专题：2025《中国企业家》影响力企业家年会 12月5日-7日，由《中国企业家》杂志社主办的"2025（第二十三届）《中国企业家》影响力企业家年 会"（原中国企业领袖年会）在北京举行，主题为"涌现·无限——共创智能商业新形态"。峰瑞资本创始 合伙人李丰出席并演讲。 作为企业家，要不要今天开始和大家一样赶紧用智能体？李丰回答称，需要先评价一下企业内部和行业 链条中数字化水平是不是比较高了。"如果不高，也许还要再等一等。如果企业自身数字化水平很高 了，也许内部可以用一些智能体。" 他提到，这一轮AI是从大语言模型开始的，数据来自于过去超过40年互联网公开文本数据的积累，才 喂出了今天的大语言模型。 除了企业外，李丰表示，大语言模型最适用的垂直智能体场景，就是在任何商务过程和价值实现过程 中，作为自然语言作为交互方式，进行多轮对话，且实现了价值兑现，这样的行业最容易使用和利用到 AI智能体的垂直场景。 他举例到，比如金融行业，全链条数字化，主要靠专业技术和专业技能进行对话，告诉客户投资原因， 如何选金融产品，风险和潜在收益是什么。还有医疗行业，医生会用数字化设备检测身体，告诉患者如 何预防疾病等。这些行业都是最容易 ...

Core Insights - The focus of academia and industry is shifting towards VLA (Vision-Language-Action) for enhancing autonomous driving capabilities, providing human-like reasoning in vehicle decision-making processes [1][4] - Traditional methods in perception and lane detection are becoming mature, leading to a decline in interest, while VLA is seen as a critical area for development by major players in the autonomous driving sector [4][6] Summary by Sections Introduction to VLA - VLA is categorized into modular VLA, integrated VLA, and reasoning-enhanced VLA, which are essential for improving the reliability and safety of autonomous driving [1][4] Course Overview - A comprehensive course on autonomous driving VLA has been designed, covering foundational algorithms and practical applications, aimed at deepening understanding of the perception systems in autonomous driving [6][21] Course Structure - The course consists of six chapters, starting with an introduction to VLA algorithms, followed by foundational knowledge in Vision, Language, and Action, and culminating in practical assignments [11][19] Chapter Highlights - Chapter 1 provides an overview of VLA algorithms and their development history, along with benchmarks and evaluation metrics [12] - Chapter 2 focuses on the foundational algorithms related to Vision, Language, and Action, including deployment of large models [13] - Chapter 3 discusses VLM (Vision-Language Model) as an interpreter in autonomous driving, covering classic and recent algorithms [14] - Chapter 4 delves into modular and integrated VLA, emphasizing the evolution of language models in planning and control [15] - Chapter 5 explores reasoning-enhanced VLA, introducing new modules for decision-making and action generation [16][18] Practical Applications - The course includes hands-on coding exercises, allowing participants to engage with real-world applications of VLA technologies, such as ReCogDrive and Impromptu VLA [15][18] Learning Outcomes - Participants are expected to gain a thorough understanding of current advancements in VLA, master core algorithms, and apply their knowledge to projects in the autonomous driving field [23][21]

Core Viewpoint - The U.S. government has allowed NVIDIA to export its H200 chips to China and other qualified customers, with a 25% fee on each chip sold, which is expected to benefit NVIDIA significantly in the Chinese market [1][3]. Group 1: Product Details - The NVIDIA H200 chip, launched in November 2023, features a groundbreaking 141GB HBM3e memory system, enhancing its capability to process large models [1]. - Compared to the H100, the H200 has a 76% increase in memory capacity, a 43% increase in bandwidth, and up to a 90% improvement in AI inference performance, making it particularly suitable for large language models and scientific computing [2]. - The current market price for the H200 ranges from 200,000 to 250,000 RMB (approximately 28,000 to 35,000 USD), while complete systems incorporating multiple H200 GPUs can exceed 300,000 USD, potentially reaching over 600,000 USD [2]. Group 2: Market Impact - Major customers for the H200 include cloud service providers (Microsoft, Amazon, Google, Oracle), AI research giants (OpenAI, Meta, Cohere, Mistral), high-performance computing institutions, and enterprise clients across various sectors [2]. - As of December 2025, over 100 large organizations globally are expected to deploy the H200, with numerous smaller clients using it indirectly through cloud services [2]. Group 3: Competitive Landscape - The ability to sell H200 chips to China is seen as advantageous for NVIDIA, as the Chinese market is substantial and developers recognize the CUDA ecosystem [3]. - Despite the potential sales to China, domestic GPU manufacturers are increasingly catching up, with companies like Huawei and Alibaba developing competitive chips that may reduce reliance on NVIDIA [4]. - NVIDIA's CEO has indicated that even with the H200 available for sale in China, local companies may not necessarily choose to purchase it due to the performance of domestic alternatives [4].

