01 按参数算，人脑相当于多大的模型？ 答案是：要看怎么算。 如果只看神经元的个数，人脑大概是860亿个神经元，也就是86B的模型，并不大。 参考一下，DeepSeek V3是671B，Kimi K2.5大概1000B，即1T； 但事实上人脑每个神经元又有7000个突触，从技术的角度类比，颗粒度更小的突触才更像AI模型的权重参数。 如果这么算，860亿*7000，那么人类大脑相当于大约600T模型。 而这么大的模型，今天的硬件肯定暂时还跑不动。 这么类比略糙。 但也说明——咱们这颗脑子的架构还是很复杂的，属于先进制程。 有点牛逼。 02 那么，大脑的制程到底有多先进呢？ 我随即问了Claude opus 4.6和Gemini 3.1 Pro一个问题（实在受不了GPT无比谄媚的风格）—— "如果人脑是一块芯片，那么它的制程是几纳米的？" 他们的答案出奇一致： 如果看神经元细胞体直径的直径，大概 10000-100000 纳米 (10-100微米)。 这么看大脑相当于几十年前的电子管计算机。 这TM也太落后了。 但逻辑显然不是这样的： 神经元并非一个简单的开关，它更像处理器的一个核，真正的开关和信号传递发生在突触 ...

Core Insights - The article highlights the unconventional path of Keller Jordan, who secured a position at OpenAI without a PhD or traditional research background, emphasizing the importance of open-source projects and practical contributions in the AI field [1][3]. Group 1: Keller Jordan's Journey - Keller Jordan graduated from UCSD in 2020 with dual degrees in mathematics and computer science, without having published any papers [5]. - His first job was at an AI content moderation startup, where he began to explore improvements in existing research [5]. - After reaching out to Google researcher Behnam for guidance, he collaborated on a project that led to a paper presented at ICLR [8]. Group 2: Contributions to AI Research - Keller's work on "NanoGPT speed run" significantly improved the training efficiency of Transformer models, achieving a 3.8 times increase in token efficiency, reducing the required tokens from approximately 10 billion to 2.7 billion [9][10]. - The design of the speed run was innovative, allowing for low-cost experimentation, with a single attempt costing as little as $8, making it accessible for individual researchers [12][13]. Group 3: Development of Muon - Keller developed an optimizer named Muon, which optimized the hidden layers of neural networks, achieving record training speeds for NanoGPT and CIFAR-10 [14][19]. - Muon demonstrated superior performance compared to the widely used AdamW optimizer, particularly as model sizes increase, indicating a potential breakthrough in AI model training [19]. Group 4: Entry into OpenAI - Keller officially joined OpenAI in December 2024, following the success of Muon in the developer community [20]. - He expressed a preference for continuing his research over publishing a paper, criticizing the prevalence of low-quality optimization papers in the field [21]. Group 5: Other Success Stories - The article also mentions other individuals who have successfully transitioned into major AI companies without traditional academic credentials, such as Sholto Douglas at Google DeepMind and Andy Jones at Anthropic, highlighting a trend of talent recognition based on practical contributions rather than formal publications [23][25][28].

证券研究报告 2026.02.11 人工智能十年展望（二十七）：越过"遗忘" 的边界，模型记忆的三层架构与产业机遇 SAC 执证编号：S0080518070011 SFC CE Ref：BOP246 于钟海 分析员 韩蕊 分析员 王之昊 分析员 SAC 执证编号：S0080523070010 SFC CE Ref：BXD683 rui.han@cicc.com.cn SAC 执证编号：S0080522050001 SFC CE Ref：BSS168 zhihao3.wang@cicc.com.cn 纵轴：相对值（%） 88 100 112 124 136 148 2025-02 2025-05 2025-08 2025-10 2026-01 沪深300 中金软件及服务 投资建议 大模型的演进史，本质上是一部与"遗忘"抗争的历史。当我们惊 叹于模型的推理能力时，往往忽视了一个重要短板：在缺乏记忆留 存的架构下，模型每一次对历史信息的处理，本质上都是一次昂贵 的"重复计算"。这种以高昂算力对抗遗忘的粗放模式，正面临着显 存墙与上下文窗口的物理极限。我们认为，2026 年及之后的AI Infra主战场将增加"模型记 ...

