AI & Robotics

Asad Norouzi

Robotics, AI, and computer vision, explained for builders, researchers, and the technically curious.

Episodes

  1. 6d ago

    Code vs. Concrete: Why Your AI Can't Fold a Towel

    AI can write a flawless Shakespearean sonnet in two seconds, but ask it to fold a laundry towel and the entire system violently crashes. Why? In this episode, we cut through the Silicon Valley hype machine and dive into the brutal math of the physical world. While the tech industry pours billions into chatbots and generative software, we explore the massive 80% failure rate of enterprise AI and the terrifying reality of the "Sim-to-Real Gap." Join us as we break down why the true future of artificial general intelligence (AGI) won't emerge from a disembodied server farm, but from robots learning how to fail, fall, and feel gravity. If you want to know what happens when code actually hits concrete, this is the ultimate reality check. In this episode, we unpack: The Sim-to-Real Gap: Why robotic brains that ace digital simulations instantly crash when faced with real-world physics and unexpected friction. Vision-Language-Action (VLA) Models: How mapping pixels and text to actual physical motor torque is revolutionizing machine learning. The Hardware Bottleneck: Why the smartest software in the world is useless if it shatters a titanium arm on a desk, and how bio-hybrid designs and optical tactile sensors (like GelSight) are saving the day. Embodied Cognition Theory: The philosophical shift proving that true intelligence requires physical consequence. Listen now to discover why the most profound revolution in technology isn't happening behind a screen. It's learning how to pick up a box. (Don't forget to follow, rate 5 stars, and share this episode with the tech optimist in your life!)

    Code vs. Concrete: Why Your AI Can't Fold a Towel
  2. Jul 22

    Embodied: The Brutal Physics of Robotic AI

    Digital AI is plateauing, but physical robotics is just having its "ChatGPT moment." Why is the tech industry suddenly desperate to stuff fragile, hallucinating software into 200-pound heavy metal bodies? In this deep dive, we cut through the sci-fi hype to uncover the brutal reality of "Embodied AI." We explore why teaching an AI to walk is a mathematical nightmare and how the physical world acts as the ultimate, unforgiving fact-checker for language models. From the engineering nightmare of the "mass penalty spiral" to the explosive dangers of the "reflected inertia trap," we break down the exact physics keeping robots from taking over your kitchen for now. In this episode, we cover: The Software Failure Rate: Why 80% of corporate AI projects are failing and how physical embodiment is the proposed cure. The Walking Problem: The brutal math of neural scaling laws and why it takes 24 times the data to make a robot move. The $16,000 Humanoid: A look at the sudden democratization of hardware (like the Unitree G1) and how university labs are replacing massive corporate budgets. Emergent Capabilities: The haunting reality of what happens when a physical machine suddenly figures out a complex physical task entirely on its own. Grab your headphones and join us as we explore the exact intersection where digital brains crash headfirst into the laws of physics. #ArtificialIntelligence #Robotics #EmbodiedAI #TechNews #Engineering #FutureOfWork #MachineLearning

    Embodied: The Brutal Physics of Robotic AI

About

Robotics, AI, and computer vision, explained for builders, researchers, and the technically curious.