当人工智能的浪潮席卷全球，北京正以科技创新之姿，成为AI大模型领域的战略高地。从智源研究院的"悟道"大模型问世，到"天使投资人"模式孵化顶尖 学者，再到月之暗面、DeepSeek、智谱等人工智能独角兽崛起，这座城市不仅汇聚了前沿技术，更以开放生态孕育突破性成果。 如今，北京正积极打造"全球开源之都"，一大批研发机构、企业积极拥抱开源，而开源也已深入到汽车、机器人等众多行业。发展AI将是一场科技长征， 新京报AI研究院将深度访谈此次AI浪潮的亲历者与见证人，讲述AI竞争新格局与背后的故事。 开栏语 清程极智联合创始人师天麾。受访者供图 人工智能是年轻的事业，也是年轻人的事业。 清程极智联合创始人师天麾正成为这句话的一个生动的注脚，而他的经历也是当前中国年轻一代AI高端人才的缩影——高中拿下信息学奥林匹克竞赛金 奖保送清华大学，大学确定了系统和高性能计算的研究方向，博士毕业后成为中国科学院计算技术研究所课程讲师、中国信通院万卡智算集群服务能力推 进方阵技术专家。 多个身份标签加持，互联网大厂曾向师天麾抛出高薪的橄榄枝，他最终却选择自己创业，理由也很简单，"做一些不同的事"。在他眼中，大厂"老板安 排"和KPI均是 ...

Core Viewpoint - The article discusses the escalating tensions between U.S. AI companies, particularly Anthropic and NVIDIA, regarding AI export controls and the implications for global AI competition, especially with China [1][9]. Group 1: Anthropic's Background and Position - Anthropic was founded in 2021 by former OpenAI researchers, including Dario Amodei, who aimed to develop responsible AI that benefits humanity [2][4]. - The company has received significant funding, totaling $5.8 billion, including $1.25 billion from Amazon and $500 million from FTX, which has not hindered its financial stability despite FTX's collapse [5]. - Anthropic's Claude series of language models, particularly Claude 3, has emerged as a competitor to OpenAI's ChatGPT, emphasizing ethical AI through its "Constitutional AI" framework [4][5]. Group 2: AI Export Controls and Geopolitical Implications - Anthropic supports stricter AI export controls to maintain U.S. leadership in AI and prevent authoritarian regimes from gaining access to advanced technologies [7][9]. - Dario Amodei argues that export controls are necessary to protect democratic values and prevent the misuse of AI technologies, particularly in the context of potential threats from Chinese companies like DeepSeek [7][9]. - The article highlights the ideological framing of AI competition as a struggle between "democratic AI" and "authoritarian AI," which Anthropic uses to justify its stance on export controls [10][11]. Group 3: Tensions with NVIDIA and Market Dynamics - NVIDIA, a leading AI chip supplier, criticizes Anthropic's claims about chip smuggling and argues that overregulation could harm U.S. competitiveness in the global market [8][9]. - The article notes a division within the U.S. AI industry, where chip suppliers prioritize market openness while model developers like Anthropic focus on the risks of technology proliferation [8]. - Anthropic's reliance on Amazon's Trainium chips rather than NVIDIA's GPUs may provide it with a strategic advantage in the ongoing debate over export controls [8]. Group 4: Chinese AI Competitors and Market Trends - Chinese AI companies, such as DeepSeek, are rapidly advancing, with models that challenge Western counterparts in performance and cost-effectiveness [9][10]. - The article emphasizes that while U.S. companies currently hold an advantage in advanced chips and data centers, China's AI ecosystem is becoming increasingly resilient and innovative [10][11]. - Export controls may provide short-term benefits but could ultimately accelerate China's self-sufficiency in AI technology, as evidenced by the growing capabilities of Chinese models [10][11].

Core Viewpoint - The multi-modal model industry has not yet reached a "GPT-4 moment," as the lack of an integrated understanding-generating architecture is a significant bottleneck for development [1][3]. Company Overview - The company, founded by CEO Jiang Daxin in 2023, focuses on multi-modal models and has undergone internal restructuring to form a "generation-understanding" team from previously separate groups [1][2]. - The company currently employs over 400 people, with 80% in technical roles, fostering a collaborative and open work environment [2]. Technological Insights - The understanding-generating integrated architecture is deemed crucial for the evolution of multi-modal models, allowing for pre-training with vast amounts of image and video data [1][3]. - The company emphasizes the importance of multi-modal capabilities for achieving Artificial General Intelligence (AGI), asserting that any shortcomings in this area could delay progress [12][31]. Market Position and Competition - The company has completed a Series B funding round of several hundred million dollars and is one of the few in the "AI six tigers" that has not abandoned pre-training [3][36]. - The competitive landscape is intense, with major players like OpenAI, Google, and Meta releasing numerous new models, highlighting the urgency for innovation [3][4]. Future Directions - The company plans to enhance its models by integrating reasoning capabilities and long-chain thinking, which are essential for solving complex problems [13][18]. - Future developments will focus on achieving a scalable understanding-generating architecture in the visual domain, which is currently a significant challenge [26][28]. Application Strategy - The company adopts a dual strategy of "super models plus super applications," aiming to leverage multi-modal capabilities and reasoning skills in its applications [31][32]. - The focus on intelligent terminal agents is seen as a key area for growth, with the potential to enhance user experience and task completion through better contextual understanding [32][34].