Core Insights - OpenAI CEO Sam Altman has declared a "red alert" status due to increasing competition from Google and other AI rivals, particularly following the release of Google's Gemini 3 model [2][6] - The competitive landscape has shifted, with Google now posing a significant threat to OpenAI's ChatGPT, which previously led the market [5][6] Group 1: Competitive Landscape - Google has launched its Gemini 3 model, which has been integrated into its vast ecosystem, achieving 650 million monthly active users in October [4] - Altman acknowledged the need for OpenAI to improve ChatGPT significantly, indicating that the company is under pressure to respond to Google's advancements [6] - The AI race has intensified, with OpenAI needing to secure additional funding of $100 billion while also increasing subscription revenue to meet investor expectations [5] Group 2: Historical Context - Google was once considered the leader in AI research, having developed foundational technologies like the Transformer architecture and the BERT model [4] - The emergence of ChatGPT marked a pivotal moment, shifting the focus of AI development and forcing Google to defend its position in the market [5] - Altman's internal memo suggests that OpenAI employees may need to cancel winter plans to focus on improving ChatGPT, reflecting the urgency of the situation [6]

Core Viewpoint - IBM's CEO Arvind Krishna questions the expected returns on the massive investments made by tech giants like Google and Amazon in AI infrastructure, suggesting that such investments are unlikely to yield reasonable returns due to the high costs associated with data centers [2][3]. Investment and Costs - Goldman Sachs estimates that the global data center market currently consumes about 55 gigawatts of power, with only approximately 14% related to AI. This demand is projected to rise to 84 gigawatts by 2027 due to increasing AI needs [2]. - Krishna calculates that building a 1-gigawatt data center requires an investment of about $80 billion. If a company commits to constructing 20 to 30 gigawatts of data centers, the capital expenditure could reach $1.5 trillion, nearly equivalent to Tesla's current market value [2]. - If all major cloud providers expand to around 100 gigawatts of capacity, it would necessitate an investment of approximately $8 trillion, with the required profit scale to cover this expenditure being staggering [2][3]. Profitability Concerns - Krishna emphasizes that $8 trillion in capital expenditure would require around $800 billion in profits just to cover interest payments, making it highly unlikely for such investments to be profitable [3]. - The rapid technological advancements mean that the chips relied upon in data centers quickly become obsolete, further complicating the return on investment [3]. AI Development and Market Trends - Despite the ongoing investment surge, Krishna believes the probability of achieving general artificial intelligence with current technologies is at most 1%. He acknowledges the significant value of this technology, which could unlock trillions of dollars in productivity potential, but asserts that the technological requirements far exceed those of current large language models [5]. - Major cloud providers are accelerating their investments in AI infrastructure, with expected expenditures reaching about $380 billion this year. Alphabet has raised its 2025 capital expenditure forecast from $85 billion to between $91 billion and $93 billion, while Amazon has increased its forecast from $118 billion to $125 billion [5].

Core Insights - The future of human-machine coexistence is an inevitable trend, and universities should move beyond traditional knowledge transmission to focus on cultivating meta-knowledge, tacit knowledge, and practical knowledge while enhancing human empathy and collaborative abilities to address the challenges posed by changes in communication forms [3][7]. Group 1: AI Development and Challenges - The foundation of artificial intelligence is based on three elements: chips, data, and algorithms, driven by Moore's Law, which has led to the emergence of large language models through massive data generated by mobile sensors and powerful chip analysis capabilities [3][7]. - Large language models have a critical shortcoming due to their lack of physical embodiment, which prevents them from providing essential variables such as gender, age, and region necessary for human communication, making it difficult to establish stable trust relationships [3][7]. Group 2: Human Interaction and Education - Despite the presence of AI with physical forms, humans still desire real-life interactions and connections, emphasizing that the essence of human learning and communication is a multi-faceted, multi-channel, and social process that cannot be satisfied by mere artificial voice or online interactions [3][7]. - AI may replace certain types of knowledge production and some cognitive tasks, but human empathy and collaborative abilities, rooted in carbon-based life, remain irreplaceable core competencies [4][8]. Group 3: The Role of Universities - The mission of universities is to help individuals become better people in a future characterized by human-machine coexistence, maintaining their social and practical characteristics while cultivating core knowledge and unique abilities [4][8]. - In the face of communication changes and knowledge iteration brought about by AI, universities must assist humanity in maintaining core values amid competition and collaboration with technology, achieving harmonious coexistence between humans and machines [4][8].