我来为您详细解读这份腾讯研究院发布的《塑造自己的下一个版本：2026前沿科技趋势》报告。这份报告以"用户视 角"出发，眺望2030年的自己，围绕五个维度展开前沿科技趋势分析。 --- 一、生命力2030：从"活得久"到"活得好" 核心观点：人类生命正经历"第三次转型" 报告开篇指出一个关键转折：过去一百年人类寿命翻了一倍，但从1900年到2000年的快速增长后，预期寿命增速已大 幅放缓。2024年《自然·衰老》期刊的研究表明，通过消除早夭和中年疾病来延长寿命的"容易摘的果实"已被采摘殆 尽。 - Alnylam Pharmaceuticals开发的RNA干扰技术仅需每六个月一次皮下注射，即可控制高血压 - 斯坦福大学开发的mRNA CAR-T技术在小鼠淋巴瘤模型中实现75%的长期无瘤生存 新范式诞生：全球正在从追求单纯的"寿命"（Lifespan）转向追求"健康寿命"（Healthspan）——即在没有严重慢性病、 残疾或认知功能衰退的情况下维持良好生活质量的年限。据世界经济论坛报告，若将人类健康寿命延长1年，产生的全 球经济价值将高达38万亿美元。 三大技术支柱 基因疗法进入"生命代码优化"时代 - CRI ...

Core Insights - Demis Hassabis, CEO of DeepMind, predicts a new era of discovery and a renaissance in the next 10 to 15 years, driven by advancements in artificial intelligence (AI) [2] - AI is expected to revolutionize medicine through personalized treatments and breakthroughs in energy solutions, enabling interstellar exploration [2] - Google faces a critical turning point with the rise of generative AI, necessitating a self-disruption to maintain its competitive edge [3] Company Restructuring - In response to competition from OpenAI's ChatGPT, Google underwent a significant internal restructuring, merging its research divisions Google Brain and DeepMind under Hassabis's leadership [3] - The integration aims to leverage the strengths of both teams and enhance the computational capabilities required for training advanced models like Gemini [3] - Following the release of models such as Gemini 3 and Nano Banana, Alphabet's stock surged approximately 65% by the end of the year, indicating the success of the restructuring strategy [3] AI in Biology - Hassabis highlights the application of AI in biology as foundational for the new era, citing DeepMind's AlphaFold model, which addresses the protein folding problem and predicts the 3D structures of over 200 million proteins [4] - AlphaFold has been utilized by over 3 million researchers and has contributed to Hassabis receiving the 2024 Nobel Prize in Chemistry [4] - Isomorphic Labs, a subsidiary of Google, is applying AlphaFold technology to drug development, aiming to enhance efficiency by 1000 times through computer simulations [4] Future Aspirations - Hassabis emphasizes that the challenging schedule and company restructuring are necessary steps toward achieving the ultimate goal of solving intelligence and leveraging it to address other global issues [5] - The next decade is anticipated to be a period of significant technological transformation, but Hassabis remains confident in the long-term vision of utilizing intelligence to tackle various challenges, potentially including space exploration [5]

Project Genie is Google’s research prototype for creating playable worlds. With just a text or image prompt, you can generate interactive environments that build in real time, as you play them. Project Genie also features a gallery of premade worlds that users can play or remix. In this world, you can fly through the air as you descend from a mountaintop, and attempt to fly through the rings in your path. Learn more: https://labs.google/projectgenie ___ Subscribe to our channel https://www.youtube.com/@goog ...

Project Genie is Google’s research prototype for creating playable worlds. With just a text or image prompt, you can generate interactive environments that build in real time, as you play them. Project Genie also features a gallery of premade worlds that users can play or remix. In this world, scan the wetlands with your flashlight, and try to spot the foxes wandering in the dark. Learn more: https://labs.google/projectgenie ___ Subscribe to our channel https://www.youtube.com/@googledeepmind Find us on X h ...

Project Genie is Google’s research prototype for creating playable worlds. With just a text or image prompt, you can generate interactive environments that build in real time, as you play them. Project Genie also features a gallery of premade worlds that users can play or remix. In this world, you can explore the ruins of a mysterious monument as an RC car. Learn more: https://labs.google/projectgenie ___ Subscribe to our channel https://www.youtube.com/@googledeepmind Find us on X https://twitter.com/Googl ...

Project Genie is Google’s research prototype for creating playable worlds. With just a text or image prompt, you can generate interactive environments that build in real time, as you play them. Project Genie also features a gallery of premade worlds that users can play or remix. In this world, the blank space is your canvas, as your blue orb leaves a trail of paint everywhere you go, testing Genie's memory of your actions, reflections in the environment, and the physics of water. Learn more: https://labs.go ...

Project Genie is Google’s research prototype for creating playable worlds. With just a text or image prompt, you can generate interactive environments that build in real time, as you play them. Project Genie also features a gallery of premade worlds that users can play or remix. In this world, you can hop through a rainy day on watercolor canvas as a pink rabbit. Learn more: https://labs.google/projectgenie ___ Subscribe to our channel https://www.youtube.com/@googledeepmind Find us on X https://twitter.com ...