Good morning. It’s Wednesday, January 8th.

Did you know: On this day in 1982, AT&T agreed to give up control of the Bell Operating Companies that provided local telephone service in the US, to settle the government's anti-trust lawsuit.

In today’s email:

• NVIDIA’s Massive Announcements at CES

• 3 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Nvidia’s Ambitions Stretch Far Beyond GPUs - Here are some of NVIDIA’s biggest announcements at CES Las Vegas:

GeForce RTX 50-Series GPUs

Nvidia CEO Jensen Huang wasted no time flexing the GeForce RTX 50 Series, headlined by the $1,999 RTX 5090 and $999 RTX 5080.

Nvidia’s GeForce RTX 50 series isn’t just about better graphics—it’s about AI running the show. With the RTX 5090 and 5080, machine learning is now baked directly into rendering processes. RTX Neural Shaders transform how we think about textures, lighting, and materials, while RTX Neural Faces and RTX Mega Geometry push character and world-building realism to new heights.

NVIDIA Cosmos

Nvidia’s Cosmos, is an AI platform set to redefine robotics and autonomous vehicle development. Built to tackle the complexities of real-world training, Cosmos uses world foundation models (WFMs) to generate precise video simulations, trained on 20 million hours of real-world data.

Available as open-source software, Cosmos supports synthetic data generation, reducing the need for massive datasets. Experts, including Nvidia's CEO, say the platform presents a ‘ChatGPT moment in robotics,’ but challenges such as cost, regulation, and safety remain. Early adopters include Uber and Agility Robotics.

Project DIGITS

Nvidia's "Project Digits" mini PC, offers petaFLOP computing power in a compact form, packing 1,000 times more power than typical laptops. Designed for AI and data science, it enables local processing of AI models, previously reliant on massive data centers. Powered by Nvidia's Grace Blackwell Superchip, it includes 128GB of memory and 4TB SSD storage. Researchers in fields like robotics can run large AI models, with two units handling up to 405 billion parameters. Slated for a $3,000 release in May, Project Digits is a game-changer for developers, data scientists, and educators.

Automotive and Robotics Initiatives

Nvidia’s Isaac GR00T blueprint, revealed at CES 2025, aims to accelerate humanoid robotics development by streamlining synthetic motion data generation. The blueprint enables developers to generate vast datasets from minimal human demonstrations using Nvidia’s Omniverse and Cosmos platforms.

Isaac GR00T workflows, including GR00T-Teleop, GR00T-Mimic, and GR00T-Gen, help robots learn through imitation learning, reducing the need for costly real-world data collection. With support from major robotics companies like Boston Dynamics, this initiative targets a $38 billion humanoid robot market. Early access to Nvidia’s humanoid developer program is now available.

AI Agents

Nvidia has launched "Blueprints," pre-built templates designed to simplify the creation of autonomous AI agents. These "launchables" include everything needed for rapid development of AI systems without starting from scratch. Companies like CrewAI, Daily, and LangChain have already created specialized blueprints for tasks like code documentation and language processing. Nvidia’s own blueprints include tools for converting PDFs to podcasts and conducting video analysis 30 times faster than real-time. Powered by Nvidia's Nemotron models, based on Meta's Llama technology, these solutions require Nvidia’s AI Enterprise software to operate across various data centers and cloud platforms.

• Google's DeepMind is recruiting AI researchers to advance world model development

• New o1 benchmark raises more questions about AI benchmarking in general

• Robots can now walk through muddy and slippery terrain, thanks to moose-like feet

• Microsoft plans $3 billion AI investment in India, Nadella says

• OpenAI's "Operator" agent for computer use could launch this month

• Meta proposes new scalable memory layers that improve knowledge, reduce hallucinations

• Nvidia CEO says his AI chips are improving faster than Moore's Law

• Meta’s Fact-Checking Partners Say They Were ‘Blindsided’ by Decision to Axe Them

• New AI brain boosts Tesla Cybertruck range by 20%, battery life by 40%

• Grok on Teslas

• Anthropic nears $2 billion investment, pushing valuation to $60 billion 

• Nvidia’s value is now worth more than AMD, Arm, Broadcom, and Intel combined—and doubled

• A scalable framework for learning the geometry-dependent solution operators of partial differential equations

• Memphis warns it may not be able to power Elon Musk’s lofty 'Colossus' supercomputer expansion plans

• Inflection point: ChatGPT was used to plan Las Vegas Cybertruck explosion

📄 Test-time Computing: from System-1 Thinking to System-2 Thinking  

📄 Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction ​​

📄 BoostStep: Boosting mathematical capability of Large Language Models via improved single-step reasoning

📄 METAGENE-1: Metagenomic Foundation Model for Pandemic Monitoring

📄 MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on 𝕏!