Physical AI that Moves the World — Qasar Younis & Peter Ludwig, Applied Intuition
Summary
Applied Intuition, co-founded by Peter Ludwig and Kasar Younis, focuses on "physical AI" for various moving systems, including cars, trucks, construction, mining, agriculture, and defense. The company's mission is to build technology for a safer, more prosperous world, deploying intelligence in safety-critical environments. Unlike large language models that operate on screens, Applied Intuition's solutions are for physical machines, exemplified by their L4 driverless trucks in Japan. The company, which has over 30 products and 1,000 engineers (83% of its workforce), evolved from initial work in autonomy, simulation, and data infrastructure. They serve 18 of the top 20 global non-Chinese automakers and government entities, providing a spectrum of solutions from development suites to full technology stacks, positioning themselves as a technology provider akin to Nvidia or AMD but without silicon.
Key takeaway
For AI Engineers and ML Architects developing systems for physical machines, prioritize robust, low-latency operating systems and highly efficient models for on-board deployment. Your focus should extend beyond model intelligence to the practical constraints of hardware, power, and real-time performance in safety-critical environments. Invest in statistical verification and validation processes, as this is paramount for regulatory acceptance and public trust, especially given the industry's shift from binary requirements to reliability metrics.
Key insights
Applied Intuition builds physical AI for safety-critical moving machines, focusing on simulation, operating systems, and fundamental AI models.
Principles
- Physical AI requires robust, low-latency operating systems.
- Statistical verification is crucial for AI in safety-critical systems.
- Compounding technology drives long-term value in hard tech.
Method
Applied Intuition's approach involves a three-bucket strategy: simulation and data infrastructure (including neural simulation and reinforcement learning), true operating systems for embedded control, and fundamental AI models for autonomy and human-machine teaming.
In practice
- Utilize simulation for virtual development and real-world correlation.
- Implement on-board (embedded) and off-board (data center) AI deployments.
- Adopt AI agents for configuring complex systems like sensor suites.
Topics
- Physical AI
- Autonomous Systems
- AI Operating Systems
- Simulation & Verification
- Embedded Machine Learning
Best for: AI Engineer, Machine Learning Engineer, AI Architect
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Latent.Space - Www.latent.space.