Core Insights - DeepSeek has released two new models, DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, which have garnered attention for their performance in comparison to leading models like OpenAI's GPT-5 and Google's Gemini3 Pro [1][2] Model Features - DeepSeek-V3.2 is designed as a high-efficiency assistant with strong reasoning and agent capabilities, aimed at automating complex tasks such as report generation and coding [2] - DeepSeek-V3.2-Speciale focuses on solving high-difficulty mathematical problems and academic research, pushing the limits of open-source model reasoning [2] Technological Innovations - The new models incorporate two significant breakthroughs: Domain-Specific Architecture (DSA) and Thinking Tool Invocation technology [2] - DSA enhances efficiency by allowing the model to retrieve only the most relevant information, reducing resource consumption [2] - Thinking Tool Invocation enables multi-round reasoning and tool usage, allowing the model to think, execute, and iterate on tasks autonomously [2] Market Positioning - The release of these models aims to bridge the performance gap between open-source and closed-source models, providing a competitive edge for open-source development [3][4] - DeepSeek's focus on practicality and generalization is intended to create pressure on closed-source vendors, transforming aspirations into competitive realities [4] Community Engagement - DeepSeek has updated its official web platform, app, and API to the new version, while the Speciale version is currently available only as a temporary API for community evaluation [4]

Core Insights - UBS highlights the recent advancements in AI with the launch of new large language models (LLMs) by companies like Google, Anthropic, and DeepSeek, intensifying competition in the industry [1] - The report emphasizes the continued relevance of the "scaling laws" in model performance, indicating that computational power will remain a critical factor in determining the AI competitive landscape [1] Model Performance - The latest generation of models has shown significant breakthroughs, with Gemini 3 Deep Think and Claude Opus 4.5 achieving multi-step reasoning task scores of 45% and 38%, respectively, surpassing previous models that scored between 10%-20% [2] - This performance aligns with the effectiveness of the AI model pre-training scaling laws, where increased computational investment leads to non-linear improvements in model capabilities [2] Chip Technology Competition - Google’s Gemini 3 Pro is trained entirely on self-developed TPU chips, sparking discussions about the competition between GPUs and AI-specific ASIC chips [2] - ASIC chips are noted for their higher efficiency in specific AI tasks, while GPUs maintain a 90% market share in data center chips due to their flexible architecture and extensive software ecosystem [2] - The collaboration between OpenAI and Broadcom, as well as Anthropic and Google, is expected to enhance the focus on ASIC chips, with both chip types anticipated to coexist in the future [2] Market Trends - The introduction of next-generation chips like NVIDIA's Blackwell and Rubin is expected to sustain the competition for computational expansion, leading to an upward revision of AI capital expenditure forecasts by UBS [3] - The advancements from Google, Anthropic, and DeepSeek are increasing competitive pressure on companies like OpenAI, driving the AI industry towards a multi-model and multi-vendor landscape, a trend expected to persist at least until 2026 [3]

Core Insights - Reinforcement Learning (RL) has become a key technology paradigm for enhancing the complex reasoning and problem-solving capabilities of Large Language Models (LLMs) [2] - The main challenge in RL for LLMs is the mismatch between sequence-level rewards and token-level optimization objectives, raising concerns about theoretical soundness and training stability [2][5] - A new RL formulation method proposed by Alibaba's Qianwen team focuses on optimizing the expected value of sequence-level rewards using a surrogate token-level objective as a first-order approximation [2][11] Methodology - The team defines an autoregressive LLM represented by a policy π_θ, focusing on sequence-level rewards where a scalar reward R(x, y) is assigned to the entire response y [6] - The decision to avoid value function methods stems from the difficulty in constructing a general, scalable, and reliable value model [7] - Directly optimizing the expected sequence-level reward is challenging due to numerical differences between training and inference [9] Key Findings - The team conducted extensive experiments using a 30 billion parameter MoE model, consuming hundreds of thousands of GPU hours [4] - The introduction of on-policy training with importance sampling correction achieved the highest training stability [10] - In off-policy updates, both clipping and Routing Replay are essential for maintaining training stability, as their absence leads to performance degradation [23] Experimental Results - The MiniRL algorithm, which incorporates importance sampling, demonstrated the best performance and stability during training [22] - The removal of importance sampling correction during training led to rapid collapse and a sharp decrease in entropy, confirming its critical role in the first-order approximation [22] - Different cold-start initialization methods yielded similar final performance, indicating that the focus should be on the RL methods themselves rather than initialization details [27]

AI之争就是生态之争 作者 | 吕敬之 来源 | #融中财经 11月20日，Gemini3推出两天后，在被称为"硅谷投资人春晚"的Cerebral Valley AI峰会上，OpenAI就被 选成了"第二大可能倒闭的AI独角兽"。 同一天，Sam Altman推送了一条内部备忘录，承认了OpenAI在预训练上已经落后于谷歌的表现。 十天后，在12月1日，Altman再次推送全员内部信，这一次，口气更加严厉，发起内部"红色预警"叫停 广告商业化、AI agent的所有尝试，把所有人的所有注意力重新调回到ChatGPT性能提升上。硅谷知名 投资人Deedy Das在X上评价，Gemini3上线十五天，ChatGPT的日均访问量已经掉了约1200万，这也是 OpenAI拉响红色警报的真正原因。 随着谷歌的穷追猛打，用户和投资人也开始意识到，AI之争，争的不只是用户数据、快速商业化，而 是长远的生态之争。 被谷歌抢走1000万流量的ChatGPT 在Gemini3上线的第十六天，OpenAI传出发布新的大模型开启反击战的消息。 据The information最新报道，OpenAI在近几周的人工智能开发竞赛中似乎已落 ...