Alex on AI
@alexatx.bsky.social
Everything is linked to everything else.
We are just connecting the dots.
Sharing news, insights and perspectives on Artificial Intelligence
We are just connecting the dots.
Sharing news, insights and perspectives on Artificial Intelligence
when Lisboa-Madrid TGV?
January 18, 2025 at 12:23 PM
when Lisboa-Madrid TGV?
🔥 FINAL THOUGHTS 🔥
Jensen didn’t just talk about the future—he showed us the AI-powered present.
The token economy, GPU breakthroughs, AI agents, and digital twins are already reshaping our world.
Which innovation excites you the most? 🚀
Comment below! 👇
Jensen didn’t just talk about the future—he showed us the AI-powered present.
The token economy, GPU breakthroughs, AI agents, and digital twins are already reshaping our world.
Which innovation excites you the most? 🚀
Comment below! 👇
January 8, 2025 at 8:09 AM
🔥 FINAL THOUGHTS 🔥
Jensen didn’t just talk about the future—he showed us the AI-powered present.
The token economy, GPU breakthroughs, AI agents, and digital twins are already reshaping our world.
Which innovation excites you the most? 🚀
Comment below! 👇
Jensen didn’t just talk about the future—he showed us the AI-powered present.
The token economy, GPU breakthroughs, AI agents, and digital twins are already reshaping our world.
Which innovation excites you the most? 🚀
Comment below! 👇
1️⃣1️⃣ NVIDIA’S BIG VISION: AN AI-DRIVEN WORLD
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
January 8, 2025 at 8:09 AM
1️⃣1️⃣ NVIDIA’S BIG VISION: AN AI-DRIVEN WORLD
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
🔟 PROJECT DIGITS: THE PERSONAL AI SUPERCOMPUTER
Imagine a cloud-connected AI supercomputer on your desk:
Powered by GB110 chips.
Runs the full NVIDIA AI stack.
Supports generative AI natively.
AI computing isn’t exclusive to data centers anymore—it’s personal.
Imagine a cloud-connected AI supercomputer on your desk:
Powered by GB110 chips.
Runs the full NVIDIA AI stack.
Supports generative AI natively.
AI computing isn’t exclusive to data centers anymore—it’s personal.
January 8, 2025 at 8:09 AM
🔟 PROJECT DIGITS: THE PERSONAL AI SUPERCOMPUTER
Imagine a cloud-connected AI supercomputer on your desk:
Powered by GB110 chips.
Runs the full NVIDIA AI stack.
Supports generative AI natively.
AI computing isn’t exclusive to data centers anymore—it’s personal.
Imagine a cloud-connected AI supercomputer on your desk:
Powered by GB110 chips.
Runs the full NVIDIA AI stack.
Supports generative AI natively.
AI computing isn’t exclusive to data centers anymore—it’s personal.
9️⃣ GENERAL ROBOTICS: THE NEXT AI FRONTIER
The ChatGPT moment for humanoid robots is near.
NVIDIA’s Isaac Groot uses:
Synthetic motion generation.
Imitation learning from human demonstrations.
Physics-based AI training.
The robotics industry is about to explode.
The ChatGPT moment for humanoid robots is near.
NVIDIA’s Isaac Groot uses:
Synthetic motion generation.
Imitation learning from human demonstrations.
Physics-based AI training.
The robotics industry is about to explode.
January 8, 2025 at 8:09 AM
9️⃣ GENERAL ROBOTICS: THE NEXT AI FRONTIER
The ChatGPT moment for humanoid robots is near.
NVIDIA’s Isaac Groot uses:
Synthetic motion generation.
Imitation learning from human demonstrations.
Physics-based AI training.
The robotics industry is about to explode.
The ChatGPT moment for humanoid robots is near.
NVIDIA’s Isaac Groot uses:
Synthetic motion generation.
Imitation learning from human demonstrations.
Physics-based AI training.
The robotics industry is about to explode.
8️⃣ AUTONOMOUS VEHICLES: THE THREE NVIDIA COMPUTERS
AVs require NVIDIA’s AI trifecta:
DGX: Train AI models.
Omniverse: Test & simulate.
Drive AGX: AI supercomputer in the car.
Synthetic data scales training from thousands of drives to billions of miles.
AVs require NVIDIA’s AI trifecta:
DGX: Train AI models.
Omniverse: Test & simulate.
Drive AGX: AI supercomputer in the car.
Synthetic data scales training from thousands of drives to billions of miles.
January 8, 2025 at 8:09 AM
8️⃣ AUTONOMOUS VEHICLES: THE THREE NVIDIA COMPUTERS
AVs require NVIDIA’s AI trifecta:
DGX: Train AI models.
Omniverse: Test & simulate.
Drive AGX: AI supercomputer in the car.
Synthetic data scales training from thousands of drives to billions of miles.
AVs require NVIDIA’s AI trifecta:
DGX: Train AI models.
Omniverse: Test & simulate.
