Ghost Autonomy develops autonomous driving software intended for integration with automaker partners. The company is piloting multimodal LLM-based systems to perform higher-complexity scene interpretation and to suggest driving decisions from camera images. Ghost says it will fine-tune existing models and train application-specific models to improve reliability and handle long-tail situations such as construction zones and unusual road users. The company is testing LLM-driven decision making via its development fleet and working with automakers to jointly validate and integrate new models into its autonomy stack. Ghost partnered with OpenAI through the OpenAI Startup Fund to gain early access to OpenAI systems and, via that relationship, access to Azure resources from Microsoft. As part of that partnership Ghost received a $5 million investment from the OpenAI Startup Fund. Company leadership argues multimodal models will add reasoning and broad-based world knowledge to autonomy systems, while outside experts quoted in the article express skepticism about applying LLMs to driving. Ghost is developing a universal collision-avoidance system for autonomous driving that shifts away from classification-first approaches. Its technology tracks clusters of pixels to detect any mass, estimate distance and relative velocity, and make avoidance decisions before objects are classified. Ghost argues this method reduces uncertainty and requires less computational power, which benefits fuel efficiency and electric vehicle range. The company has conducted most testing off-road using physical obstacles and augmented reality and has not yet tested in urban environments. Public highway testing with a human safety driver is planned to begin this year and scale up next year. Ghost previously planned a consumer retrofit kit for highway autonomy but is also pursuing direct partnerships with automakers; it recently hired Jacqueline Glassman, former NHTSA Chief Counsel and Acting Administrator, as General Counsel. Ghost Locomotion develops an aftermarket hardware and software kit that lets privately owned cars drive autonomously on highways. The kit includes sensors and a computer installed in the trunk, connects to the vehicle CAN bus, and requires electronic steering; cameras are mounted in locations such as a license-plate-style rear unit, behind the rearview mirror, and around the door window frame. Ghost says it will deliver the product in 2020 and prices will be below Tesla’s $7,000 Full Self-Driving package. The company is starting with the 20 most popular car brands as an aftermarket product, plans retail installation spaces, and expects the kit to eventually be integrated into cars by manufacturers. Ghost employs about 75 people and has already installed kits on the cars of high-mileage drivers (including Uber and Lyft drivers and commuters) to collect video and driving data. It plans to have kits in hundreds of cars by the end of the year and thousands by the next year to build training data.

Total raised

$169M

Funding rounds

3

Latest round

Equity

Latest activity

Nov 2023