The Rundown AI -每日精选 07月31日 15:33
Stanford's virtual AI lab speeds discovery
index_new5.html
../../../zaker_core/zaker_tpl_static/wap/tpl_guoji1.html

 

斯坦福大学的虚拟AI实验室正通过一支由专业AI科学家组成的团队,以前所未有的速度推动科学研究。这些AI科学家能够自主进行会议、设计实验并产出实验室级结果,将原本需要数月的研究周期缩短至数天。该虚拟实验室的AI首席研究员可以组建专门的AI代理,进行高效的会议和实验设计,人类研究员仅需介入1%的时间。AI团队已成功设计出92种纳米抗体,其中两种在实际测试中能有效结合COVID-19变种病毒。这种模式的出现,意味着科学发现不再受限于人类的时间、精力、资源和专业知识,预示着科学研究进入了一个自主、全天候的新时代。

🔬 斯坦福大学的AI虚拟实验室构建了一支由AI科学家组成的团队,能够自主设计、讨论并测试生物医学发现,显著加速了科研进程,例如在数天内就生成了COVID-19的纳米抗体候选物。

💡 该实验室的核心特点是其“AI首席研究员”能够组建专门的AI代理,进行高效的会议并独立请求AlphaFold等工具辅助研究策略,人类研究员的干预率极低,仅为1%,极大地提升了研究效率。

🧬 AI团队不仅设计了92种纳米抗体,其中两种在实际物理实验室测试中成功结合了近期的SARS-CoV-2变种病毒,证明了其在解决实际科学问题上的能力。

🚀 AI研究团队的出现打破了传统科学研究中对人类时间、精力、资源和专业知识的限制,随着AI能力的不断提升,科学发现的速度和模式将发生颠覆性变化。

📚 OpenAI推出的ChatGPT“学习模式”旨在通过引导性问题和反馈,而非直接提供答案,帮助学生深入理解问题,将AI定位为教育的辅助伙伴,而非简单的答案提供者。

Read Online | Sign Up | Advertise

Good morning, {{ first_name | AI enthusiasts }}. AI is already being integrated into laboratories across scientific fields… But what happens when a team of agents runs its own lab entirely?

Stanford’s virtual AI lab is doing just that, with specialized AI scientists running meetings, designing experiments, and producing lab-ready results in days instead of months — speeding us into the era of autonomous, 24/7 scientific discovery.


In today’s AI rundown:

    Stanford’s AI-powered virtual scientists

    Meta targets Mira Murati's startup with massive offers

    Build apps with leading open-source AI

    ChatGPT’s new Study Mode for deeper learning

    4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

AI & SCIENCE

👨‍🔬 Stanford’s AI-powered virtual scientists

Image source: James Zou (@james_y_zou on X)

The Rundown: Researchers from Stanford and the Chan Zuckerberg Biohub just developed a “virtual lab” of AI scientists that design, debate, and test biomedical discoveries — already generating COVID-19 nanobody candidates in days.

The details:

    The lab features an “AI principal investigator” that assembles specialized agents that conduct meetings lasting seconds instead of hours.

    Human researchers needed to intervene just 1% of the time, allowing AI agents to request tools like AlphaFold to aid in research strategy independently.

    The AI team produced 92 nanobody designs, with two successfully binding to recent SARS-CoV-2 variants when tested in physical laboratories.

    The AI lab also releases full transcripts of the AI team’s reasoning, letting human researchers review, steer, or validate the process as needed.

Why it matters: The arrival of teams of AI research teams means science is no longer capped by human limits on time, energy, resources, and expertise. With agentic capabilities only continuing to scale, the pace of discovery is about to completely change, along with the traditional notions of scientific research.

TOGETHER WITH HUBSPOT

🧠 100+ ChatGPT prompts to revolutionize your workflow

The Rundown: HubSpot’s free, comprehensive  “How to Use ChatGPT at Work” guide provides 100+ ready-to-use prompts to help professionals boost efficiency and adopt AI-driven workflows.

Inside, you’ll find:

    A quick crash course to master ChatGPT in under 30 minutes

    Practical industry use cases to spark real-world inspiration

    100+ prompts to streamline tasks and accelerate productivity

    Expert tips to tackle common AI roadblocks with confidence

Get your free copy and join 10,000+ professionals leveling up with AI.

META & THINKING MACHINES LAB

💰 Meta targets Mira Murati's startup with massive offers

Image source: Ideogram / The Rundown

The Rundown: Meta has approached over a dozen employees at ex-OpenAI CTO Mira Murati's Thinking Machines Lab, according to Wired, offering massive compensation packages (including one exceeding $1B) to join its superintelligence team.

The details:

    Zuckerberg’s outreach reportedly includes personally messaging recruits via WhatsApp, followed by interviews with him and other executives.

    Compensation packages ranged from $200-500M over four years, with first-year guarantees between $50-100M for some, and one offer over $1B.