Drive AGX: AI supercomputer in the car.
Synthetic data scales training from thousands of drives to billions of miles.
7️⃣ AI AGENTS: THE NEW DIGITAL WORKFORCE
AI agents are the next step:
NIMs: Pre-packaged AI microservices.
NEMO: Digital onboarding for AI agents.
Blueprints: Custom AI models for enterprise use.
Your IT department won’t just manage software—they’ll manage AI employees.
AI agents are the next step:
NIMs: Pre-packaged AI microservices.
NEMO: Digital onboarding for AI agents.
Blueprints: Custom AI models for enterprise use.
Your IT department won’t just manage software—they’ll manage AI employees.
January 8, 2025 at 8:09 AM
7️⃣ AI AGENTS: THE NEW DIGITAL WORKFORCE
AI agents are the next step:
NIMs: Pre-packaged AI microservices.
NEMO: Digital onboarding for AI agents.
Blueprints: Custom AI models for enterprise use.
Your IT department won’t just manage software—they’ll manage AI employees.
AI agents are the next step:
NIMs: Pre-packaged AI microservices.
NEMO: Digital onboarding for AI agents.
Blueprints: Custom AI models for enterprise use.
Your IT department won’t just manage software—they’ll manage AI employees.
6️⃣ OMNIVERSE + COSMOS: A MULTIVERSE FOR AI
Omniverse+Cosmos=A physics-grounded multiverse generator.
It creates endless synthetic scenarios, allowing AI to:
Learn real-world physics.
Simulate complex environments.
Make grounded, reliable predictions.
The digital twin revolution is here.
Omniverse+Cosmos=A physics-grounded multiverse generator.
It creates endless synthetic scenarios, allowing AI to:
Learn real-world physics.
Simulate complex environments.
Make grounded, reliable predictions.
The digital twin revolution is here.
January 8, 2025 at 8:09 AM
6️⃣ OMNIVERSE + COSMOS: A MULTIVERSE FOR AI
Omniverse+Cosmos=A physics-grounded multiverse generator.
It creates endless synthetic scenarios, allowing AI to:
Learn real-world physics.
Simulate complex environments.
Make grounded, reliable predictions.
The digital twin revolution is here.
Omniverse+Cosmos=A physics-grounded multiverse generator.
It creates endless synthetic scenarios, allowing AI to:
Learn real-world physics.
Simulate complex environments.
Make grounded, reliable predictions.
The digital twin revolution is here.
5️⃣ NVIDIA COSMOS: TEACHING AI TO UNDERSTAND THE WORLD
NVIDIA Cosmos is a world foundation model.
It’s trained on 20 million hours of physical interactions—gravity, inertia, object permanence.
This isn’t just for simulation—it’s the foundation for robotics, AVs, and synthetic data generation.
NVIDIA Cosmos is a world foundation model.
It’s trained on 20 million hours of physical interactions—gravity, inertia, object permanence.
This isn’t just for simulation—it’s the foundation for robotics, AVs, and synthetic data generation.
January 8, 2025 at 8:09 AM
5️⃣ NVIDIA COSMOS: TEACHING AI TO UNDERSTAND THE WORLD
NVIDIA Cosmos is a world foundation model.
It’s trained on 20 million hours of physical interactions—gravity, inertia, object permanence.
This isn’t just for simulation—it’s the foundation for robotics, AVs, and synthetic data generation.
NVIDIA Cosmos is a world foundation model.
It’s trained on 20 million hours of physical interactions—gravity, inertia, object permanence.
This isn’t just for simulation—it’s the foundation for robotics, AVs, and synthetic data generation.
4️⃣ AI IS EVERYWHERE: FROM DATA CENTERS TO DESKTOPS
AI started in the cloud, but now it’s coming home.
With Windows WSL2 and NVIDIA GPUs, your PC becomes an AI powerhouse.
Generative APIs will power everything—language, sound, and 3D rendering—right from your desktop.
AI started in the cloud, but now it’s coming home.
With Windows WSL2 and NVIDIA GPUs, your PC becomes an AI powerhouse.
Generative APIs will power everything—language, sound, and 3D rendering—right from your desktop.
January 8, 2025 at 8:09 AM
4️⃣ AI IS EVERYWHERE: FROM DATA CENTERS TO DESKTOPS
AI started in the cloud, but now it’s coming home.
With Windows WSL2 and NVIDIA GPUs, your PC becomes an AI powerhouse.
Generative APIs will power everything—language, sound, and 3D rendering—right from your desktop.
AI started in the cloud, but now it’s coming home.
With Windows WSL2 and NVIDIA GPUs, your PC becomes an AI powerhouse.
Generative APIs will power everything—language, sound, and 3D rendering—right from your desktop.
3️⃣ SCALING LAWS: THE SECRET TO AI’S EXPLOSIVE GROWTH
Three scaling laws drive AI:
Pre-training Scaling: More data, bigger models, more compute.