民营经济促进法创下多个"第一"：我国第一部专门关于民营经济发展的基础性法律，第一次将坚持"两 个毫不动摇"写入其中的法律，第一次明确了民营经济的法律地位。 根据法律规定，民营经济是社会主义市场经济的重要组成部分，是推进中国式现代化的生力军，是高质 量发展的重要基础，是推动我国全面建成社会主义现代化强国、实现中华民族伟大复兴的重要力量。促 进民营经济持续、健康、高质量发展，是国家长期坚持的重大方针政策。 民营经济促进法共9章78条，涉及企业公平竞争、投融资促进、科技创新、规范经营、服务保障、权益 保护等内容。从首次提出到完成立法，仅用不到18个月时间。 作为一部基础性法律，如此迅速地推出与当前背景有关。王振江表示，民营经济在我国国民经济和社会 发展中的地位、作用持续提升。截至3月底，全国登记在册民营企业超过5700万户，占企业总量的 92.3%。 尽快完成立法，"营造有利于包括民营经济在内的各种所有制经济共同发展的法治环境和社会氛围，坚 定不移办好自己的事，不断巩固中国经济企稳回升、长期向好态势，以高质量发展的确定性应对外部环 境急剧变化的不确定性。"王振江说。 法律的生命在于实施。司法部副部长王振江表示，民营 ...

Core Insights - The departure of Dai Zonghong, co-founder and CTO of Zero One Everything, marks a significant shift as he embarks on a new entrepreneurial venture supported by Innovation Works [2] - Zero One Everything has reported an overall revenue of several hundred million this year, indicating a strong financial performance despite recent challenges [2] - The AI Infra team, once a pride of Zero One Everything, faced setbacks in 2024, leading to a strategic pivot towards consumer and business AI applications [4] Group 1: Company Developments - Dai Zonghong was responsible for the AI Infra team, which focused on optimizing training efficiency and reducing costs, achieving a 40% reduction in training costs for the Yi-34B model [3] - The company has been adjusting its project focus based on market product-market fit (PMF), including encouraging independent financing for commercially viable projects [2] - Following the departure of key personnel, including co-founders and early core members, the company has seen a significant reduction in its technical talent pool [4] Group 2: Market Position and Challenges - Zero One Everything struggled to keep pace with the emerging MoE (Mixture of Experts) architecture, resulting in a delayed launch of its flagship model Yi-Lightning [4] - The competitive landscape has shifted, with companies like ByteDance and DeepSeek dominating the high-performance model market, limiting Yi-Lightning's market impact [4] - The company has shifted its business focus towards consumer AI search applications and business model solutions, indicating a strategic realignment in response to market pressures [4]

Macro News - The Ministry of Commerce of China expresses willingness to strengthen cooperation with other countries to promote the healthy and sustainable development of cross-border e-commerce, despite the recent cancellation of the tax exemption policy for small packages from China by the US, which is seen as detrimental to both countries' businesses and consumers [1][2] - The Federal Reserve maintains the federal funds rate target range at 4.25%-4.5%, in line with market expectations, and indicates that it does not plan to preemptively lower interest rates due to ongoing high inflation [2][3] Market Trends - The Shanghai Gold Exchange issues a notice to strengthen market risk prevention measures amid significant fluctuations in international precious metal prices, urging member units to prepare risk emergency plans [3] - Bitcoin approaches the $100,000 mark for the first time since February, leading to a rise in related US stocks, with notable increases in companies such as Strategy (+5.1%) and Hut 8 (+7.7%) [4] - The first quarter results of 85 property insurance companies reveal a total insurance business income of approximately 516.15 billion yuan and a net profit of about 25.60 billion yuan, with 70 companies reporting profits and 15 reporting losses [5] Company Developments - OpenAI appoints Fidji Simo, the CEO of Instacart, as the new Chief Applications Officer, allowing current CEO Sam Altman to focus more on research and safety [5] - Huawei's HarmonyOS makes its debut on computers, with the first HarmonyOS computer set to be released on May 19, featuring deep integration of AI capabilities [6][7]

