PyTorch
Search documents
陈天奇、贾扬清点赞:Vibe Coding版PyTorch,连论文都是AI写的
机器之心· 2026-01-23 08:29
编辑|Panda、泽南 前两天,Node.js 之父 Ryan Dahl 在 X 上断言:「 人 类编写 代码的 时代已经结束了。 」该帖引发广泛讨论,浏览量更是已经超过了 700 万。而现在,我们迎来了 一个对这一判断的有力证明。 刚刚,英伟达杰出工程师许冰(Bing Xu)在 GitHub 上开源了一个新项目 VibeTensor ,让我们看到了 AI 在编程方面的强大实力。 从名字也能看出来,这是 Vibe Coding 的成果。事实也确实如此,这位谷歌学术引用量超 20 万的工程师在 X 上表示:「 这是第一个完全由 AI 智能体生成的深度 学习系统,没有一行人类编写的代码。 」 更重要的是,许冰强调:「自 2025 年夏天以来,我一行代码都没写过。」他说这项工作是他看过 Andrej Kaparthy 的播客之后开始的。「我当时并不认同他的观 点,所以我和 Terry Chen(英伟达首席工程师)开始用它来测试我们的智能体的能力。弗兰肯斯坦效应最终暴露了我们智能体的一些局限性 —— 但方向很明 确。」 更具体来说,VibeTensor 是一个可运行的深度学习系统,配备了 RCU 风格的调度器、缓存分 ...
硅谷真实「无间道」,OpenAI前CTO怒斩泄密联创,奥特曼打包收了
3 6 Ke· 2026-01-16 12:42
Core Insights - The recent departure of CTO Barret Zoph from Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has created significant turmoil within the company, which is rumored to be valued at $50 billion [2][12] - OpenAI has welcomed Zoph and two other key members back, indicating a strategic move to strengthen its team amidst ongoing competition in the AI sector [1][17] Group 1: Company Developments - Mira Murati announced the separation from Barret Zoph via a brief statement, indicating a lack of amicability in the departure [3][8] - Zoph's dismissal was reportedly due to "misconduct," with allegations of leaking company secrets to competitors [6][9] - OpenAI's CEO Fidji Simo expressed excitement over the return of Zoph, Luke Metz, and Sam Schoenholz, suggesting that this move was planned weeks in advance [8][9] Group 2: Impact on Thinking Machines Lab - The loss of Zoph and other core team members poses a significant challenge for Thinking Machines Lab, especially as it is in a critical fundraising phase [12][28] - The company has appointed Soumith Chintala, known as the "father of PyTorch," as the new CTO to stabilize the situation and maintain engineering capabilities [13][15] - The departure of key personnel raises concerns about the company's governance and internal stability, potentially affecting its market perception [13][28] Group 3: Competitive Landscape - The incident highlights the ongoing "poaching" dynamics in Silicon Valley, where companies like OpenAI and Anthropic are actively recruiting talent from each other [28][30] - The return of Zoph and his colleagues to OpenAI is seen as a reinforcement of its technical strength and leadership position in the AI industry [17][27] - The competitive environment is intensifying, with expectations of further talent shifts in the coming weeks [28][30]
OpenAI核心旧部,再创业又内讧了
3 6 Ke· 2026-01-16 00:31
这都什么样的硅谷剧情! 创业不到一年,联创被开!然后马上跑回了老东家… 最新消息称,2024年年底才从OpenAI出走、联合众OpenAI老将创立Thinking Machines Lab的Barret Zoph,在不到一年的时间里,已选择重返老东家。 虽然在没有竞业协议的硅谷,这样的"折返跑"并不罕见。 但这一次,Barret Zoph离场方式并不大……体面。 一头是前《连线》杂志资深记者Kylie Robison发文表示,Zoph的离职系因发生"不当行为"(unethic problem)被开除。 推特、领英也都光速同步更新。 这是怎么一回事? Barret Zoph跳槽(被开)始终 咱们先来快速过一下时间线。 另一头是,前OpenAI CTO,代班CEO(内乱版),现Thinking Machines Lab CEOMira在爆料后,火速现身说法: 我们已经与Barret Zoph结束了合作关系,新一任CTO由Soumith Chintala担任。 更离谱的是,一位接近Thinking Machines的消息人士向连线杂志表示,Zoph曾向竞争对手泄露公司机密信息。 这一顿刷屏给网友整的是一愣一愣。 最 ...
