Drishtant Ghosh (Drix10)
drix10blogs
GitHubLinkedIn
AI Driven Vehicles and Transportation••6 min read•1178 words

🤖 AI Engineering - Physical AI Fundamentals

⚡Direct Technical Summary

Physical AI requires a deep understanding of low-level systems and mechanisms. Simply collecting data and using imitation learning is not enough to achieve true physical AI. Key P

🤖 AI Engineering - Physical AI Fundamentals

Physical AI requires a deep understanding of low-level systems and mechanisms. Simply collecting data and using imitation learning is not enough to achieve true physical AI.

Key Points:

  • Low-Level System Understanding: Physical AI requires a deep understanding of low-level systems and mechanisms, including sensors, actuators, and control systems.

  • Imitation Learning Limitations: Imitation learning alone is not sufficient to achieve true physical AI, as it does not provide a deep understanding of the underlying systems.

  • Real-World Experience: Physical AI requires real-world experience and experimentation to develop a deep understanding of the underlying systems and mechanisms.

🔗 Resources:

🚀 AI Engineering - Audi Vision Systems

13 years ago, my team at NVIDIA helped Audi develop vision systems for self-parking. The video itself still looks impressive, even today, considering the tiny compute power at the time.

Key Points:

  • Early Vision Systems: Audi's early vision systems for self-parking were developed 13 years ago, using NVIDIA's compute power.

  • CUDA Graph Optimization: The system required significant CUDA graph optimization to run efficiently.

  • Tiny Compute Power: The compute power at the time was tiny, but the system still managed to produce impressive results.

🔗 Resources:

🚀 AI Engineering - Gatik AI

$600M+ in contracted revenue. Fortune 50 customers like PepsiCo. Commercial operations across Texas, Arizona, Arkansas, and Canada. And now, following our $200M Series D, backing from Qatar Investment Authority, KDT, and Cathie Wood ARK Invest.

Key Points:

  • Gatik AI Revenue: Gatik AI has achieved $600M+ in contracted revenue.

  • Fortune 50 Customers: Gatik AI has customers like PepsiCo, a Fortune 50 company.

  • Commercial Operations: Gatik AI has commercial operations across multiple states and countries.

  • Series D Funding: Gatik AI has received $200M in Series D funding.

🔗 Resources:

🚀 AI Engineering - Flock Security

Free idea to fix Flock: encrypt all video and give residents control over unlocking it. Police can request bounded access per incident, but a majority of residents has to ok. Threshold cryptography / m-of-n authorization, maybe MPC so no single party ever holds the master key.

Key Points:

  • Flock Security: Flock's security can be improved by encrypting all video and giving residents control over unlocking it.

  • Threshold Cryptography: Threshold cryptography can be used to ensure that no single party holds the master key.

  • M-of-N Authorization: M-of-n authorization can be used to ensure that a majority of residents must ok before access is granted.

🔗 Resources:

🚀 AI Engineering - HOA Flock Requests

Maybe? Sounds like my last HOA to be honest where 2 retired guys would decide for the entire subdivision. Who has time to approve every flock request? In places where it's needed there might be multiple a day.

Key Points:

  • HOA Flock Requests: HOAs may struggle to approve every flock request, especially in places where it's needed multiple times a day.

  • Retired Decision-Makers: In some cases, retired individuals may be making decisions for the entire subdivision.

  • Time-Consuming Process: The process of approving every flock request can be time-consuming.

🔗 Resources:

🚀 AI Engineering - Nuro's First Decade

New Podcast Episode: This week we spoke with Dave Ferguson, Co-Founder and Co-CEO of @nuro, about Nuro's first decade and its partnership with Lucid and Uber to deploy driverless robotaxis.

Key Points:

  • Nuro's First Decade: Nuro's first decade has been marked by significant milestones, including its partnership with Lucid and Uber.

  • Driverless Robotaxis: Nuro has been working on deploying driverless robotaxis, a significant step towards autonomous driving.

  • Partnership with Lucid and Uber: Nuro has partnered with Lucid and Uber to deploy driverless robotaxis.

🔗 Resources:

🚀 AI Engineering - Nuro's Company Culture

“We were adamant that we wanted to build a company where the very best people in the world could come do the very best work of their career” Dave told The Driverless Digest.

Key Points:

  • Nuro's Company Culture: Nuro's company culture is focused on attracting the best talent and providing them with the opportunity to do their best work.

  • Dave Ferguson's Vision: Dave Ferguson, Co-Founder and Co-CEO of Nuro, has a clear vision for the company's culture.

🔗 Resources:

🚀 AI Engineering - The Road to Driverless Robotaxis

You can listen/watch the podcast here: Substack - https://thedriverlessdigest.com/p/theroad-to-driverless-robotaxis… ↗ Apple Podcasts - https://podcasts.apple.com/us/podcast/theroad-to-driverless-robotaxis-with-nuro-co-ceo/id1811181944?i=1000790959221… ↗ Spotify - https://open.spotify.com/episode/7udvKVBeBO4v3gYsZnxw6C?si=bV7Vde4fRG6Gc--fMP3rAw… ↗ Youtube - https://youtu.be/db9EUdSW6Co ↗

Key Points:

  • The Road to Driverless Robotaxis: The podcast episode explores Nuro's journey towards deploying driverless robotaxis.

  • Nuro's Co-CEO Dave Ferguson: Nuro's Co-CEO Dave Ferguson shares his insights on the company's journey.

🔗 Resources:

🚀 AI Engineering - Autonomous Driving Index

As the autonomous driving licensing market heats up, follow The @RoadToAutonomy Autonomous Driving Index. The only index following the sector. Updates every 12 hours.

Key Points:

  • Autonomous Driving Index: The Autonomous Driving Index tracks the autonomous driving licensing market.

  • Updates Every 12 Hours: The index is updated every 12 hours to reflect changes in the market.

🔗 Resources:

🚀 AI Engineering - Astra's Toy Toyota

For hours, Astra refused to consistently drive our toyota irl even though we told it it was in an empty lot, 7 mph cap, human foot on the brake etc. Telling it the whole thing was a "simulation" also failed, it would just look at the camera and realized it was real. Then we

Key Points:

  • Astra's Toy Toyota: Astra struggled to consistently drive a toy Toyota in a real-world setting.

  • Simulation Failure: Even when told it was a simulation, Astra failed to understand the difference between reality and simulation.

  • Real-World Experience: Astra's performance in real-world settings is still a work in progress.

🔗 Resources:

📂Source / Implementation:AI Driven Vehicles and Transportation / resources-250.md
GitHub Repository↗

Related AI Driven Vehicles and Transportation Breakdowns

🚀 Autonomous Era - Building an American Industrial Base
7m readRead ↗
🚀 Autonomous Vehicles - AI-Powered Coaching for Driver Safety
5m readRead ↗
🚀 Tech News - Twitter Threads
6m readRead ↗
🤖 AI and Human Collaboration
4m readRead ↗
Drishtant Ghosh (Drix10)
Drishtant Ghosh (Drix10)•Author & Engineer

Technical founder and engineer working across AI systems, developer infrastructure, and cybersecurity.

Portfolio·GitHub·LinkedIn·X·Email
← Back to all breakdowns
GitHub: AI Driven Vehicles and Transportation/resources-250.md ↗https://blogs.drix10.com/articles/ai-driven-vehicles-and-transportation/ai-engineering-physical-ai-fundamentals-250