彭博资讯说，苹果的动作是为因应与Google长期的伙伴关系可能终止，同时代表了搜寻服务这个产业将 会快速出现的重大变化。 库伊告诉法院，苹果计划将新创公司的搜寻工具列入选单之内，供iPhone、iPad等苹果装置的使用者， 在Safari浏览器内可以选择自己偏好的搜寻。不过库伊并没有提及这些可能举措落实的时间表。 库伊也表示，苹果亦考虑增加中国AI公司DeepSeek的AI服务到，以及其他的一些选项到iPhone之中。 苹果一旦改向采用新的搜寻供应商，将威胁到Google与苹果每年约200亿美元的商业关系，目前双方这 样的关系，苹果iPhone和Safari浏览器预设的搜寻引擎是Google，为Google带来数亿使用者，助攻它的 广告事业。 库伊作证是司法部控告Google垄断搜寻市场胜利后，法庭要决定救济方案所进行的听证的一部分。焦点 放在苹果与Google一年200亿美元的合约，让Google成为苹果浏览器的预设搜寻。 库伊作证指出，AI打开Google在搜寻领域首次面对严峻竞争的大门。他说，「在AI出现之前，我的感 觉是其他的选择都不可靠」，「我认为现在的潜力要大得多，因为有新的参与者以不同的方式来解 ...

Group 1 - The core theme of the recent earnings reports from major semiconductor companies is uncertainty, primarily due to changes in U.S. tariff policies and export restrictions to China, leading to unclear product demand [1][3]. - AMD reported that despite exceeding first-quarter earnings expectations, it anticipates a loss of $1.5 billion by the end of the fiscal year due to restrictions on AI chip exports to China [3]. - Marvell postponed its investor day originally scheduled for June 10 to an unspecified date in 2026, citing the current uncertain macroeconomic environment, with its stock dropping 4.4% in pre-market trading [3][4]. Group 2 - Samsung indicated that demand fluctuations are expected to be significant due to tariff policy changes and macroeconomic uncertainty, making it difficult to predict the impact of tariffs and countermeasures on business [4]. - Analysts highlight that the semiconductor industry is facing complex demand signals and geopolitical headwinds, with Marvell's decision to delay its investor day adding to the uncertainty [4][5]. - NVIDIA's CEO stated that China could become a $50 billion AI market within two to three years, emphasizing the potential loss for U.S. companies if they cannot address the issue [5][6]. Group 3 - Chinese companies are increasing investments in self-developed technologies, with firms like Huawei and Alibaba aiming to create competitive products against companies like NVIDIA [7]. - The competitive landscape is intensifying, as U.S. companies must recognize that they are not the only players in the race for AI technology [7].

Group 1 - The core viewpoint of the article emphasizes the need for investors to adapt to the rapidly changing landscape of 2025, focusing on long-term investment strategies amidst uncertainty [2] - The investment philosophy of the company is to find "certainty cuts" in uncertainty, leveraging long-termism to hedge against short-term fluctuations [3][22] - The company has invested in over 600 enterprises, focusing on the integration of "people, events, time, and value" in its investment logic [3][22] Group 2 - The success of DeepSeek is attributed to the resonance between technological iteration and scene demand, highlighting the importance of focusing on specific verticals rather than general models [4][5] - The company is shifting its focus from general AI models to application-based companies that can encapsulate large model capabilities into SaaS tools and niche models trained with industry data [6] - The company emphasizes the need for local governments to move away from "policy following" thinking and instead foster an ecosystem that supports innovation and collaboration [7] Group 3 - The company identifies three key traits in young entrepreneurs: obsessive focus, technical originality from 0 to 1, and rapid error correction ability [8] - The company sees high certainty in three sectors for 2025: AI in the real economy, synthetic biology, and the silver economy [9] - The company advises entrepreneurs to focus on single-point breakthroughs, bind to industry chains, and prioritize cash flow management [10][11] Group 4 - The company is restructuring its fundraising strategy, with government-guided funds now accounting for over 70% of its LP structure, aiming for precise matching with strong industrial cities [12] - The exit strategy is evolving from "betting on IPOs" to a more diversified approach, including mergers and acquisitions [13][14] - The company predicts a strong year for hard tech IPOs, with over 10 companies in semiconductor, AI, and biomedicine expected to reach a market value of over 10 billion [25] Group 5 - The company believes that consumer investment is diversifying, focusing on new domestic products and aging population needs [17] - The company sees structural opportunities in cross-border investments, particularly in Southeast Asia and the Middle East [25][20] - The investment logic of the company is centered around identifying "unicorn" companies with high growth potential and ensuring exit strategies are in place [22]