OpenAI核心旧部,再创业又内讧了
量子位· 2026-01-15 23:57
Core Viewpoint - The article discusses the unexpected departure of Barret Zoph from Thinking Machines Lab due to alleged unethical behavior and his swift return to OpenAI, raising questions about the circumstances surrounding his exit and the implications for both companies [4][12][41]. Group 1: Departure and Return - Barret Zoph was reportedly terminated from Thinking Machines Lab due to "unethical behavior" and was quickly replaced by Soumith Chintala as the new CTO [4][5][8]. - Following his termination, Zoph announced his return to OpenAI, expressing excitement about rejoining the team, which had been in preparation for several weeks [12][13][41]. - The rapid transition from Thinking Machines Lab to OpenAI has sparked speculation about the nature of Zoph's departure and the internal dynamics at both companies [16][23][41]. Group 2: Company Dynamics - Thinking Machines Lab, co-founded by Zoph and others, is currently valued at $50 billion, making it one of the hottest startups in Silicon Valley [32]. - The article highlights a trend of co-founders leaving top AI labs, with OpenAI losing 8 out of 11 co-founders and Thinking Machines Lab losing 3 out of 6 [44]. - The internal conflicts at Thinking Machines Lab, particularly regarding Zoph's departure, suggest deeper issues within the company, as it lost a key co-founder [43][44]. Group 3: Background on Barret Zoph - Barret Zoph was a significant contributor to OpenAI, particularly in the development of GPT-4, and had previously worked at Google Brain [26][30]. - His expertise in optimizing foundational models has been crucial for the practical applications of AI technologies like ChatGPT [28][30]. - The return of Zoph, along with Luke Metz and Sam Schoenholz, is seen as a substantial gain for OpenAI, especially after the recent loss of another research vice president [41][42].
Mira公司内乱?CTO被开除,带团队回OpenAI,翁荔上推发言
机器之心· 2026-01-15 09:17
今天对于 Thinking Machines Lab 和 OpenAI 来说都是不同寻常的一天。 Thinking Machines Lab 创始人兼 CEO Mira Murati 官宣了 与 联合创始人兼 CTO Barret Zoph 的分道扬镳 。 同时,她也宣布了 新任 CTO 的人选 ——Pytorch 之父 Soumith Chintala 。这位在现代 AI 基础设施领域颇具影响力的研究者在去年 11 月初离开了 Meta,并选择 加入 Thinking Machines Lab。 大约 1 个小时后,OpenAI 应用 CEO Fidji Simo 宣布, Barret Zoph 将重返 OpenAI 。 连同他一起回归 OpenAI 的还有 另一位 Thinking Machines Lab 联合创始人 Luke Metz 以及 创始团队成员 Sam Schoenholz 。 机器之心编辑部 两位联合创始人同时从 Thinking Machines Lab「出走」,这一消息在圈内造成了不小的冲击。 根据有人获悉的内部消息, 此次是由于 Barret Zoph 个人的不道德行为,Thinki ...