    The report also detailed that Meta CTO Andrew Bosworth’s pitch has centered on commoditizing AI with open source models to undercut rivals like OpenAI.

    Despite the offers, not a single person from the company has accepted, with WIRED reporting industry skepticism over MSL’s strategy and roadmap.

Why it matters: We thought the naming of Shengjia Zhao as chief scientist might be a final bow on the MSL team, but Zuck clearly isn’t stopping in his pursuit of top AI talent at all costs. TML’s staff decline is both a potential testament to their incoming first product and a window into how the industry is viewing Meta’s new venture.

AI TRAINING

💻 Build apps with leading open-source AI

The Rundown: In this tutorial, you’ll learn how to use Alibaba’s new Qwen 3 Coder, a competitive coding model that matches premium offerings — featuring both browser-based chat and CLI access for developers.

Step-by-step:

    Visit Qwen Chat, create a free account, and select Qwen3-Coder as your model

    Test with simple prompts like “Create a Twitter clone in one file” and use the Preview button to see results

    Refine with follow-up prompts: “Add images and make it more complete” to expand functionality

    Install CLI with npm install -g qwen-code/qwen-code, then type qwen in your terminal for command-line access

Pro tip: Qwen gives you 1M free tokens and performance on par with premium tools—plus, it’s fully open-source. Perfect for devs who want power without the paywall.

PRESENTED BY GOOGLE CLOUD

📶 Scale your AI startup with Google Cloud

The Rundown: Google for Startups Cloud Program connects AI innovators with cutting-edge technical applications to fuel your product’s journey from early development to massive scale, leveraging infrastructure built to handle your rapid growth.

Inside the program, you’ll gain:

    Up to $350K in cloud credits to scale faster

    Access to Google’s core AI tools like Vertex AI and Gemini

    Dedicated expert guidance for customized support and solutions

Start building with Google Cloud — apply now.

OPENAI

📚 ChatGPT’s Study Mode for deeper learning

Image source: OpenAI

The Rundown: OpenAI just rolled out Study Mode for ChatGPT, a new feature designed to guide students through problems step-by-step, using Socratic questions and feedback instead of just providing solutions.

The details:

    Study Mode uses custom instructions created with teaching experts to ask guiding questions and provide interactive learning instead of direct answers.

    The AI guides learners with interactive prompts, hints, and knowledge checks to encourage active participation and deeper understanding.

    ChatGPT will actively resist requests for quick solutions in this mode, instead redirecting students back to the learning process.

    The rollout begins immediately for Free, Plus, Pro, and Team users, with educational institutions receiving access within the next few weeks.

Why it matters: AI has shown huge promise in offering personalized learning, but many teachers would also say it’s upended the educational system. Tools like Study Mode offer the potential for AI to be an educational partner instead of a shortcut, but right now, their success might largely depend on student buy-in.

QUICK HITS

🛠️ Trending AI Tools

    🎥 Wan2.2 - Alibaba’s new open MoE model for AI video generation

    🤖 GLM-4.5 - Z.ai’s new SOTA open-source agentic AI family

    📊 Shortcut AI - AI agent for Excel spreadsheets

    🖥️ Copilot Mode - New agentic capabilities for Microsoft Edge

💼 AI Job Opportunities

📰 Everything else in AI today

Meta’s superintelligence team poached AI researcher Bowen Zhang from Apple’s foundation models group, marking the fourth departure in the last month.

Google’s NotebookLM is rolling out Video Overviews, giving users the ability to generate narrated slides on any topic or document.

Microsoft is reportedly nearing a deal to retain access to OpenAI’s tech even after the company’s AGI milestone, a current point of contention in terms of the partnership.

xAI opened the waitlist for its upcoming “Imagine” image and video generation feature, which will reportedly include audio capabilities similar to Google’s Veo 3.

Adobe unveiled new AI features for editing in Photoshop, including Harmonize for realistic blending, Generative Upscale, and more.

Ideogram released Character, a character consistency model allowing users to place a specific person into existing scenes and new outputs from a single reference photo.

Writer launched Action Agent, an enterprise AI agent that executes tasks and uses tools in its own environment, beating Manus and OAI Deep Research on benchmarks.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, August 1, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of this workshop, you’ll have practical strategies to get the AI to do exactly what you want.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Fish AI Reader

Fish AI Reader

AI辅助创作,多种专业模板,深度分析,高质量内容生成。从观点提取到深度思考,FishAI为您提供全方位的创作支持。新版本引入自定义参数,让您的创作更加个性化和精准。

FishAI

FishAI

鱼阅,AI 时代的下一个智能信息助手,助你摆脱信息焦虑

联系邮箱 441953276@qq.com

相关标签

AI科学家 虚拟实验室 科学发现 ChatGPT学习模式 生物医学
相关文章