{"author":{"@type":"Person","name":"Alexander Bock","url":"https://abxk.ai/en/autor/alexander-bock/"},"categories":["Physical AI"],"content":{"markdown":"Most AI you use lives on a screen: you type a question, and a language model writes an answer. **Physical AI** leaves the screen. It is AI that senses the real world through cameras, microphones or electrodes and acts in it through wheels, motors or a computer cursor. A robotaxi that drives without a driver, a robot arm that picks parts it has never seen and an implant that turns brain signals into words are all physical AI.\n\nThis is part 1 of Level 0 in our Physical AI learning path. You will learn what physical AI is, the loop every such system runs, the fields this learning path covers and why the physical world is so much harder for AI than a chat window. You do not need any prior knowledge.\n\n## The loop: sense, decide, act\n\nEvery physical AI system runs the same loop many times per second. It **senses** the world with sensors, **understands** what it sees, for example \"a cyclist, two meters ahead, moving left\", **decides** what to do and **acts** through motors or other actuators. The action changes the world, and the loop starts again with fresh sensor data.\n\n![A loop of four steps around the real world: sense with cameras and other sensors, understand what is there, decide what to do and act with motors; the action changes the world, which is sensed again](/images/learn/en/what-is-physical-ai-loop.webp \"The loop that every physical AI system runs: sense, understand, decide, act.\")\n\nA language model works differently: it gets one input, writes one output and waits. If its answer is wrong, you can ask again. A self-driving car cannot pause the street. It has to decide within fractions of a second, and it cannot take a wrong decision back.\n\n## The four fields in this learning path\n\nPhysical AI covers many machines, from vacuum robots to drones. This learning path follows four fields where AI already does real work, plus the infrastructure behind them, and each has public data you can check yourself.\n\n| Field | Senses | Decides | Acts | Public data on ABXK.AI |\n|---|---|---|---|---|\n| Self-driving vehicles | Cameras, radar, lidar | Where to drive and how fast | Steering, brakes, motor | [Crash reports filed with the US road safety authority](/en/physical-ai/self-driving-car-crash-data/) |\n| Robots | Cameras, force and touch sensors | How to grasp, where to step | Motors in joints and grippers | Notable AI models for robots |\n| Brain-computer interfaces | Electrodes on or in the brain | Which movement or word was intended | Cursor, robot arm, synthetic voice | [Clinical studies in the US trial registry](/en/physical-ai/brain-computer-interface-trials/) |\n| Chips and data centers | — | — | Supply the computing power | [Chips, data centers and power use](/en/physical-ai/ai-data-centers/) |\n\nThe last row is not a machine that acts on its own. It is the physical foundation of all AI: every model, including the ones in robots and cars, is trained in data centers full of chips that draw large amounts of electricity.\n\n## Why the physical world is harder than a chat\n\nPhysical AI is still a small part of AI research. Of the {{\u003c num \"ai_models\" \"notable_since_2020\" \u003e}} notable AI models that Epoch AI, a research institute, recorded since 2020, only {{\u003c num \"ai_models\" \"robotics_share_since_2020_pct\" \u003e}}% were built for robots or vehicles; most were language and image models.\n\n![Horizontal bars: notable AI models since 2020 by domain, with language far ahead and robotics and driving near the bottom](/images/learn/en/what-is-physical-ai-domains.webp \"Notable AI models since 2020 by domain. Physical AI is a small share.\")\n\nThere are four main reasons:\n\n- **Mistakes cost more.** A wrong word in a chat is annoying. A wrong steering decision can hurt people. Physical AI has to be safe in rare situations, not just in typical ones.\n- **There is little data.** Language models learn from text collected from the internet. There is no internet of robot movements: every example of a robot grasping a cup has to be recorded with real hardware or produced in a simulation.\n- **Time is short.** A car at 50 km/h (31 mph) covers almost 14 meters per second. Sensing, understanding and deciding have to fit into fractions of a second, on a computer inside the vehicle.\n- **The world is open-ended.** A street has endless rare situations, called **edge cases**: a mattress on the highway, a police officer waving drivers through a red light. A model cannot have seen them all.