Ghost Autonomy is a technology company specializing in the development of advanced software solutions for autonomous driving. Their focus is on creating a robust and scalable autonomy stack that can be integrated into various vehicle platforms to enable self-driving capabilities.

At Ghost, our mission is to make self-driving for everyone. We build autonomous driving software for automakers, based on a breakthrough in artificial intelligence that finally makes highway autonomy safe and scalable for the consumer car market.

Ghost was founded in 2017 by John Hayes and Volkmar Uhlig. Before Ghost, John co-founded Pure Storage, taking the company public in 2015. 

Ghost is backed by investors including Mike Speiser at Sutter Hill Ventures, Keith Rabois at Founders Fund, and Vinod Khosla at Khosla Ventures.

Sutter Hill Ventures

Khosla Ventures

Founders Fund

Coatue

OpenAI Startup Fund

Ghost Autonomy

Imagine a future where your car has an open API that you, or anyone can program to?  This is the world Ghost CEO John Hayes envisions in a recently-published article on Forbes.  Software-defined vehicle, to the extreme!

https://lnkd.in/g6PKnNxs

LLMs are disrupting everything - and we think autonomy is next.  Multi-Modal LLMs (MLLMs) have the potential power to reason about driving scenes holistically, combining perception and planning to provide autonomous vehicles deeper scene understanding and guidance on the correct driving maneuver by considering the scene in totality.

In this video we share a glimpse at the reasoning capabilities of MLLMs to understand intersections, construction scenes, and urban environments.  

At Ghost Autonomy we're optimizing LLMs for autonomy and building a platform that makes it easy for automakers to bring large models to the cars of tomorrow.  Read more here: https://lnkd.in/g-Htfrcm

We’re excited to announce that OpenAI Startup Fund has invested in Ghost Autonomy to bring Multi-Modal LLMs to autonomous driving.

We believe that MLLMs will have a profound impact on the autonomy software stack, with the potential power to reason about driving scenes holistically to provide AVs deeper understanding and guidance on the correct driving maneuver. We're optimizing MLLMs for autonomy and building a platform that enables automakers to bring large models to the cars of tomorrow.
https://lnkd.in/gGtrFdr4

Watch some examples of MLLMs reasoning about intersections, understanding construction scenes, and interpreting urban road environments in our deeper-dive blog: https://lnkd.in/g-Htfrcm

Ghost CEO John Hayes op-ed on the software applications that will define the next-generation of winners in automotive.

"Fusion" is where various sensor outputs (mono vision, stereo vision, and radar) are brought together to create a unified view of obstacles and the scene.  This step is critical to smooth and safe driving, as it's where input differences are disambiguated, where noise and false positives/negatives are corrected, and where resiliency is achieved in the perception layer.  In this episode of Talking Autonomy, Model Engineer Prannay Khosla walks through how Ghost implements late fusion techniques for resilient perception.

Accident reports for AVs in California suggest a majority of the collisions were caused by software issues, either misinterpreting the surroundings or being too rigid, causing the vehicles to act in unexpected ways. 

AV systems need to be designed differently with smarter software enabled by generalized AI to drive on roads they’ve never seen before.

Read more from Kevin Jost at Inside Autonomous Vehicles about how at Ghost we are developing a new software stack for autonomy, designed to generalize and scale to every car on every highway. ⬇

https://lnkd.in/ehxW4RbT 

#AV #AI #autonomousvehicles #softwaredefinedvehicles

"Any very complex computer science problem is being subject to AI, and with AI you almost always get better performance."

John Hayes presentation at CAR MBS on developing reliable and robust AI for autonomous driving. 

Video and full transcript here:
https://lnkd.in/gMArXexd

Automakers tell us that one of the biggest challenges with current L2+/L3 systems is HD Maps.  Fragile, inaccurate, constantly out-of-date, and hard to localize to given the fidelity of GPS.  

At Ghost, we asked ourselves - "if humans don't need HD Maps to drive, why should AVs?"  In this episode of Talking Autonomy, Arunachalam Narayanaswamy explains how Ghost sees and understands the road with vision and neural networks alone.

Ghost continues to validate generalization by testing Ghost in more and more places.  In this video you'll see Ghost's first drive in Canada, taking Highway 401 for 132 autonomous miles from Detroit, MI to Toronto, ON!

COM:GHOSTAUTONOMY

Ghost Autonomy

Analysis Not Available