- The Renaissance Times
- Posts
- NVIDIA's Future: Robots and Bigger Chips
NVIDIA's Future: Robots and Bigger Chips
Dear Artisan,
This week is the Super Bowl for residents of San Francisco.
NVIDIA is hosting its annual GTC conference over the coming days, where it will share its plans for the future of the company and the industry itself.
CEO Jensen Huang kicked things off with his keynote yesterday, and the message was clear: the future is robots.
Let’s dive in.
Hot Off The Press
NVIDIA Announces GR00T, a General Purpose Foundation Model for Robotics
NVIDIA Announces Blackwell Chips that are 30x Faster Than Current State-of-the Art Chips
Apple Releases MM1, A Multi-Modal Foundation Model for Text and Images
YouTube to Start Labeling AI-Generated Content Tag on Videos
NVIDIA Released NIM software technology, which packages optimized inference engines and support for AI models into containers for easy deployment
The One Big Thing
What is the one big thing?
Definitely Not a Chip for Ants
Look at the man pictured above. The fate of humanity is in this man’s hands.
I’m only sort of joking. After watching the first day of NVIDIA’s GTC conference, it’s easy to get the feeling that this company is on the path to invincibility. If AGI is the tool enabling the next industrial revolution, NVIDIA’s chips are the steel enabling the revolution to take place.
The Basics:
Blackwell: a significant advancement in GPU technology, featuring a dual-chip design to create the biggest AI chip ever
GR00T: a general-purpose foundation model for humanoid robots
Hyperion 8: a production-ready platform that includes sensors, compute and software needed for AV development
NIM: a new software platform designed to streamline the deployment of custom and pre-trained AI models into production environments
CEO Jensen Huang’s keynote shows that this is just the beginning for this company. Chips and compute are the infrastructure any AI company needs to build on, and NVIDIA has all of it.
Now, it has come time to expand their reach and broaden the stack within which they operate. NVIDIA’s chips are already used in most vehicles and used to power robotics across the world. They are pushing the boundaries further by creating their own robotics lab aimed at achieving AGI for robots. OpenAI, but for the physical world.
It’s hard to overstate how big of a deal this is. If there is one group in the world with the resources and capabilities to create a world model of the entire physical world (which is what an AGI would require), it will need so much compute that essentially every other company in the world is ruled out.
NVIDIA has one of the biggest competitive advantages of all time with its chip lead: if they get far enough along towards developing robust models for robotics, it will be impossible for anyone to come close to catching up, OpenAI included.
This is awe inspiring:
Today is the beginning of our moonshot to solve embodied AGI in the physical world. I’m so excited to announce Project GR00T, our new initiative to create a general-purpose foundation model for humanoid robot learning.
The GR00T model will enable a robot to understand multimodal… twitter.com/i/web/status/1…
— Jim Fan (@DrJimFan)
10:54 PM • Mar 18, 2024
The Gallery
Sora Videos Continue to Impress
Sora.
How wonderful.
9 new clips:
1)
— Proper 🧐 (@ProperPrompter)
3:16 PM • Mar 18, 2024
Sam Altman on Lex Fridman Podcast
Apple and Google Meeting in Secret
Photo evidence of Tim Apple cheating on Siri.
— Harrison Kinsley (@Sentdex)
5:16 PM • Mar 18, 2024
Tools
Must-have tools for every artisan to add to their toolkit:
DGX Cloud: a new service built with NVIDIA to easily train LLMs and image generation models on NVIDIA H100
HuggingChat Assistants with Internet Access
Pipo: AI dubbing over video
APISR: Anime Production Inspired Real-World Anime Super-Resolution
Deep Tech
The newest and coolest in the research world that you need to know about:
HumanoidBench: the first-of-its-kind simulated humanoid benchmark with 27 distinct whole-body tasks
Stable Video 3D: a generative model based on Stable Video Diffusion
NVIDIA Blackwell Specs and Deep Dive
Anticipatory Music Transformer: A 780M parameter model, trained on a larger corpus of music: Lakh + MetaMIDI
Synthesizer v2 in DSPy: Generate diverse quality synthetic data based on DSPy Signature
Closing Thought
If 2024 is the year of robots, what can 2025 possibly bring?
Work With Us!
The AI Renaissance is coming and we are building the best community of the people making it happen.
Contact us to sponsor your product or brand and reach the exact audience for your needs across our newsletter and podcast network.