GPT-4 技术功臣疑似泄密被开除,OpenAI 系创业天团上演「无间道」
3 6 Ke· 2026-01-15 02:29
Core Insights - The AI company Thinking Machines Lab, valued at $12 billion, has dismissed its first CTO Barret Zoph due to alleged unethical behavior [1][2][5] - Soumith Chintala, a prominent figure in AI and a core team member, has been appointed as the new CTO [1][9] - Zoph's dismissal comes amid reports of him leaking confidential information to competitors [2][3] Company Developments - Mira Murati, the founder of Thinking Machines Lab and former CTO of OpenAI, announced Zoph's termination during an all-hands meeting, indicating a swift decision [1][3] - Zoph, along with co-founder Luke Metz and employee Sam Schoenholz, is reportedly returning to OpenAI, which is seen as a significant advantage for the company [5][11] - The internal dynamics at Thinking Machines Lab suggest a complex power struggle and differing interests between the parties involved [13] Leadership Changes - Soumith Chintala, the new CTO, is recognized as the "father of PyTorch" and has a strong background in AI, having previously served as VP at Meta [9][11] - The founding team of Thinking Machines Lab includes several former key members from OpenAI, indicating a strong talent pool [11] Financial Aspects - Thinking Machines Lab raised $2 billion in a seed round led by a16z, with participation from major investors like Nvidia and AMD, setting a record for the largest seed funding round in Silicon Valley history [13]
DeepSeek等8大产品都是意外?! 改变世界的项目们,最初都没被“当个事儿办”
Sou Hu Cai Jing· 2026-01-13 01:47
Core Insights - Many groundbreaking products initially started as side projects, which were not considered significant at their inception [1][2][3][5][6] - Side projects are defined as non-core, non-KPI driven initiatives that are not part of a company's strategic plan [1] - The success of side projects can be attributed to their ability to operate without the constraints typically associated with mainline projects, allowing for greater innovation and flexibility [2][3][6] Group 1: Examples of Successful Side Projects - DeepSeek, a side project of Huansquare Quantitative, emerged from internal technical research and has become a significant tool in quantitative trading [2] - Qwen, developed by Alibaba, was initially a side project that allowed for more autonomy and faster iteration, ultimately leading to its integration into the company's main offerings [3] - Claude Code, initially a simple experimental project by an engineer, evolved into a key product for Anthropic, demonstrating the potential of side projects to gain traction unexpectedly [5] Group 2: Impact of AI on Project Development - The integration of AI into software engineering has lowered the cost of experimentation, enabling individuals to validate ideas more quickly and easily [7][8] - Side projects often begin by addressing specific problems and evolve through real-world usage, which enhances their maturity and relevance [8] - The shift towards AI-driven development suggests that early signals of future trends may increasingly emerge from projects that were initially overlooked [10] Group 3: Strategic Considerations - While AI enhances execution efficiency, it does not necessarily improve the accuracy of strategic judgments, highlighting a potential limitation of mainline projects [10] - The evolving landscape indicates that side projects may play a crucial role in validating directions before scaling up to mainline initiatives [10]
DeepSeek等8大产品都是意外?! 改变世界的项目们,最初都没被“当个事儿办”
量子位· 2026-01-11 04:02
Core Viewpoint - Side projects, often overlooked initially, can lead to groundbreaking products and innovations in the tech industry, demonstrating that exploration and experimentation can yield significant results [1][2][3]. Group 1: Definition and Characteristics of Side Projects - A side project is defined as a non-core, non-KPI driven initiative that is not strategically planned at its inception [2]. - These projects are less constrained by traditional business structures, allowing for more creative freedom and innovation [3][12]. - The lack of formal oversight enables these projects to evolve organically, often leading to unexpected successes [13][40]. Group 2: Examples of Successful Side Projects - DeepSeek, a side project of Huafang Quantitative, emerged from internal technical research and has become a significant tool in quantitative trading [4][11]. - Qwen, initially a side project at Alibaba, has successfully transitioned into a prominent open-source model, benefiting from reduced decision-making constraints [18][22]. - Claude Code started as an experimental project by engineer Boris Cherny and evolved into a key product for Anthropic, showcasing the potential of side projects to disrupt traditional product development [27][32]. Group 3: Advantages of Side Projects - Side projects can enhance the likelihood of success due to less bureaucratic interference, allowing teams to iterate quickly and adapt based on real-world feedback [22][25]. - The cost of experimentation is lower in the AI era, enabling individuals to validate ideas more swiftly without extensive resource coordination [37][44]. - The flexibility of side projects allows for rapid adjustments and improvements, ultimately leading to more robust and mature products [41][43]. Group 4: Implications for Future Projects - The trend indicates that early signals of future innovations may increasingly arise from projects initially deemed non-essential [53]. - While not all side projects guarantee success when scaled, they provide a foundation for larger initiatives once their value is proven [54][55].
