. has narrowed to just 0.3%, highlighting the competitive advancements made by Alibaba [11] - The competition is framed as a battle between two AI philosophies: ChatGPT's closed-source, paid model versus Qianwen's open-source, free approach, which aims to democratize AI access [13][15] Group 4: Industry Implications and Future Outlook - The emergence of Qianwen signifies a shift towards a more inclusive and open AI future, challenging the dominance of closed-source models that have historically defined the industry [19] - As more companies join the open-source movement, the AI industry is expected to return to its essence, allowing everyone to benefit from technological advancements [19]