10+ hours of podcasts, in 5 minutes.
12 quotes from 1 episode on Practical AI, each with a timestamped link to the source.
Ming-Yu Liu argues that physical AI represents the next major market for hardware because robots interacting with reality run local, instantaneous processing. Creating world models and open software frameworks directly shows NVIDIA how to design future GPU architectures to meet these on-device demands.
Model weights serve as the modern equivalent of software packages that engineers pull down and adapt.
“In the modern world, you have computers and you have libraries still, but you now have models…”
Autonomous hardware evaluates processing efficiency by measuring intelligence delivered per watt of energy consumed.
“In the end, the primary evaluation would be intelligence per watt.”
Humanoid robots create a new market for onboard compute.
“One day if we have humanoid robot surrounding us, they will require computer…”
Cloud applications efficiently process hundreds of aggregated user requests at once. Physical machines operate alone and process commands one at a time.
“A lot of tasks is a batch size one inference problem. So API, you can aggregate users' API call from different users, and then process them in whole.”
Constructing foundation models provides direct feedback on compute limitations. Understanding these constraints shows NVIDIA exactly what kind of GPU architectures developers demand.
“Through building the model, we better understand what kind of computer we should build, what kind of GPU architecture we should have…”
A complete open release provides developers with specific data sets and software frameworks. Engineers use these tools to train base models on proprietary information.
“Open model, it's not just have the model open weight allow you to use. We also provide the training framework, so that you can take this open model and your own data, and post-train to something more tailored for your use case.”
“We even open-source data that we created to help you to use those data if you want to build your own open model.”
Simulated world models generate pixel-level environmental changes. These environments provide autonomous systems with exact control signals to complete physical tasks safely.
“With a world model, instead of having your policy deploy the real car, drive in the real world, you can have your policy interact with the world model…”
From NVIDIA Cosmos: Why Physical AI Needs World Models, Not LLMs
“When you model the world, it generates the pixel space evolution, and the pixel space evolution has a strong correlation to the control signal you might need to use to complete certain manipulation tasks.”
From NVIDIA Cosmos: Why Physical AI Needs World Models, Not LLMs
“There will be applications that were not foreseen by the foundation model frontier companies…”
“If we can make physical AI come, you know, the the dream of physical AI come true, there will be a lot of opportunity for NVIDIA, so we are very aligned…”
“To achieve full automation, we need our robot be smarter, even the robot arm smarter and more, I would say, more configurable…”
How we attribute quotes. Every quote was matched against the episode transcript, so the words and the timestamp are real (we trim filler words like "um", nothing else). The name comes from our written summary of the episode, and we use it only when a separate check of the captions finds that person on the episode. YouTube gives us no voice-by-voice transcript, so open the timestamp to hear who is talking. See a wrong name? Tell us and we fix or remove it.
10+ hours of podcasts, distilled into one 5-minute read. Free, every Sunday.
For now, every subscriber gets both newsletters. No spam. Unsubscribe with one click.