那个固执的法国老头走了,带走了硅谷最后的理想主义
AI科技大本营· 2026-01-05 10:12
Core Viewpoint - The departure of Yann LeCun from Meta marks the end of an era characterized by a focus on fundamental AI research, transitioning to a more commercially driven approach under the leadership of Alexandr Wang, emphasizing scale and immediate results over theoretical exploration [4][5][50]. Group 1: Historical Context - In 2013, Facebook was a burgeoning company seeking to integrate AI into its operations, leading to the recruitment of Yann LeCun, a prominent figure in AI research, to establish the Facebook AI Research (FAIR) lab [8][12][13]. - LeCun's vision for FAIR was to create a research environment that prioritized scientific inquiry over commercial pressures, fostering a culture of open exploration [14][23]. Group 2: Contributions and Innovations - LeCun played a pivotal role in the development of PyTorch, a flexible and user-friendly deep learning framework that emerged as a significant competitor to Google's TensorFlow, largely due to the open-source philosophy he championed [17][22][24]. - The success of PyTorch led to a major shift in the academic landscape, with a significant majority of top research papers adopting it, effectively sidelining TensorFlow in the academic community [22][24]. Group 3: Philosophical Divergence - LeCun's philosophical stance on AI emphasized the importance of understanding the underlying principles of intelligence, contrasting sharply with the emerging trend of large language models (LLMs) that he criticized for lacking true comprehension [30][32][36]. - His belief that LLMs were fundamentally flawed due to their reliance on statistical predictions rather than genuine understanding created a rift between him and the evolving priorities at Meta [32][36][50]. Group 4: Transition and Challenges - The rise of Alexandr Wang at Meta signified a shift towards a more aggressive, commercially focused strategy, prioritizing rapid development and deployment of AI technologies over the foundational research ethos that LeCun embodied [48][50]. - LeCun's eventual departure from Meta was driven by a growing disconnect with the company's new direction, which emphasized short-term commercial gains over long-term scientific exploration [52][56]. Group 5: Future Implications - The evolution of FAIR into a more commercially oriented entity under Wang raises questions about the future of AI research and the balance between commercial viability and scientific integrity [42][44][56]. - The legacy of LeCun's contributions, particularly in fostering an open-source culture and prioritizing fundamental research, may influence future developments in AI, as the industry grapples with the implications of prioritizing scale and immediate results [60][62].
不要死磕CUDA,国内首个Triton技术大会官宣,AI芯片编程迎来新范式
AI科技大本营· 2025-12-26 05:42
Core Viewpoint - The article discusses the emergence of Triton as a user-friendly programming tool for AI chip development, aiming to lower the barriers for developers previously reliant on complex languages like CUDA [1][2][3]. Group 1: Triton Overview - Triton allows developers to write high-performance GPU code in a Python-like syntax, making it accessible to a broader audience [3]. - The integration of Triton with the PyTorch ecosystem enhances its usability and performance, making AI chip programming more approachable [3]. Group 2: Triton Next Conference - The Triton Next conference is scheduled for January 9, 2026, in Beijing, organized by the FlagOS community and the Beijing Academy of Artificial Intelligence [4][16]. - The conference aims to explore the current state and future developments of Triton, including discussions on its compiler and potential applications [5][6]. Group 3: Conference Agenda - The morning sessions will focus on the foundational principles of Triton, recent academic research, and the latest developments in the FlagOS community [7]. - Afternoon sessions will highlight practical applications of Triton, featuring insights from various teams on building next-generation AI models and addressing hardware compatibility challenges [11]. Group 4: Workshops and Training - The conference will include hands-on workshops designed to help developers apply Triton in real-world scenarios, covering topics like operator training and compiler usage [15][18].