\n\n## How physical AI learns\n\nBecause real data is scarce, physical AI combines several ways of learning. Engineers **program** rules where the task is fixed, such as a welding robot on an assembly line. Robots learn from **demonstrations**, when a person guides the robot or steers it remotely and the model copies the movement. They practice in **simulation**, millions of attempts in a virtual world before the first attempt in the real one. Since 2023, **foundation models for robots** have joined these methods: large models that link images, language and movement, so that a robot can follow an instruction such as \"put the red cup in the sink\". {{\u003c num \"ai_models\" \"robotics_since_2022\" \u003e}} of the {{\u003c num \"ai_models\" \"robotics_models\" \u003e}} notable robotics models in Epoch AI's data were published in 2022 or later.\n\n{{\u003c data \"models-robotics\" \"n=8\" \u003e}}\n\n## What to check when you read news about physical AI\n\nPhysical AI attracts bold claims. These questions help you separate demonstrations from real use:\n\n1. **Demo or service?** A video of a robot folding laundry shows one success. Ask how often it works and in how many homes it runs.\n2. **Who is in control?** Many robots and some vehicles are steered remotely by people in difficult moments. That is legitimate, but it matters for what the AI can do alone.\n3. **How much real use?** Miles driven without a driver, hours of operation or patients treated say more than any announcement.\n4. **What do the public records say?** For self-driving cars and brain implants, regulators publish crash reports and trial registrations. This learning path uses them.\n\nThe next parts of Level 0 look at each field in turn, starting with how self-driving cars see the road. For the data behind this learning path, see the overview of [Physical AI](/en/physical-ai/).","text":"Most AI you use lives on a screen: you type a question, and a language model writes an answer. Physical AI leaves the screen. It is AI that senses the real world through cameras, microphones or electrodes and acts in it through wheels, motors or a computer cursor. A robotaxi that drives without a driver, a robot arm that picks parts it has never seen and an implant that turns brain signals into words are all physical AI.\nThis is part 1 of Level 0 in our Physical AI learning path. You will learn what physical AI is, the loop every such system runs, the fields this learning path covers and why the physical world is so much harder for AI than a chat window. You do not need any prior knowledge.\nThe loop: sense, decide, act Every physical AI system runs the same loop many times per second. It senses the world with sensors, understands what it sees, for example \u0026ldquo;a cyclist, two meters ahead, moving left\u0026rdquo;, decides what to do and acts through motors or other actuators. The action changes the world, and the loop starts again with fresh sensor data.\nThe loop that every physical AI system runs: sense, understand, decide, act. A language model works differently: it gets one input, writes one output and waits. If its answer is wrong, you can ask again. A self-driving car cannot pause the street. It has to decide within fractions of a second, and it cannot take a wrong decision back.\nThe four fields in this learning path Physical AI covers many machines, from vacuum robots to drones. This learning path follows four fields where AI already does real work, plus the infrastructure behind them, and each has public data you can check yourself.\nWhat Is Physical AI? Machines That Sense, Decide and Act Field Senses Decides Acts Public data on ABXK.AI Self-driving vehicles Cameras, radar, lidar Where to drive and how fast Steering, brakes, motor Crash reports filed with the US road safety authority Robots Cameras, force and touch sensors How to grasp, where to step Motors in joints and grippers Notable AI models for robots Brain-computer interfaces Electrodes on or in the brain Which movement or word was intended Cursor, robot arm, synthetic voice Clinical studies in the US trial registry Chips and data centers — — Supply the computing power Chips, data centers and power use The last row is not a machine that acts on its own. It is the physical foundation of all AI: every model, including the ones in robots and cars, is trained in data centers full of chips that draw large amounts of electricity.\nWhy the physical world is harder than a chat Physical AI is still a small part of AI research. Of the 628 notable AI models that Epoch AI, a research institute, recorded since 2020, only 2.9% were built for robots or vehicles; most were language and image models.