Post-training Scaling: Fine-tuned with reinforcement learning.
Test-time Scaling: Real-time reasoning
The result? Smarter AI at every stage.
Three scaling laws drive AI:
Pre-training Scaling: More data, bigger models, more compute.
Post-training Scaling: Fine-tuned with reinforcement learning.
Test-time Scaling: Real-time reasoning
The result? Smarter AI at every stage.
January 8, 2025 at 8:09 AM
3️⃣ SCALING LAWS: THE SECRET TO AI’S EXPLOSIVE GROWTH
Three scaling laws drive AI:
Pre-training Scaling: More data, bigger models, more compute.
Post-training Scaling: Fine-tuned with reinforcement learning.
Test-time Scaling: Real-time reasoning
The result? Smarter AI at every stage.
Three scaling laws drive AI:
Pre-training Scaling: More data, bigger models, more compute.
Post-training Scaling: Fine-tuned with reinforcement learning.
Test-time Scaling: Real-time reasoning
The result? Smarter AI at every stage.
2️⃣ RTX BLACKWELL: THE AI SUPERCHIP
The RTX 50 series isn’t just a GPU—it’s a computational marvel:
92 billion transistors
4 petaflops of AI performance
Neural material shading
AI doesn’t just use these GPUs—it powers them.
This is the AI-GPU feedback loop in action.
The RTX 50 series isn’t just a GPU—it’s a computational marvel:
92 billion transistors
4 petaflops of AI performance
Neural material shading
AI doesn’t just use these GPUs—it powers them.
This is the AI-GPU feedback loop in action.
January 8, 2025 at 8:09 AM
2️⃣ RTX BLACKWELL: THE AI SUPERCHIP
The RTX 50 series isn’t just a GPU—it’s a computational marvel:
92 billion transistors
4 petaflops of AI performance
Neural material shading
AI doesn’t just use these GPUs—it powers them.
This is the AI-GPU feedback loop in action.
The RTX 50 series isn’t just a GPU—it’s a computational marvel:
92 billion transistors
4 petaflops of AI performance
Neural material shading
AI doesn’t just use these GPUs—it powers them.
This is the AI-GPU feedback loop in action.
1️⃣1️⃣ NVIDIA’S BIG VISION: AN AI-DRIVEN WORLD
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
January 8, 2025 at 8:09 AM
1️⃣1️⃣ NVIDIA’S BIG VISION: AN AI-DRIVEN WORLD
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
Every token, every pixel, every prediction—it’s all powered by NVIDIA’s AI stack.
From autonomous cars to digital twins to AI agents, NVIDIA isn’t building the future.
They’re scaling the present.
You can Apply here:
openai.com/index/early-...
openai.com/index/early-...
December 21, 2024 at 9:27 PM
You can Apply here:
openai.com/index/early-...
openai.com/index/early-...
Mark your calendars: O3 Mini is slated for release around late January, and full O3 will follow soon after. The AI future is unfolding fast! #AI #StayTuned
December 21, 2024 at 9:25 PM
Mark your calendars: O3 Mini is slated for release around late January, and full O3 will follow soon after. The AI future is unfolding fast! #AI #StayTuned
Public Access for Security & Safety testing is rolling out now—applications close on January 10. Help OpenAI find and fix vulnerabilities before the official release! #AI #CyberSecurity
December 21, 2024 at 9:25 PM
Public Access for Security & Safety testing is rolling out now—applications close on January 10. Help OpenAI find and fix vulnerabilities before the official release! #AI #CyberSecurity
O3 Mini also supports function calling and structured outputs, giving developers the advanced capabilities they’ve been asking for—at a lower cost. #Developers #OpenAI
December 21, 2024 at 9:25 PM
O3 Mini also supports function calling and structured outputs, giving developers the advanced capabilities they’ve been asking for—at a lower cost. #Developers #OpenAI
In math, O3 Mini hits comparable or better performance than O1 but with dramatically reduced latency, bringing near-instant responses. #AI #Performance
December 21, 2024 at 9:25 PM
In math, O3 Mini hits comparable or better performance than O1 but with dramatically reduced latency, bringing near-instant responses. #AI #Performance
Efficiency milestone! O3 Mini delivers higher coding ELO than O1 Mini at a fraction of the cost. Perfect for large-scale code generation tasks. #AI #CostEfficiency
December 21, 2024 at 9:25 PM
Efficiency milestone! O3 Mini delivers higher coding ELO than O1 Mini at a fraction of the cost. Perfect for large-scale code generation tasks. #AI #CostEfficiency
O3 Mini supports “adaptive thinking time,” letting you choose low, medium, or high reasoning effort based on your use case—striking a new balance between speed and accuracy. #AI #Innovation
December 21, 2024 at 9:25 PM
O3 Mini supports “adaptive thinking time,” letting you choose low, medium, or high reasoning effort based on your use case—striking a new balance between speed and accuracy. #AI #Innovation