探索环线（Discovery Loop）成立：Jeff Dean 等四位谷歌元老创办 AI 自动化科研公司

自动化发现，加速全球科学与工程的发展。

科学方法是人类曾经发明的最伟大的工具之一，但其执行涉及重复的实验循环，而以今天的手工方式很难实现规模化：你提出一个实验、实施并运行它、检查结果，然后迭代以改进你的方法。

从历史上看，科学进步依赖于这些连续的人类迭代。在许多领域，这一过程仍然极其缓慢且劳动密集。

在Discovery Loop，我们正在构建系统来自动化整个实验循环。通过利用前沿的人工智能模型和大规模计算基础设施，我们的系统将能够快速提出、运行并从评估中学习。

这种方法允许成千上万的实验并行执行，大大压缩迭代时间，并提高科学和工程产出的数量和质量。

我们将首先专注于自动化机器学习研究和工程的过程。

我们将使用这些自动化的机器学习能力快速优化我们自己的技术堆栈，然后再扩展到其他领域。

我们相信，我们的方法能够在科学和工程领域解决任何有可衡量结果的学习循环。最终，我们正在构建能够应对国家工程院（NAE）重大挑战的系统——例如设计更好的药物、推进健康信息学、使太阳能经济可行、提供洁净水源、保障网络安全以及开发科学发现的工具。

我们的使命很简单：我们正在构建能够自动解决机器学习、科学和工程中重要问题的人工智能解决方案。通过加快我们进行工程和科学发现的速度，我们可以更快地将科学和技术的益处带给世界。最终，我们的目标是构建人工智能系统，作为一种深远积极的、赋能人类的力量，提供能够在全球范围内改善人们生活的技术解决方案。

我们的创始团队——Jeff Dean、Sanjay Ghemawat、Quoc Le 和 Oriol Vinyals——拥有深厚的友谊历史以及数十年的紧密且具有影响力的合作经验。

总体而言，我们代表了人工智能领域中三位被引用次数最多的研究人员，以及分布式系统领域中两位被引用次数最多的研究人员。

在我们之间，我们开创了大规模计算，并领导了关键基础设施、产品以及世界依赖的基础性人工智能进展的创造，包括多代 Google 搜索、Google 广告、Google 新闻、Google 翻译、Google 文件系统 (GFS)、MapReduce、BigTable、Spanner、TensorFlow、Pathways、TPU、AlphaChip、AlphaStar、AlphaCode、AlphaFold、Gemini、模型蒸馏、专家混合模型架构、word2vec、序列到序列模型、链式思维推理、神经结构搜索以及多代大型语言模型 (LLM) 等。

我们的相对优势不仅在于技术能力，更在于我们曾经构建过的系统前所未有的规模。我们拥有从芯片、硬件基础设施、软件基础设施、机器学习模型到产品的真正全栈深度。

想象一个未来，少数人就能比如今庞大的科学家和工程师团队更快速、更高质量地进行科学研究和工程任务。通过自动化发现循环，世界将在无数科学领域实现更快速的进步。

我们正在组建一个精干的现场团队，以执行这一变革性愿景。

Automating discovery to accelerate science and engineering for the world.

The scientific method is one of the greatest tools humanity has ever devised, yet execution entails repetitive experimental loops that are hard to scale with today's manual efforts: you propose an experiment, implement and run it, examine the results, then iterate to refine your approach.

Historically, scientific progress has relied on these sequential human iterations. In many domains, this process remains incredibly slow and labor-intensive.

At Discovery Loop, we are building systems to automate these entire experimental loops. By utilizing frontier AI models and large-scale computational infrastructure, our systems will be able to rapidly propose, run, and learn from evaluations.

This approach allows for the parallel execution of thousands of experiments, drastically compressing iteration time and driving up the quantity and quality of scientific and engineering output.

We will initially focus on automating the process of machine learning research and engineering.

We will use these automated ML capabilities to rapidly optimize our own technology stack before expanding to other domains.

We believe our approach will be able to solve any learning loop with measurable outcomes within the domains of science and engineering. Ultimately, we are building systems capable of taking on National Academy of Engineering (NAE) Grand Challenges—such as engineering better medicines, advancing health informatics, making solar energy economical, providing access to clean water, securing cyberspace, and engineering the tools of scientific discovery.

Our mission is straightforward: we are building AI solutions that can automatically solve important problems in machine learning, science, and engineering. By advancing the pace at which we conduct engineering and scientific discovery, we can bring the benefits of science and technology to the world much faster. Ultimately, our goal is to build AI systems that act as a deeply positive, empowering force for humanity, delivering technology solutions that improve people's lives on a global scale.

Our founding team — Jeff Dean, Sanjay Ghemawat, Quoc Le, and Oriol Vinyals — has a shared history of deep friendship and decades of close and impactful collaboration.

Collectively, we represent three of the most-cited researchers in artificial intelligence and two of the most-cited researchers in distributed systems.

Between us, we have pioneered massive scale computing and led the creation of critical infrastructure, products, and foundational AI advances that the world relies on, including multiple generations of Google Search, Google Ads, Google News, Google Translate, Google File System, MapReduce, BigTable, Spanner, TensorFlow, Pathways, TPUs, AlphaChip, AlphaStar, AlphaCode, AlphaFold, Gemini, model distillation, mixture-of-experts model architectures, word2vec, sequence-to-sequence models, chain of thought reasoning, neural architecture search, and multiple generations of Large Language Models (LLMs) among others.

Our relative advantage isn't just our technical ability; it is the unprecedented scale of the systems we have previously built. We possess true full-stack depth that spans chips, hardware infrastructure, software infrastructure, ML models, and products.

Imagine a future where a handful of people can conduct scientific research and engineering tasks much more rapidly, and with higher quality, than massive teams of scientists and engineers do today. By automating the loops of discovery, the world will be able to make much more rapid advances across countless fields of science.

We are building a lean, in-person team to execute this transformative vision.