\nNotable AI models since 2020 by domain. Physical AI is a small share. There are four main reasons:\nMistakes cost more. A wrong word in a chat is annoying. A wrong steering decision can hurt people. Physical AI has to be safe in rare situations, not just in typical ones. There is little data. Language models learn from text collected from the internet. There is no internet of robot movements: every example of a robot grasping a cup has to be recorded with real hardware or produced in a simulation. Time is short. A car at 50 km/h (31 mph) covers almost 14 meters per second. Sensing, understanding and deciding have to fit into fractions of a second, on a computer inside the vehicle. The world is open-ended. A street has endless rare situations, called edge cases: a mattress on the highway, a police officer waving drivers through a red light. A model cannot have seen them all. How physical AI learns Because real data is scarce, physical AI combines several ways of learning. Engineers program rules where the task is fixed, such as a welding robot on an assembly line. Robots learn from demonstrations, when a person guides the robot or steers it remotely and the model copies the movement. They practice in simulation, millions of attempts in a virtual world before the first attempt in the real one. Since 2023, foundation models for robots have joined these methods: large models that link images, language and movement, so that a robot can follow an instruction such as \u0026ldquo;put the red cup in the sink\u0026rdquo;. 19 of the 31 notable robotics models in Epoch AI\u0026rsquo;s data were published in 2022 or later.\nNotable AI models for robots and vehicles, newest first ModelOrganizationPublishedTask GEN-1.5GeneralistAugust 19, 2026Grasping and moving objects MolmoAct 2Allen Institute for AIApril 5, 2026Grasping and moving objects π0.6 (pi-0.6)Physical IntelligenceNovember 17, 2025Grasping and moving objects Eagle 2NvidiaJanuary 20, 2025– RDT-1BTsinghua UniversityOctober 10, 2024Grasping and moving objects GR-2ByteDanceOctober 8, 2024Video Table Tennis AgentGoogle DeepMindAugust 7, 2024Sports OpenVLAStanford UniversityJune 13, 2024Grasping and moving objects Octo-BaseUniversity of California (UC) BerkeleyMay 20, 2024Grasping and moving objects ManiGaussianTsinghua UniversityMarch 13, 2024Grasping and moving objects RT-TrajectoryGoogle DeepMindNovember 3, 2023Grasping and moving objects RT-2-XGoogle DeepMindOctober 13, 2023Grasping and moving objects Robot ParkourShanghai Qi Zhi instituteSeptember 12, 2023Imitating movement SwiftIntel LabsAugust 30, 2023Drone flight RT-2Google DeepMindJuly 28, 2023Grasping and moving objects Agile Soccer RobotGoogle DeepMindApril 26, 2023Imitating movement PaLM-EGoogleMarch 6, 2023Questions about images RT-1GoogleDecember 13, 2022Grasping and moving objects GatoDeepMindMay 12, 2022Atari 6-Act TetherFacebook AI ResearchAugust 3, 2021Object detection SemExpCarnegie Mellon University (CMU)July 2, 2020Object detection DD-PPOGeorgia Institute of TechnologyDecember 19, 2019Object detection Walking Minotaur robotUniversity of California (UC) BerkeleyJune 19, 2019Imitating movement MetaMimicGoogleOctober 11, 2018Grasping and moving objects Deep Deterministic Policy GradientsGoogle DeepMindSeptember 9, 2015Grasping and moving objects Truck backer-upperStanford UniversityJune 18, 1989Self-driving Q-learningUniversity of LondonJanuary 1, 1989Route finding ASE+ACEUniversity of Massachusetts AmherstSeptember 1, 1983Pole balancing Heuristic Reinforcement LearningPurdue UniversityOctober 1, 1965System control SNARCHarvard UniversityJanuary 8, 1952Maze solving TheseusBell LaboratoriesJuly 2, 1950Maze solving Source as for the first table on this page.\nWhat to check when you read news about physical AI Physical AI attracts bold claims. These questions help you separate demonstrations from real use:\nDemo or service? A video of a robot folding laundry shows one success. Ask how often it works and in how many homes it runs. Who is in control? Many robots and some vehicles are steered remotely by people in difficult moments. That is legitimate, but it matters for what the AI can do alone. How much real use? Miles driven without a driver, hours of operation or patients treated say more than any announcement. What do the public records say? For self-driving cars and brain implants, regulators publish crash reports and trial registrations. This learning path uses them. The next parts of Level 0 look at each field in turn, starting with how self-driving cars see the road. For the data behind this learning path, see the overview of Physical AI.\n","words":1153},"date":"2026-10-18","description":"Physical AI explained: how machines sense, decide and act in the real world, from robotaxis to robot arms and brain implants, and why it is harder than a chatbot.","image":"https://abxk.ai/images/cover/what-is-physical-ai.webp","json":"https://abxk.ai/en/physical-ai/learn/what-is-physical-ai/index.json","kind":"page","language":"en","learning_path":{"category":"physical-ai","level":0,"level_name":"Start here"},"modified":"2026-10-18","section":"physical-ai","seo_title":"What Is Physical AI? Machines That Sense and Act","tags":["physical-ai","robotics","autonomous-driving","beginner","physical-ai-basics"],"title":"What Is Physical AI? Machines That Sense, Decide and Act","translations":[{"json":"https://abxk.ai/de/physical-ai/learn/what-is-physical-ai/index.json","language":"de","url":"https://abxk.ai/de/physical-ai/learn/what-is-physical-ai/"}],"url":"https://abxk.ai/en/physical-ai/learn/what-is-physical-ai/"}