This isn’t traditional business news

Welcome to Morning Brew—the free newsletter designed to keep you in the know on the business news impacting your career, company, and life—in a way you didn’t know you needed.

Note: this isn’t traditional business news. Morning Brew’s approach cuts through the noise and bore of classic business media, opting for short writeups, witty jokes, and above all—presenting the facts.

Save time, actually enjoy business news, and join over 4 million professionals reading daily.

Check it out

Today:

• Humanoid Robots Surge in 2025

• Musk to Unveils Grok 3 Monday

• Meta Enters Humanoid Robotics Race

• UK Tests Claude for Public Services

• Dell Secures $5B xAI Deal

• Tencent, Baidu Bet on DeepSeek

Legions of Robots UNLEASHED in 2025... Robotic Era BEGINS.

Robotics is surging in 2025, driven by AI and engineering breakthroughs. Humanoid robots like Tesla’s Optimus and Figure AI’s 01 are advancing rapidly for labor, service, and household tasks. Chinese company Unitree impresses with its agile, backflipping H1, while military applications are expanding globally. 

Open-source projects like ROS are also accelerating innovation, making robotics more accessible. This year could mark the mainstream adoption of robots in daily life.

WATCH THE VIDEO ON YOUTUBE

Musk to Unveils Grok 3 Monday

Elon Musk announced that xAI will launch its Grok 3 chatbot on Monday, calling it the "smartest AI on Earth." Grok 3 uses synthetic data to learn from errors, aiming to rival OpenAI’s ChatGPT. The launch comes as competition in the AI market intensifies, with Chinese startup DeepSeek recently gaining traction. xAI has secured $6 billion in funding, positioning itself as a key player in the AI industry.

Meta Enters Humanoid Robotics Race

Meta Platforms is launching a new division to develop AI-powered humanoid robots for household tasks, leveraging its Llama AI platform. Led by Marc Whitten, the initiative aims to enhance robotics with advanced AI capabilities. Meta plans to build its own hardware while partnering with robotics firms like Unitree and Figure AI. The move marks Meta’s significant investment in robotics amid growing competition from Tesla and other AI-driven companies.

UK Tests Claude for Public Services

The UK government is partnering with AI startup Anthropic to explore using its chatbot, Claude, for public services. This initiative aims to simplify access to information and boost efficiency. Backed by Google and Amazon, Anthropic has already helped the European Parliament streamline document searches. The move aligns with Prime Minister Keir Starmer’s goal to make the UK a global leader in AI innovation.

Dell Secures $5B xAI Deal

Dell Technologies is close to finalizing a $5 billion deal to supply servers for Elon Musk's AI startup, xAI. The servers, equipped with Nvidia GB200 semiconductors, will support xAI's supercomputer project in Memphis. Delivery is set for this year, as xAI ramps up its AI capabilities. The partnership highlights growing investments in AI infrastructure amid increasing competition in the industry.

Tencent, Baidu Bet on DeepSeek

Tencent's Weixin app is testing DeepSeek's AI model to enhance search capabilities, marking a significant move in China’s AI race. The integration focuses on domestic users, while Baidu is also connecting its search engine to DeepSeek. The AI startup recently gained attention by surpassing ChatGPT in downloads with its cost-efficient model, prompting companies like Great Wall Motor to adopt its technology.

🧠RESEARCH

The Stochastic Parrot on LLM's Shoulder: A Summative Assessment of Physical Concept Understanding

The study tests if large language models (LLMs) truly understand physical concepts using the PhysiCo benchmark. Results show LLMs, including GPT-4o, perform 40% worse than humans, excelling in language tasks but struggling with abstract grid-based patterns, highlighting their limited comprehension beyond memorization.

InfiniteHiP: Extending Language Model Context Up to 3 Million Tokens on a Single GPU

InfiniteHiP introduces a new framework to handle long contexts in large language models (LLMs) efficiently. It prunes irrelevant tokens, adjusts attention patterns, and offloads memory, enabling processing of up to 3 million tokens on a single GPU—achieving 18.95x faster decoding without extra training or context loss.

Skrr: Skip and Re-use Text Encoder Layers for Memory Efficient Text-to-Image Generation

The Skrr technique improves memory efficiency in text-to-image (T2I) models by skipping and reusing text encoder layers. It reduces memory use without sacrificing image quality, outperforming existing pruning methods while maintaining performance across key evaluation metrics like FID, CLIP, DreamSim, and GenEval.

EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents

EmbodiedBench introduces a benchmark for evaluating vision-driven embodied agents using 1,128 tasks across various environments. Tests show MLLMs perform well in high-level tasks but struggle with low-level actions, with GPT-4o scoring only 28.9%. The platform aims to advance research in MLLM-based agents.

SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models

SelfCite introduces a self-supervised method for sentence-level citations in LLMs. It uses context ablation to assess citation necessity and sufficiency, improving citation quality without extensive human annotation. Tests show a citation F1 improvement of up to 5.3 points on the LongBench-Cite benchmark.

🛠️TOP TOOLS

Gamma AI - AI-powered tool designed to revolutionize the creation of presentations, documents, and webpages.

Haiper AI - Designed to simplify and enhance the process of creating high-quality visual content, including videos and images.

Kolors Virtual Try-On - AI-powered tool designed to transform the way users explore and interact with fashion.

FaceCheck ID - AI-powered tool designed to enhance online safety and security through advanced facial recognition technology.

Doctrina AI - AI education assistant designed to revolutionize the way teaching and learning are approached.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Elon Musk’s xAI seeks $10 billion in funding, potentially valuing the company at $75 billion, to support AI development, including Grok models, with plans to purchase $5 billion in Dell servers.

• DeepSeek founder Liang Wenfeng will meet with Chinese leaders, including Xi Jinping, signaling potential government support for private-sector growth as DeepSeek's AI models gain global attention, challenging U.S. dominance in AI technology.

• Meta AI, in collaboration with Spanish researchers, decoded brain activity to reconstruct typed sentences with 80% accuracy. This breakthrough could advance neuroprostheses and improve communication for individuals with brain-related impairments.

• Apple plans to add AI features and a spatial content app to its Vision Pro headset, aiming to boost sales after initial enthusiasm waned. The update may launch as early as April.

• OpenAI is updating ChatGPT to promote intellectual freedom by offering diverse perspectives on controversial topics. The AI will avoid editorial bias, aiming to deliver neutral, context-rich answers while maintaining safety standards.

• Nous Research launched DeepHermes-3, an AI model with toggleable reasoning mode for deeper, more accurate responses. Users can switch between intuitive and structured reasoning, supporting diverse tasks like coding, math, and creative writing.

• A study found people struggle to distinguish ChatGPT's therapeutic responses from human therapists, with AI often rated more empathetic. While promising, researchers stress careful monitoring and responsible integration into mental health care.

• Researchers at Shanghai Jiao Tong University found LLMs can learn reasoning tasks with minimal data using curated examples. Their "LIMO" method achieved high accuracy, suggesting enterprises can train specialized models efficiently with fewer resources.