Most models aren’t helpful out of the box. They need post-training to follow instructions, reason clearly, or act safely. In our latest short course, taught by @BanghuaZ, Assistant Professor at @UW and co-founder of @NexusflowX, you’ll learn how to post-train large language

¯\_(ツ)_/¯. Original seen on Memes for Programmers/Reddit.

This week, in The Batch, Andrew Ng discusses how the U.S. “One Big Beautiful Bill” will shape AI regulation. Plus:. 🕵️ How Anthropic researchers got 16 leading LLMs to commit blackmail.🐝 How an AI-powered beehive keeps bees healthier.🛒 How Walmart built a cloud- and

RT @BanghuaZ: Really excited to work with @AndrewYNg and @DeepLearningAI on this new course on post-training of LLMs—one of the most creat….

Anthropic researchers asked Claude 3.7 Sonnet and DeepSeek-R1 multiple-choice questions, then fed them hints that pointed to wrong answers. When the models followed a hint, they mentioned it in their chain of thought only 25 percent (Claude) and 39 percent (DeepSeek) of the

The U.S. National Hurricane Center (NHC) is testing a graph neural network built by Google’s Weather Lab. The model predicts where and how hard tropical storms will hit two weeks in advance more accurately than conventional methods. This partnership between Google and the NHC

The right data job for you is out there, but how do you find it?. In the final course of the Data Analytics Professional Certificate, you'll learn how to job search smarter:. ✅ Where to look (and why).✅ What job titles to search.✅ How to evaluate a role based on your skills

Biomni is an AI agent for biology research that taps Claude 4 Sonnet and 150 tools, 60 databases, and 100 software packages. While previous biology-research agents typically focus on one subdiscipline, Biomni is built to perform tasks from genomics to pathology. On the

Amazon's Project Rainier will build an interconnected "supercluster" of dozens of next-generation data centers, part of a planned $100 billion in capital spending this year. Amazon has built 7 data centers in New Carlisle, Indiana, with up to 30 planned at that site and more.

Microsoft and academic partners updated BitNet b1.58, constraining all linear-layer weights to –1, 0, or +1 (≈1.58 bits each) yet achieving competitive performance. Across 16 language, math, and coding benchmarks, the model averaged 54.2 percent accuracy while generating 34.5

Meta’s Aria Gen 2 glasses pack five cameras, seven microphones, motion sensors, and a Qualcomm processor into a 75-gram frame, letting researchers capture synchronized video, depth maps, eye-gaze, hand motion, sound, location, and heart rate for up to eight hours. Units will be

Amazon’s Andy Jassy and chief executives at Bank of America, IBM, Shopify, and Williams-Sonoma say generative AI will let them shrink hiring or cut corporate staff. Their statements echo a World Economic Forum survey that found 41 percent of employers expect to downsize thanks to.

This week, in The Batch, Andrew Ng advises scaling down your AI projects so you can practice building AI apps even when you have only a short block of time. Plus:. 🏗️ Amazon plans AI cluster in Indiana.🕶️ Meta details Aria 2 smart glasses.🌀 AI weather prediction gains traction

🚀 AI for Good in Action!. We’re proud to be partners of the AI for Changemakers Program, supporting nonprofits in building AI-driven solutions for world’s most pressing problems, including global education. Over the past weeks, participants of the program - mission-driven

To supercharge its AI push without triggering a government review that would come with acquiring other companies, Meta bought a 49 percent non-voting stake in data-labeling firm Scale AI for $14.3 billion and hired its founder and CEO, Alexandr Wang, and key staff. Wang will.

📈 In business, your insight is only as valuable as the action it inspires. "Data Storytelling," the final course in the Data Analytics Professional Certificate teaches you how to structure your analysis into compelling, actionable narratives, tailored to the people making

RT @predibase: 🧠 Join the 10k developers supercharging their #LLM skills with Reinforcement Fine-tuning—and it's free! 🧠. Reinforcement Fin….

Researchers from Stanford and Princeton fine-tuned a language model to detect racial discrimination buried in millions of historical property deeds in Santa Clara County, California. These clauses, once used to bar people of color from owning or occupying certain homes, remain

This week, in The Batch, Andrew Ng discusses a U.S. court ruling that training LLMs on copyrighted books is fair use. Plus:. 💰 Meta invests $14.3 B in Scale AI.🧬 Biomni AI agent spans life-science research.⚠️ Top CEOs flag AI job cuts.⚡ Ternary BitNet beats most 2B models

RT @AndrewYNg: New Course: ACP: Agent Communication Protocol. Learn to build agents that communicate and collaborate across different frame….