Today:

China's Humanoid Robot JUST SHATTERED World Record! $90,000 Model Begins Deployment Across US and EU

This new humanoid robot just broke the speed record, zipping around at 3.3 meters per second. That's as fast as someone hustling past you, making you do a double-take. Not just fast, it's got moves too, busting out TikTok dances with the kind of coordination and power that's seriously impressive. 

This robot's not just about speed; it's decked out with tech for navigating and carrying stuff, aiming to be a game-changer in rescue operations and more. Plus, there's a robot dog that's fast and strong, perfect for industrial and emergency scenarios. This tech is fresh, and it's pushing the envelope on what robots can do in our world.

WATCH THE VIDEO ON YOUTUBE

Copilot for OneDrive will fetch your files and summarize them

Microsoft is gearing up to launch Copilot for OneDrive, set to debut in late April. This tool acts as a personal research assistant, capable of summarizing and extracting info from various file types, including text docs, presentations, and PDFs. Users can customize summaries to their preference, focusing on key points. 

Copilot also responds to natural language queries, offering tailored responses and assisting in tasks like creating outlines and lists. Early access begins soon for OneDrive users, with broader availability slated for late April for those with a Copilot for Microsoft 365 license.

THE VERGE

Nvidia CEO says AI could pass human tests in five years

NVIDIA's CEO, Jensen Huang, predicts artificial general intelligence (AGI) could match human capabilities within five years. He clarified that this timeline hinges on defining AGI achievement as passing human tests. Huang noted AI's current ability to pass exams like the legal bar and expects AI to excel in medical tests soon. 

However, he acknowledged differing views on AGI's definition, which complicates engineering goals. Huang also addressed the need for more chip factories to support AI expansion, emphasizing ongoing improvements in computing efficiency. This insight emerged during an economic forum at Stanford University, reflecting NVIDIA's pivotal role in AI advancement.

REUTERS

OpenAI execs reject Elon Musk's claims

OpenAI's top brass rebuffed Elon Musk's allegations in a memo, asserting independence and dedication to humanity's benefit. They denied achieving Artificial General Intelligence (AGI) and refuted claims of being a Microsoft subsidiary. CEO Sam Altman acknowledged challenges ahead, including government scrutiny post-Altman's reinstatement. Rebuttals focused on GPT-4's limitations compared to AGI, commitment to societal benefits, and autonomous decision-making. 

Meanwhile, the SEC probes OpenAI's internal communications for potential investor misinformation during a past crisis. With OpenAI valued over $80 billion, its governance attracts antitrust scrutiny. The saga highlights tensions in AI development and corporate governance.

AXIOS

China’s bureaucrats turn to AI, handwriting robots to cope with official paperwork

Chinese bureaucrats are adopting AI and robotic tools to handle paperwork, despite government efforts to reduce bureaucracy. These officials face a heavy workload, including handwritten reports, prompting some to utilize technology for writing tasks. 

While these tools ease mundane tasks, they lack training on current information. Some officials have been exposed for using uniformly suspicious or suddenly improved handwriting. This shift highlights a tech-savvy approach to administrative challenges, reflecting the ongoing tension between traditional bureaucracy and modern efficiency in China.

SCMP

🧠RESEARCH

MOSAIC: A Modular System for Assistive and Interactive Cooking

MOSAIC is a cutting-edge home robot system designed for cooking alongside humans. It blends natural language interaction, coordination of multiple robots, and an extensive knowledge of everyday objects through modularity. MOSAIC integrates large-scale pre-trained models for broad tasks and specific modules for precise control. Tested in 60 real-world cooking trials with humans, achieving a 68.3% success rate across six recipes, it highlights both achievements and future challenges in robotic collaboration.

Trajectory Consistency Distillation

Trajectory Consistency Distillation (TCD) is an advanced technique improving text-to-image synthesis by addressing clarity and detail issues found in existing models. By analyzing and identifying errors in latent space consistency, TCD introduces a trajectory consistency function and strategic stochastic sampling to enhance image quality and detail. This method not only improves image generation at low computational costs but also surpasses teacher models in producing more detailed images at higher costs, showcasing significant advancements in image synthesis technology. 

ViewFusion: Towards Multi-View Consistency via Interpolated Denoising

ViewFusion introduces a novel, no-extra-training-needed algorithm that brings consistency to images created from multiple viewpoints. It smartly uses previous images as context for generating new ones, ensuring that all views match up seamlessly. This method enhances diffusion models, enabling them to produce more coherent and detailed scenes from several angles without the need for additional tweaking. This breakthrough demonstrates significant improvements in generating realistic multi-view images.

Simple linear attention language models balance the recall-throughput tradeoff

BASED, a novel architecture that balances the trade-off between recall ability and memory efficiency in language models. BASED integrates linear and sliding window attention to navigate the trade-off curve effectively, achieving high recall with manageable memory usage. It outperforms comparable models in perplexity and real-world tasks by significant margins. The approach also includes IO-aware algorithms boosting language generation throughput by 24x compared to existing solutions. 

DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models

"DistriFusion" is a groundbreaking method that significantly speeds up the creation of high-resolution images using diffusion models, without sacrificing quality. By dividing the input into patches and distributing them across multiple GPUs, it cleverly overcomes the computational challenges. A unique approach called displaced patch parallelism allows for efficient communication between patches, leading to up to 6.1 times faster processing on eight NVIDIA A100 GPUs. The technique promises great potential for real-time, high-quality image generation applications. 

🛠️TOP TOOLS

Wondera - Sing One Song to Create Your AI Voice.

Meals.chat - Offers a Telegram bot for diet tracking by analyzing pictures or descriptions of your meals.

Dart - AI project management tool.

Verb - A new space for writers.

Beehiiv - Newsletter platform designed to boost growth, offering intuitive tools for creating, publishing, and monetizing content.

🗞️MORE NEWS

Here Come the AI Worms

Security researchers developed an AI worm capable of spreading among generative AI agents, potentially compromising data and sending spam. Dubbed Morris II, it poses a new cyberattack threat. By exploiting vulnerabilities in systems like ChatGPT and Gemini, it demonstrates risks in AI ecosystems. Countermeasures and awareness are crucial as such threats evolve. WIRED

Elon Musk vs. Everyone: The New Fight in AI

Elon Musk is gearing up for an AI showdown with tech giants like Microsoft, Google, and OpenAI's Sam Altman. Dubbed the "AI wars," Musk is pushing his nascent xAI against the established players. It's a battle of the titans in the realm of artificial intelligence, with Musk leading the charge against the industry heavyweights. THE WALL STREET JOURNAL

Huawei's AI chip capabilities under intense scrutiny after market leader Nvidia taps it as potential rival

Huawei's AI chip, the Ascend 910B, is turning heads as it rivals Nvidia's A100 GPUs in power, signaling Huawei as a key player in AI tech despite US sanctions. This development challenges Nvidia's dominance, especially with US export restrictions limiting its sales in China. Huawei's move into AI chips, backed by China's top chipmaker, showcases its resilience and the shifting dynamics in the global tech landscape. YAHOO!

AI meets Old Masters in the fight to authenticate paintings

Artificial intelligence (AI) is shaking up the art world, even in traditional areas like Old Masters. At events like the Tefaf fair, AI companies like Art Recognition are showcasing their ability to authenticate artworks, aiding in attributions and boosting prices. However, skepticism remains about AI replacing human expertise entirely. FINANCIAL TIMES

Smart Glasses and AI Badges: All the Ways AI Appeared at MWC 2024

At Mobile World Congress 2024, AI stole the spotlight with wearable assistants and enhanced phone features. The Humane AI Pin offers discreet aid, projecting information via laser, while phones like Xiaomi 14 and Honor Magic 6 Pro showcase on-device AI for tasks like expanding photos and eye-tracking controls. Chipmakers like Qualcomm and MediaTek also debuted improved AI-powered modems for faster, more efficient mobile experiences. CNET

Microsoft Copilot: AI chatbot gives questionable answers on teaching sex, DEI, LGBTQ topics to preschool kids

Microsoft's AI chatbot Copilot has stirred controversy by suggesting it's okay to discuss topics like diversity, equity, inclusion (DEI), transgenderism, and sex education with preschool kids, while deeming discussions on pedophilia and White privilege inappropriate for the age group. The chatbot's responses, varying due to its programming and training data, have sparked debate on what's suitable for early childhood education. Copilot emphasized tailoring discussions to kids' understanding levels and encouraged creating inclusive environments without explicit content. FOX BUSINESS

AI finds several early risk factors to predict Alzheimer’s 7 years early

Researchers at the University of California San Francisco have harnessed AI to pinpoint early risk factors for Alzheimer’s, potentially flagging the disease up to seven years before symptoms emerge. This breakthrough could revolutionize how we approach Alzheimer’s, allowing for earlier interventions and possibly even preventive measures. Interestingly, the study highlighted not only shared risk factors between men and women, such as high cholesterol and vitamin D deficiency but also uncovered sex-specific risks like erectile dysfunction and an enlarged prostate for men, and osteoporosis for women. MEDICAL NEWS TODAY

AI could help replicate smells in danger of being lost to history

AI is revolutionizing the way we recreate and preserve scents, potentially saving fragrances at risk of being lost, like specific incenses or forest smells altered by climate change. Researchers have successfully used AI to replicate existing perfumes by analyzing their chemical makeup and scent intensity. The process, which considers how smells change over time due to evaporation, could make perfume development cheaper and more sustainable. While still needing human testing for validation, this blend of AI, physics, and chemistry marks a significant step forward in fragrance creation and preservation. NEWSCIENTIST

AI programming languages power today’s innovations like ChatGPT. These are some of the most popular

AI's shaping the future, with a $15.7 trillion boost by 2030. Python's king for its simplicity and vast library, crucial for AI gigs. Other hot skills include Java and SQL. Learning's key, with courses and real-world projects. Soft skills like problem-solving and adaptability are gold. AI's rapid changes demand continuous learning and ethical understanding. FORTUNE